State of AI Report 2022 - OnLINE

stateof.
ai #stateofai
State of AI Report
October 11, 2022
Nathan Benaich Ian Hogarth

Introduction | Research | Industry | Politics | Safety | Predictions #stateofai | 2
About the authors

Nathan is the General Partner of Air Street Capital, a Ian is a co-founder at Plural, an investment platform for
venture capital firm investing in AI-first technology and experienced founders to help the most ambitious
life science companies. He founded RAAIS and European startups. He is a Visiting Professor at UCL
London.AI (AI community for industry and research), the working with Professor Mariana Mazzucato. Ian was co-
RAAIS Foundation (funding open-source AI projects), founder and CEO of Songkick, the concert service. He
and Spinout.fyi (improving university spinout creation). started studying machine learning in 2005 where his
He studied biology at Williams College and earned a PhD Masters project was a computer vision system to classify
from Cambridge in cancer research. breast cancer biopsy images.
stateof.ai 2022
Artificial intelligence (AI) is a multidisciplinary field of science and engineering whose goal is to create intelligent machines.
We believe that AI will be a force multiplier on technological progress in our increasingly digital, data-driven world. This is because
everything around us today, ranging from culture to consumer products, is a product of intelligence.
The State of AI Report is now in its fifth year. Consider this report as a compilation of the most interesting things we’ve seen with a
goal of triggering an informed conversation about the state of AI and its implication for the future.
We consider the following key dimensions in our report:

- Research: Technology breakthroughs and their capabilities.
- Industry: Areas of commercial application for AI and its business impact.
- Politics: Regulation of AI, its economic implications and the evolving geopolitics of AI.
- Safety: Identifying and mitigating catastrophic risks that highly-capable future AI systems could pose to us.
- Predictions: What we believe will happen in the next 12 months and a 2021 performance review to keep us honest.
Produced by Nathan Benaich (@nathanbenaich), Ian Hogarth (@soundboy), Othmane Sebbouh (@osebbouh) and
Nitarshan Rajkumar (@nitarshan).
stateof.ai 2022
Thank you!
Othmane Sebbouh Nitarshan Rajkumar
Research Assistant Research Assistant

Othmane is a PhD student in ML at Nitarshan is a PhD student in AI at the
ENS Paris, CREST-ENSAE and University of Cambridge. He was a
CNRS. He holds an MsC in research student at Mila and a software
management from ESSEC Business engineer at Airbnb. He holds a BSc
School and a Master in Applied from University of Waterloo.
Mathematics from ENSAE and Ecole
Polytechnique.
stateof.ai 2022
Definitions
Artificial intelligence (AI): a broad discipline with the goal of creating intelligent machines, as opposed to the natural intelligence
that is demonstrated by humans and animals.
Artificial general intelligence (AGI): a term used to describe future machines that could match and then exceed the full range of
human cognitive ability across all economically valuable tasks.
AI Safety: a field that studies and attempts to mitigate the catastrophic risks which future AI could pose to humanity.
Machine learning (ML): a subset of AI that often uses statistical techniques to give machines the ability to "learn" from data
without being explicitly given the instructions for how to do so. This process is known as “training” a “model” using a learning
“algorithm” that progressively improves model performance on a specific task.
Reinforcement learning (RL): an area of ML in which software agents learn goal-oriented behavior by trial and error in an
environment that provides rewards or penalties in response to their actions (called a “policy”) towards achieving that goal.
Deep learning (DL): an area of ML that attempts to mimic the activity in layers of neurons in the brain to learn how to recognise
complex patterns in data. The “deep” refers to the large number of layers of neurons in contemporary models that help to learn rich
representations of data to achieve better performance gains.
stateof.ai 2022
Definitions
Model: once a ML algorithm has been trained on data, the output of the process is known as the model. This can then be used to
make predictions.
Self-supervised learning (SSL): a form of unsupervised learning, where manually labeled data is not needed. Raw data is instead
modified in an automated way to create artificial labels to learn from. An example of SSL is learning to complete text by masking
random words in a sentence and trying to predict the missing ones.
(Large) Language model (LM, LLM): a model trained on textual data. The most common use case of a LM is text generation. The
term “LLM” is used to designate multi-billion parameter LMs, but this is a moving definition.
Computer vision (CV): enabling machines to analyse, understand and manipulate images and video.
Transformer: a model architecture at the core of most state of the art (SOTA) ML research. It is composed of multiple “attention”
layers which learn which parts of the input data are the most important for a given task. Transformers started in language modeling,
then expanded into computer vision, audio, and other modalities.
stateof.ai 2022
Executive Summary
Research
- Diffusion models took the computer vision world by storm with impressive text-to-image generation capabilities.
- AI attacks more science problems, ranging from plastic recycling, nuclear fusion reactor control, and natural product discovery.
- Scaling laws refocus on data: perhaps model scale is not all that you need. Progress towards a single model to rule them all.
- Community-driven open sourcing of large models happens at breakneck speed, empowering collectives to compete with large labs.
- Inspired by neuroscience, AI research are starting to look like cognitive science in its approaches.
Industry
- Have upstart AI semiconductor startups made a dent vs. NVIDIA? Usage statistics in AI research shows NVIDIA ahead by 20-100x.
- Big tech companies expand their AI clouds and form large partnerships with A(G)I startups.
- Hiring freezes and the disbanding of AI labs precipitates the formation of many startups from giants including DeepMind and OpenAI.
- Major AI drug discovery companies have 18 clinical assets and the first CE mark is awarded for autonomous medical imaging diagnostics.
- The latest in AI for code research is quickly translated by big tech and startups into commercial developer tools.
Politics
- The chasm between academia and industry in large scale AI work is potentially beyond repair: almost 0% of work is done in academia.
- Academia is passing the baton to decentralized research collectives funded by non-traditional sources.
- The Great Reshoring of American semiconductor capabilities is kicked off in earnest, but geopolitical tensions are sky high.
- AI continues to be infused into a greater number of defense product categories and defense AI startups receive even more funding.
Safety
- AI Safety research is seeing increased awareness, talent, and funding, but is still far behind that of capabilities research.
stateof.ai 2022
Scorecard: Reviewing our predictions from 2021
stateof.ai 2022
Our 2021 Prediction Grad Evidence
e
DeepMind’s Gato model makes progress in this direction in which a transformer
Transformers replace RNNs to learn world models with which RL
Yes predicts the next state and action, but it is not trained with RL. University of Geneva’s
agents surpass human performance in large and rich games.
GPT-like transformer model IRIS solves tasks in Atari environments.
ASML’s market cap reaches $500B. No Current market cap is circa $165B (3 Oct 2022)
Anthropic publishes on the level of GPT, Dota, AlphaGo to establish

No Not yet.
itself as a third pole of AGI research.
A wave of consolidation in AI semiconductors with at least one of

Graphcore, Cerebras, SambaNova, Groq, or Mythic being acquired by No No new announced AI semiconductor consolidation has happened yet.
a large technology company or major semiconductor incumbent.
Small transformers + CNN hybrid models match current SOTA on

MaxViT from Google with 475M parameters almost matched (89.53%) CoAtNet-7’s
ImageNet top-1 accuracy (CoAtNet-7, 90.88%, 2.44B params) with Yes
performance (90.88%) on ImageNet top-1 accuracy.
10x fewer parameters.
DeepMind shows a major breakthrough in the physical sciences. Yes Three (!) DeepMind papers in mathematics and material science.
The JAX framework grows from 1% to 5% of monthly repos created

No JAX usage still accounts for <1% of monthly repos on Papers With Code.
as measured by Papers With Code.
A new AGI-focused research company is formed with significant

Adept.ai was co-founded by the authors of the Transformer and is focused on AGI via
backing and a roadmap that’s focused on a sector vertical (e.g. Yes
software tool use automation.
developer tools, life science).
Bonus! Predictions, revisited - better late than never!

Year Prediction Grade Evidence
Access to Taiwanese and South Korean semiconductor companies US CHIPS Act 2022 prevents recipients to expand
2018 Yes
becomes an explicit part of the trade war between US and China. operations in China. TSMC caught in the crosshairs.
The government of an OECD country blocks the acquisition of a The UK, amongst others, blocked the acquisition of Arm
2018 Yes
leading ML company by a US or Chinese HQ'd tech company. by NVIDIA.
As AI systems become more powerful, governance of AI becomes a

2019 bigger topic and at least one major AI company makes a substantial Yes Anthropic set up as a public benefit corporation.
change to their governance model.
Facebook/Meta makes a major breakthrough in AR/VR with 3D

2020 Sort of Implicitron in PyTorch3D. Not applied to AR/VR yet.
computer vision.
Chinese and European defense-focused AI startups collectively raise

2020 Yes Helsing (Germany) raised $100M Series A in 2022.
>$100M in the next 12 months.
2020 NVIDIA does not end up completing its acquisition of Arm. Yes Deal is formally cancelled in 2022.
stateof.ai 2022
Section 1: Research
stateof.ai 2022
2021 Prediction: DeepMind’s breakthroughs in the physical sciences (1/3)

In 2021, we predicted: “DeepMind releases a major research breakthrough in the physical sciences.” The
company has since made significant advancements in both mathematics and materials science.
● One of the decisive moments in mathematics is formulating a
conjecture, or a hypothesis, on the relationship between
variables of interest. This is often done by observing a large
number of instances of the values of these variables, and
potentially using data-driven conjecture generation methods.
But these are limited to low-dimensional, linear, and generally
simple mathematical objects.
● In a Nature article, DeepMind researchers proposed an iterative workflow involving mathematicians and a supervised
ML model (typically a NN). Mathematicians hypothesize a function relating two variables (input X(z) and output
Y(z)). A computer generates a large number of instances of the variables and a NN is fit to the data. Gradient saliency
methods are used to determine the most relevant inputs in X(z). Mathematicians can turn refine their hypothesis
and/or generate more data until the conjecture holds on a large amount of data.
stateof.ai 2022

● DeepMind researchers used their framework in a collaboration with
mathematics professors from the University of Sydney and the University of
Oxford to (i) propose an algorithm that could solve a 40 years-long standing
conjecture in representation theory and (ii) prove a new theorem in the study of
knots.
● DeepMind made an important contribution in materials science as
well. It showed that the exact functional in Density Functional Theory,
an essential tool to compute electronic energies, can be efficiently
approximated using a neural network. Notably, instead of constraining
the neural network to verify mathematical constraints of the DFT
functional, researchers simply incorporate them into the training data
to which they fit the NN.
stateof.ai 2022

● DeepMind repurposed AlphaZero (their RL model trained to beat the best
human players of Chess, Go and Shogi) to do matrix multiplication. This
AlphaTensor model was able to find new deterministic algorithms to multiply
two matrices. To use AlphaZero, the researchers recast the matrix multiplication
problem as a single-player game where each move corresponds to an algorithm
instruction and the goal is to zero-out a tensor measuring how far from correct
the predicted algorithm is.
● Finding faster matrix multiplication algorithms, a seemingly simple and well-
studied problem, has been stale for decades. DeepMind’s approach not only
helps speed up research in the field, but also boosts matrix multiplication based
technology, that is AI, imaging, and essentially everything happening on our
phones.
stateof.ai 2022
Reinforcement learning could be a core component of the next fusion breakthrough

DeepMind trained a reinforcement learning system to adjust the magnetic coils of Lausanne’s TCV (Variable
Configuration tokamak). The system’s flexibility means it could also be used in ITER, the promising next
generation tokamak under construction in France.
● A popular route to achieving nuclear fusion requires
confining extremely hot plasma for enough time using a
tokamak.
● A major obstacle is that the plasma is unstable, loses
heat and degrades materials when it touches the
tokamak’s walls. Stabilizing it requires tuning the
magnetic coils thousands of times per second.
● DeepMind’s deep RL system did just that: first in a
simulated environment and then when deployed in the
TCV in Lausanne. The system was also able to shape
the plasma in new ways, including making it
compatible with ITER’s design.
stateof.ai 2022
Predicting the structure of the entire known proteome: what could this unlock next?
Since its open sourcing, DeepMind’s AlphaFold 2 has been used in hundreds of research papers. The company
has now deployed the system to predict the 3D structure of 200 million known proteins from plants, bacteria,
animals and other organisms. The extent of the downstream breakthroughs enabled by this technology - ranging
from drug discovery to basic science - will need a few years to materialize.
● There are 190k empirically determined 3D structures in the Protein Data
Bank today. These have been derived through X-Ray crystallography and
cryogenic electron microscopy.
● The first release of AlphaFold DB in July 2021 included 1M predicted
protein structures.
● This new release 200x’s the database size. Over 500,000 researchers from
190 countries have made use of the database.
● AlphaFold mentions in AI research literature is growing massively and is
predicted to triple year on year (right chart).
stateof.ai 2022
Language models for proteins: a familiar story of open source and scaled models
Researchers independently applied language models to the problems of protein generation and structure
prediction while scaling model parameter. They both report large benefits from scaling their models.
● Salesforce researchers find that scaling their LMs allows them to better capture the training
distribution of protein sequences (as measured by perplexity).
● Using the 6B param ProGen2, they generated proteins with similar folds to natural proteins,
but with a substantially different sequence identity. But to unlock the full potential of scale,
the authors insist that more emphasis be placed on data distribution.
● Meta et al. introduced the ESM family of protein LMs, whose sizes range from 8M to 15B
(dubbed ESM-2) parameters. Using ESM-2, they build ESMFold to predict protein structure.
They show that ESMFold produces similar predictions to AlphaFold 2 and RoseTTAFold, but
is an order of magnitude faster.
● This is because ESMFold doesn’t rely on the use of multiple
sequence alignments (MSA) and templates like AlphaFold 2 and
RoseTTAFold, and instead only uses protein sequences.
stateof.ai 2022
OpenCell: understanding protein localization with a little help from machine learning
Researchers used CRISPR-based endogenous tagging — modifying genes by illuminating specific aspects of the
proteins’ function — to determine protein localization in cells. They then used clustering algorithms to identify
protein communities and formulate mechanistic hypotheses on uncharacterized proteins.
● An important goal of genomic research is to understand where proteins localize and how
they interact in a cell to enable particular functions. With its dataset of 1,310 tagged
proteins across ~5,900 3D images, the OpenCell initiative enabled researchers to draw
important links between spatial distribution of proteins, their functions, and their
interactions.
● Markov clustering on the graph of protein interactions successfully delineated
functionally related proteins. This will help researchers better understand so-far
uncharacterized proteins.
● We often expect ML to deliver definitive predictions. But here as with math, ML first
gives partial answers (here clusters), humans then interpret, formulate and test
hypotheses, before delivering a definitive answer.
stateof.ai 2022
Plastic recycling gets a much-needed ML-engineered enzyme

Researchers from UT Austin engineered an enzyme capable of degrading PET, a type of plastic responsible for
12% of global solid waste.
● The PET hydrolase, called FAST-PETase, is more
robust to different temperatures and pH levels than
existing ones.
● FAST-PETase was able to almost completely degrade
51 different products in 1 week.
● They also showed that they
could resynthesize PET from
monomers recovered from
FAST-PETase degradation,
potentially opening the way
for industrial scale closed-loop
PET recycling.
stateof.ai 2022
Beware of compounded errors: in science, ML in and garbage out?

With the increased use of ML in quantitative sciences, methodological errors in ML can leak to these disciplines.
Researchers from Princeton warn of a growing reproducibility crisis in ML-based science driven in part by one
such methodological error: data leakage.
● Data leakage is an umbrella term covering all cases where data that
shouldn’t be available to a model in fact is. The most common
example is when test data is included in the training set. But the
leakage can be more pernicious: when the model uses features that
are a proxy of the outcome variable or when test data come from a
distribution which is different from the one about which the
scientific claim is made.
● The authors argue that the ensuing reproducibility failures in ML-based science are systemic: they study 20 reviews
across 17 science fields examining errors in ML-based science and find that data leakage errors happened in every one
of the 329 papers the reviews span. Inspired by the increasingly popular model cards in ML, the authors propose that
researchers use model info sheets designed to prevent data leakage issues.
stateof.ai 2022
OpenAI uses Minecraft as a testbed for computer-using agents

OpenAI trained a model (Video PreTraining, VPT) to play Minecraft from video frames using a small amount of
labeled mouse and keyboard interactions. VPT is the first ML model to learn to craft diamonds, “a task that
usually takes proficient humans over 20 minutes (24,000 actions)”.
● OpenAI gathered 2,000 hours of video labeled with
mouse and keyboard actions and trained an inverse
dynamics model (IDM) to predict actions given past
and future frames – this is the PreTraining part.
● They then used the IDM to label 70K hours of video
on which they trained a model to predict actions given
only past video frames.
● They show that the model can be fine-tuned with
imitation learning and reinforcement learning (RL) to
achieve a performance which is too hard to reach
using RL from scratch.
stateof.ai 2022
Corporate AI labs rush into AI for code research

OpenAI’s Codex, which drives GitHub Copilot, has impressed the computer science community with its ability to
complete code on multiple lines or directly from natural language instructions. This success spurred more
research in this space, including from Salesforce, Google and DeepMind.
● With the conversational CodeGen, Salesforce researchers leverage the language
understanding of LLMs to specify coding requirements in multiturn language
interactions. It is the only open source model to be competitive with Codex.
● A more impressive feat was achieved by Google’s LLM PaLM, which achieves a similar
performance to Codex, but with 50x less code in its training data (PaLM was trained on a
larger non-code dataset). When fine-tuned on Python code, PaLM outperformed (82% vs.
71.7% SOTA) peers on Deepfix, a code repair task.
● DeepMind’s AlphaCode tackles a different problem: the generation of whole programs on
competitive programming tasks. It ranked in the top half on Codeforces, a coding
competitions platform. It was pre-trained on GitHub data and fine-tuned on Codeforces
problems and solutions. Millions of possible solutions are then sampled, filtered, and
clustered to obtain 10 final candidate submissions.
stateof.ai 2022
Five years after the Transformer, there must be some efficient alternative, right… right?
The attention layer at the core of the transformer model famously suffers from a quadratic dependence on its
input. A slew of papers promised to solve this, but no method has been adopted. SOTA LLMs come in different
flavors (autoencoding, autoregressive, encoder-decoders), yet all rely on the same attention mechanism.
● A Googol of transformers have been trained over the past few years,
costing millions (billions?) to labs and companies around the world. But
so-called “Efficient Transformers” are nowhere to be found in large-scale
LM research (where they would make the biggest difference!). GPT-3,
PaLM, LaMDA, Gopher, OPT, BLOOM, GPT-Neo, Megatron-Turing
NLG, GLM-130B, etc. all use the original attention layer in their
transformers.
● Several reasons can explain this lack of adoption: (i) the potential linear
speed-up is only useful for large input sequences, (ii) the new methods
introduce additional constraints that make the architectures less universal,
(iii) the reported efficiency measures don’t translate in actual
computational cost and time savings.
stateof.ai 2022
Mathematical abilities of Language Models largely surpass expectations

Built on Google’s 540B parameter LM PaLM, Google’s Minerva achieves a 50.3% score on the MATH
benchmark (43.4 pct points better than previous SOTA), beating forecasters expectations for best score in 2022
(13%). Meanwhile, OpenAI trained a network to solve two mathematical olympiad problems (IMO).
● Google trained its (pre-trained) LLM PaLM on an additional 118GB dataset of scientific
papers from arXiv and web pages using LaTeX and MathJax. Using chain of thought
prompting (including intermediate reasoning steps in prompts rather than the final answer
only) and other techniques like majority voting, Minerva improves the SOTA on most
datasets by at least double digit pct points.
● Minerva only uses a language model and doesn’t explicitly encode formal mathematics. It
is more flexible but can only be automatically evaluated on its final answer rather than its
whole reasoning, which might justify some score inflation.
● In contrast, OpenAI built a (transformer-based) theorem prover built in the Lean formal
environment. Different versions of their model were able to solve a number of problems
from AMC12 (26), AIME (6) and IMO (2) (increasing order of difficulty).
stateof.ai 2022
Fast progress in LLM research renders benchmarks obsolete, but a BIG one comes to
help
Only 66% of machine learning benchmarks have received more than 3 results at different time points, and many
are solved or saturated soon after their release. BIG (Beyond the Imitation Game), a new benchmark designed
by 444 authors across 132 institutions, aims to challenge current and future language models.
● A study from the University of Vienna, Oxford, and FHI
examined 1,688 benchmarks for 406 AI tasks and identified
different submission dynamics (see right).
● They note that language benchmarks in particular tend to
be quickly saturated.
● Rapid LLM progress and emerging capabilities seem to outrun current benchmarks. As a result, much of this progress
is only captured through circumstantial evidence like demos or one-off breakthroughs, and/or evaluated on disparate
dedicated benchmarks, making it difficult to identify actual progress.
● The new BIG benchmark contains 204 tasks, all with strong human expert baselines, which evaluate a large set of
LLM capabilities from memorization to multi-step reasoning. They show that, for now, even the best models perform
poorly on the BIG benchmark.
stateof.ai 2022
Ducking language model scaling laws: more data please

DeepMind revisited LM scaling laws and found that current LMs are significantly undertrained: they’re not
trained on enough data given their large size. They train Chinchilla, a 4x smaller version of their Gopher, on 4.6x
more data, and find that Chinchilla outperforms Gopher and other large models on BIG-bench.
● Empirical LM scaling laws determine, for a fixed compute budget, the
model and training data sizes that should be used. Past work from
OpenAI had established that model size should increase faster than
training data size as the compute budget increases.
● DeepMind claims that the model size and the number of training tokens
should instead increase at roughly the same rate.
● Compared to OpenAI’s work, DeepMind uses larger models to derive
their scaling laws. They emphasize that data scaling leads to better
predictions from multibillion parameter models.
● Following these new scaling laws, Chinchilla (70B params) is trained on 1.4T tokens. Gopher (230B) on 300B.
● Though trained with the same compute budget, the lighter Chinchilla should be faster to run.
stateof.ai 2022
Ducking language model scaling laws: emergence

While model loss can be reasonably predicted as a function of size and compute using well-calibrated scaling
laws, many LLM capabilities emerge unpredictably when models reach a critical size. These acquired capabilities
are exciting, but the emergence phenomenon makes evaluating model safety more difficult.
Arithmetics Fig. of speech
● Emergence is not fully understood: it could be that for multi-step reasoning tasks,
models need to be deeper to encode the reasoning steps. For memorization tasks, having
more parameters is a natural solution. The metrics themselves may be part of the
explanation, as an answer on a reasoning task is only considered correct if its conclusion
is. Thus despite continuous improvements with model size, we only consider a model
successful when increments accumulate past a certain point. Multi-task NLU Transliteration
● A possible consequence of emergence is that there are a range of tasks that are out of
reach of current LLMs that could soon be successfully tackled.
● Alternatively, deploying LLMs on real-world tasks at larger scales is more uncertain as
unsafe and undesirable abilities can emerge. Alongside the brittle nature of ML models,
this is another feature practitioners will need to account for. Training
FLOPs
stateof.ai 2022
Teach a machine to fish: tool use as the next frontier?

Language models can learn to use tools such as search engines and calculators, simply by making available text
interfaces to these tools and training on a very small number of human demonstrations.
● OpenAI’s WebGPT was the first model to demonstrate this convincingly by
fine-tuning GPT-3 to interact with a search engine to provide answers
grounded with references. This merely required collecting data of humans
doing this task and converting the interaction data into text that the model
could consume for training by standard supervised learning. Importantly, the
use of increasing amounts of human demonstration data significantly
increased the truthfulness and informativeness of answers (right panel, white
bars for WebGPT), a significant advance from when we covered truthfulness
evaluation in our 2021 report (slide 44).
● Adept, a new AGI company, is commercializing this paradigm. The company
trains large transformer models to interact with websites, software applications
and APIs (see more at adept.ai/act) in order to drive workflow productivity.
stateof.ai 2022
Looking back: three eras of compute in machine learning

A study documents the incredible acceleration of compute requirements in machine learning. It identifies 3 eras
of machine learning according to training compute per model doubling time. The Pre-Deep Learning Era (pre-
2010, training compute doubled every 20 months), the Deep Learning Era (2010-15, doubling every 6 months),
and the Large-Scale Era (2016-present, a 100-1000x jump, then doubling every 10 months).
stateof.ai 2022
Diffusion models take over text-to-image generation and expand into other modalities
When we covered diffusion models in the 2021 Report (slide 36), they were overtaking GANs in image generation
on a few benchmarks. Today, they are now the undisputable SOTA for text-to-image generation, and are
diffusing (pun intended) into text-to-video, text generation, audio, molecular design and more.
● Diffusion models (DMs) learn to reverse successive noise additions to images by modeling
the inverse distribution (generating denoised images from noisy ones) at each step as a
Gaussian whose mean and covariance are parametrized as a neural network. DMs generate
new images from random noise.
● Sequential denoising makes them slow, but new techniques (like denoising in a lower-
dimensional space) allow them to be faster at inference time and to generate higher-quality
samples (classifier-free guidance – trading off diversity for fidelity).
● SOTA text-to-image models like DALL-E 2, Imagen and Stable Diffusion are based on DMs.
They’re also used in controllable text generation (generating text with a pre-defined structure
or semantic context), model-based reinforcement learning, video generation and even
molecular generation.
stateof.ai 2022
DALL-E 2, Imagen and Parti…the battle for text-to-image generation rages

The second iteration of OpenAI’s DALL-E, released in April 2022, came with a significant jump in the quality of
generated images. Soon after, another at least equally impressive diffusion-based model came from Google
(Imagen). Meanwhile, Google’s Parti took a different, autoregressive, route.
DALL-E Imagen Parti-350M Parti-20B
2
● Instead of using a diffusion model, Parti treats text-to-image generation as a simple sequence-to-sequence task, where
the sequence to be predicted is a representation of the pixels of the image. Notably, as the number of parameters and
training data in Parti are scaled, the model acquires new abilities like spelling.
● Other impressive text-to-image models include GLIDE (OpenAI) and Make-a-Scene (Meta — can use both text and
sketches), which predate DALL-E 2, and CogView2 (Tsinghua, BAAI — both English and Chinese).
stateof.ai 2022
The text-to-image diffusion model frenzy gives birth to new AI labs

Stability.ai and Midjourney came out of seemingly nowhere with text-to-image models that rival those of
established AI labs. Both have APIs in beta, Midjourney is reportedly profitable, and Stability has already open-
sourced their model. But more on their emergence and research dynamics in our Politics section.
Image Credits: Fabian Stelzer stateof.ai 2022

The text-to-video generation race has started

Research on diffusion-based text-to-video generation was kicked-off around April 2022, with work from Google
and the University of British Columbia. But in late September, new research from Meta and Google came with a
jump in quality, announcing a sooner-than-expected DALL-E moment for text-to-video generation.
Make-a-Video Imagen Video Phenaki
● Meta made the first splash from Big Tech in text-to-video generation by releasing Make-a-Video, a diffusion model
for video generation.
● In an eerily similar fashion to text-to-image generation, Google then published (less than a week later) almost
simultaneously two models: one diffusion-model based, Imagen, and another non diffusion-model based, Phenaki.
The latter can dynamically adapt the video via additional prompts.
stateof.ai 2022
Closed for 14 months: community-driven open sourcing of GPT et al.

Landmark models from OpenAI and DeepMind have been implemented/cloned/improved by the open source
community much faster than we’d have expected.
GPT-NeoX (20B)
GPT-j (6B) Megatron
Turing-NLG (137B) BLOOM
Chinchilla
(70B) (176B)
GPT-3 Pan-Gu (200B) FLAN (137B) Gopher (280B) OPT (175B)
(175B)
Jan 2022 Aug 2022
June 2020 May 2021 Aug 2021 Sep 2021 May 2022
HyperCLOVA (204B) Yuan 1.0 (246B) LaMDA (280B) GLM (130B)
PaLM (540B)
Jurassic-1 Jumbo (204B)
Ernie 3.0 Titan (260B)

Open-sourced models in red stateof.ai 2022
Closed for 15 months: community-driven open sourcing of DALL-E et al.

DALL-E Make-a-scene DALL-E 2 Imagen Parti Stable Diffusion
Apr 2022
Jan 2021 Mar 2022 May 2022 June 2022 July 2022 Aug 2022
DALL-E mini
CogView2
Open-sourced models in red stateof.ai 2022

Closed for 35 months: community-driven open sourcing of AlphaFold et al.

AlphaFold 1 ESM ESM-1B AlphaFold 2 RosettaFold OpenFold
Aug 2022
Aug 2018 Apr 2019 Nov 2020 July 2021
ESM 2
Open-sourced models in red.

Note that the models we reference are not necessarily replicas, but can be improved versions or independently developed. stateof.ai 2022
LLMs empower robots to execute diverse and ambiguous instructions

Thanks to their large range of capabilities, LLMs could in principle enable robots to perform any task by
explaining its steps in natural language. But LLMs have little contextual knowledge of the robot’s environment
and its abilities, making their explanations generally infeasible for the robot. PaLM-SayCan solves this.
● Given an ambiguous instruction “I spilled my drink, can you help?”, a carefully prompt-
engineered LLM (e.g. Google’s PaLM) can devise a sequence of abstract steps to pick up and
bring you a sponge. But any given skill (e.g. pick up, put down) needs to be doable by the
robot in concordance with its environment (e.g. robot sees a sponge).
● To incentivise the LLM to output feasible instructions, SayCan maximises the likelihood of an
instruction being successfully executed by the robot.
● Assume the robot can execute a set of skills. Then, for any given instruction and state, the
system selects the skill that maximizes: the probability of a given completion (restricted to the
set of available skills) times the probability of success given the completion and the current
state. The system is trained using reinforcement learning.
● Researchers tested SayCan on 101 instructions from 7 types of language instructions. It was
successful in planning and execution 84% and 74% of the time respectively.
stateof.ai 2022
2021 Prediction: in vision, convolutional networks want a fair fight with transformers…
The introduction of Vision Transformers (ViT) and other image transformers last year as SOTA models on
imaging benchmarks announced the dawn of ConvNets. Not so fast: work from Meta and UC Berkeley argues
that modernizing ConvNets gives them an edge over ViTs.
● The researchers introduce ConvNeXt, a ResNet which is augmented with
the recent design choices introduced in hierarchical vision Transformers
like Swin, but doesn’t use attention layers.
● ConvNeXt is both competitive with Swin Transformer and ViT on
ImageNet-1K and ImageNet-22K and benefits from scale like them.
● Transformers quickly replaced recurrent neural networks in language
modeling, but we don’t expect a similar abrupt drop-off in ConvNets
usage, especially in smaller scale ML use-cases.
● Meanwhile, our 2021 prediction of small transformers + CNN hybrid
models manifested in MaxViT from Google with 475M parameters almost
matching (89.53%) CoAtNet-7’s performance (90.88%) on ImageNet top-
1 accuracy.
stateof.ai 2022
…but the inevitable vision and language modeling unification continues…

Self-supervision techniques used to train transformers on text are now transposed almost as is to images and are
achieving state of the art results on ImageNet-1K.
● Transformer-based autoencoder LMs are trained to predict randomly
masked words in large text corpora. This results in powerful models that are
SOTA in language modeling tasks (e.g. BERT).
● While masking a word in a sentence makes the sentence nonsensical and
creates a challenging task for LMs, reconstructing a few randomly masked
pixels in images is trivial thanks to neighbouring pixels.
● The solution: mask large patches of pixels (e.g. 75% of the pixels). Meta use this and other adjustments (the encoder
only sees visible patches, the decoder is much smaller than the encoder) to pre-train a ViT-Huge model on ImageNet-
1K and then fine-tune it to achieve a task-best 87.8% top-1 accuracy.
● Self-supervised learning isn’t new to computer vision (see for e.g. Meta’s SEER model). Nor are masking techniques
(e.g. Context encoders, or a more recent SiT). But this work is further evidence that SOTA techniques in language
transition seamlessly vision. Can domains unification be pushed further?
stateof.ai 2022
2021 Prediction:…culminating in a single transformer to rule them all?

Transformers trained on a specific task (via supervised or self-supervised learning) can be used for a broader set
of tasks via fine-tuning. Recent works show that a single transformer can be directly and efficiently trained on
various tasks across different modalities (multi-task multimodal learning).
● Attempts at generalist multitask, multimodal models date back to at least
Google’s “One model to learn them all” (2017), which tackled 8 tasks in
image, text and speech. DeepMind’s Gato brings this effort to another
level: researchers train a 1.2B parameter transformer to perform hundreds
of tasks in robotics, simulated environments, and vision and language.
This partially proves our 2021 Prediction.
● They showed that scaling consistently improved the model, but it was
kept “small” for live low-latency robotics tasks.
● To train their model on different modalities, all data was serialized into a sequence of tokens which are embedded in a
learned vector space. The model is trained in a fully supervised fashion.
● Separately: With data2vec, on a narrower set of tasks, Meta devised a unified self-supervision strategy across
modalities. But for now, different transformers are used for each modality.
stateof.ai 2022
2021 Prediction: transformers for learning in world models in reinforcement

learning
In 2021, we predicted: “Transformers replace RNNs to learn world models with which RL agents surpass human
performance in large and rich game environments.” Researchers from the University of Geneva used a GPT-like
transformer to simulate the world environment. They showed that their agent (dubbed IRIS) was sample
efficient and surpassed human performance on 10 of the 26 games of Atari. IRIS was notably the best method
among the ones that don’t use lookahead search.
stateof.ai 2022
Transformers are becoming truly cross-modality

In the 2020 State of AI Report we predicted that transformers would expand beyond NLP to achieve state of the
art in computer vision. It is now clear that transformers are a candidate general purpose architecture. Analysing
transformer-related papers in 2022 shows just how ubiquitous this model architecture has become.
41%
81%
22%
16%
2%
9%
7%
5%
stateof.ai 2022
NeRFs expand into their own mature field of research

The seminal NeRF paper was published in March 2020. Since then, fundamental improvements to the methods
and new applications have been quickly and continuously developed. For example, more than 50 papers on NeRF
alone appeared at CVPR in 2022.
● From last year’s Report (slide 18): Given multiple views of an image, NeRF uses a
multilayered perceptron to learn a representation of the image and to render new
views of it. It learns a mapping from every pixel location and view direction to the
color and density at that location.
● Among this year’s work, Plenoxels stands out by removing the MLP altogether and
achieving a 100x speedup in NeRF training. Another exciting direction was
rendering large scale sceneries from a few views with NeRFs, whether city-scale
(rendering entire neighborhoods of San Francisco with Block-NeRF) or satellite-
scale with Mega-NeRF*.
● Given the current quality of the results and the field’s rate of progress, we expect that in a year or two, NeRFs will
feature prominently in our industry section.
*You can better appreciate NeRF research by checking demos. E.g.
Block-NeRF, NeRF in the dark, Light Field Neural Rendering stateof.ai 2022
Treating bacterial infections by data-driven personalised selection of antibacterial agents

Resistance to antibacterial agents is common and often arises as a result of a different pathogen already present
in a patient’s body. So how should doctors find the right antibiotic that cures the infection but doesn’t render the
patient susceptible to a new infection?
● By comparing the microbiome profiles of >200,000 patients with urinary tract or wound infections who were treated
with known antibiotics before and after their infections, ML can be used to predict the risk of treatment-induced gain
of resistance on a patient-specific level.
● Indeed, urinary tract infection (UTI) patients treated with antibiotics that the ML system would not have
recommended resulted in significantly resistance (E). Both UTI (F) and wound infection (G) patients would suffer far
fewer reinfections if they’d have been prescribed antibiotics according to the ML system.
stateof.ai 2022
Interpreting small molecule mass spectra using transformers

Tandem mass spectrometry (MS/MS) is commonly used in metabolomics, the study of small molecules in
biological samples. Less than 10% of small molecules can be identified from spectral reference libraries as most
of nature’s chemical space is unknown. Transformers enable fast, accurate, in silico, characterization of the
molecules in metabolic mixtures, enabling biomarker and natural product drug discovery at scale.
● Very few biological samples can typically be
identified from reference libraries.
● Property-prediction transformers outperform
at predicting a suite of medicinally-relevant
chemical properties like solubility, drug
likeness, and synthetic accessibility directly
from MS/MS, without using structure
prediction intermediates or reference
lookups.
stateof.ai 2022
Drug discovery, the flagship “AI for good” application, is not immune to misuse
Researchers from Collaborations Pharmaceuticals and King’s College London showed that machine learning
models designed for therapeutic use can be easily repurposed to generate biochemical weapons.
● The researchers had trained their “MegaSyn” model to maximize bioactivity and
minimize toxicity. To design toxic molecules, they kept the same model, but now
simply training it to maximize both bioactivity and toxicity. They used a public
database of drug-like molecules.
● They directed the model towards generation of the nerve agent VX, known to be
one of the most toxic chemical warfare agents.
● However, as is the case with regular drug discovery, finding molecules with a high
predicted toxicity doesn’t mean it is easy to make them. But as drug discovery
with AI in the loop is being dramatically improved, we can imagine best practices
in drug discovery diffusing into building cheap biochemical weapons.
stateof.ai 2022
Compared to US AI research, Chinese papers focus more on surveillance related-tasks.

These include autonomy, object detection, tracking, scene understanding, action and
speaker recognition.
Comparing data modalities in Chinese vs. US Comparing machine learning tasks in Chinese vs. US papers
papers
Red = more common in China
Blue = more common in the US
stateof.ai 2022
While US-based authors published more AI papers than Chinese peers in 2022, China and
Chinese institutions are growing their output at a faster rate
# papers published in 2022 and change vs. 2021 # papers published in 2022 and change vs. 2021
+11% +27%
+24% +13%
-2% +13%
+3% +11%
+10% +1%
+4%
stateof.ai 2022
The China-US AI research paper gap explodes if we include the Chinese-language

database, China National Knowledge Infrastructure
Chinese institutions author 4.5x the number of papers than American institutions since 2010.
stateof.ai 2022
Section 2: Industry
stateof.ai 2022
Do upstart AI chip companies still have a chance vs. NVIDIA’s GPU?

NVIDIA’s FY 2021 datacenter revenue came in at $10.6B. In Q4 2021, they recognised $3.26B, which on an
annualised basis is greater than the combined valuation of top-3 AI semiconductor startups. NVIDIA has over
3 million developers on their platform and the company’s latest H100 chip generation is expected to deliver 9x
training performance vs. the A100. Meanwhile, revenue figures for Cerebras, SambaNova and Graphcore are
not publicly available.
Latest private valuation Annualised datacenter revenue
$5.1 billion
$4 billion $13 billion
$2.8 billion
stateof.ai 2022
NVIDIA’s chips are the most popular in AI research papers…and by a massive margin
GPUs are 131x more commonly used than ASICs, 90x more than chips from Graphcore, Habana, Cerebras,
SambaNova and Cambricon combined, 78x more than Google’s TPU, and 23x more than FPGAs.
log scale
23x
78-131x
stateof.ai 2022
For NVIDIA, the V100 is most popular, and Graphcore is most used amongst challengers
The V100, released in 2017, is NVIDIA’s workhorse chip, followed by the A100 that was released in 2020. The
H100 is hotly awaited in 2022. Of the major AI chip challengers, Graphcore is cited most often.
Number of AI papers citing use of specific NVIDIA cards Number of AI papers citing use of specific AI chip startups
stateof.ai 2022
NVIDIA fails to acquire Arm and grows its revenue 2.5x and valuation 2x during the deal
Announced at $40B, NVIDIA’s attempted acquisition of Arm fell through due to significant geopolitical and
anti competition pushback. Nonetheless, NVIDIA’s enterprise value grew by $295B during the period (!!)
Deal announced Deal cancelled
stateof.ai 2022
NVIDIA reaps rewards from investing in AI research tying up hardware and software
NVIDIA has been investing heavily in AI research and producing some of the best works in imaging over the
years. For instance, their latest work on view synthesis just won the best paper award at SIGGRAPH, one of the
most prestigious computer graphics conferences. But NVIDIA has now gone a step further and applied their
reinforcement learning work to design their next-generation AI chip, the H100 GPU.
stateof.ai 2022
David teaming up with Goliath: training large models requires compute partnerships
The hyperscalers and challenger AI compute providers are tallying up major AI compute partnerships,
notably Microsoft’s $1B investment into OpenAI. We expect more to come.
None yet?
None yet?
stateof.ai 2022
In a gold rush for compute, companies build bigger than national supercomputers
“We think the most benefits will go to whoever has the biggest computer” – Greg Brockman, OpenAI CTO
Current NVIDIA A100 GPU count Future NVIDIA H100 GPU

count
d e
at
m
sti
*e
stateof.ai 2022
The compounding effects of government contracting in AI

In 1962 the US government bought all integrated circuits in the world, supercharging the development of this
technology and its end markets. Some governments are providing that opportunity again, as “buyers of first
resort” for AI companies. With access to unique high-quality data, companies could gain an edge in building
consumer or enterprise AI software.
● Researchers examined Chinese facial recognition AI companies and
showed a causal relationship between the number of government contracts
they signed and the cumulative amount of general AI software they
produced. Unsurprisingly, leadership in the computer vision space has
largely been ceded to Chinese companies now.
● The principle should stand in other heavily regulated sectors, like defence
or healthcare, which build an expertise through unique data that is
transferable to everyday AI products.
stateof.ai 2022
How should big tech deal with their language model consumer products?
Meta’s release of the BlenderBot3 chatbot for free public use in August 2022 was faced with catastrophic press
because the chatbot was spitting misinformation. Meanwhile, Google, which published a paper on their chatbot
LaMDA in May 2021, had decided to keep the system in-house. But a few weeks after BlenderBot’s release,
Google announced a larger initiative called “AI test kitchen”, where regular users will be able to interact with
Google’s latest AI agents, including LaMDA.
● Large-scale release of AI systems to the 1B+ users of Google and Facebook
all but ensures that every ethics or safety issue with these systems will be
surfaced, either by coincidence or by adversarially querying them. But only
by making these systems widely available can these companies fix those
issues, understand user behaviour and create useful and profitable systems.
● Running away from this dilemma, 4 of the authors of the paper introducing
LaMDA went on to found/join Character.AI, which describes itself as “an AI
company creating revolutionary open-ended conversational applications”.
Watch this space…
stateof.ai 2022
DeepMind and OpenAI alums form new startups and Meta disbands its core AI group
Once considered untouchable, talent from Tier 1 AI labs is breaking loose and becoming entrepreneurial.
Alums are working on AGI, AI safety, biotech, fintech, energy, dev tools and robotics. Others, such as Meta,
are folding their centralised AI research group after letting it run free from product roadmap pressure for
almost 10 years. Meta concluded that “while the centralized nature of the [AI] organization gave us leverage
in some areas it also made it a challenge to integrate as deeply as we would hope.”
stateof.ai 2022
Attention is all you need… to build your AI startup

All but one author of the landmark paper that introduced transformer-based neural networks have left
Google to build their own startups in AGI, conversational agents, AI-first biotech and blockchain.
Amount raised in 2022
$580M
$225M
$125M
$65M
stateof.ai 2022
AI coding assistants are deployed fast, with early signs of developer productivity gains
OpenAI’s Codex quickly evolved from research (July 2021) to open commercialization (June 2022) with
(Microsoft’s) GitHub Copilot now publicly available for $10/month or $100/year. Amazon followed suit by
announcing CodeWhisperer in preview in June 2022. Google revealed that it was using an internal ML-
powered code completion tool (so maybe in a few years in a browser IDE?). Meanwhile, with its 1M+ users,
tabnine raised $15M, promising accurate multiline code completions.
Single line Multi-line
Fraction of code added by ML 2.6% 0.6%
Average characters per accept 21 73
Acceptance rate (for suggestions visible for >750ms) 25% 34%
Reduction in coding iteration duration 6% -
Metrics for Google’s coding assistant. Users are 10k+

Google-internal developers (5k+ for multi-line
experiments).
stateof.ai 2022
AI-first drug discovery companies have 18 assets in clinical trials, up from 0 in 2020
And many more assets in early discovery stages. We expect early clinical trial readouts from 2023 onwards.
# of assets per pipeline stage per company % of assets per pipeline stage overall
Updated as of 26 Aug 2022 stateof.ai 2022

Can AI and compute bend the physical reality of clinical trial chokepoints?
A study of 6,151 successful phase transitions between 2011–2020 found that it takes 10.5 years on average for a
drug to achieve regulatory approval. This includes 2.3 years at Phase I, 3.6 years at Phase II, 3.3 years at
Phase III, and 1.3 years at the regulatory stage. What’s more, it costs $6.5k on average to recruit one patient
into a clinical trial. With 30% of patients eventually dropping out due to non-compliance, the fully-loaded
recruitment cost is closer to $19.5k/patient. While AI promises better drugs faster, we need to solve for the
physical bottlenecks of clinical trials today.
# of registered studies (ClinicalTrials.gov EOY) Stepwise probability of drug success
stateof.ai 2022
Predicting the evolution of real-world covid variants using language models

mRNA vaccine leader, BioNTech, and enterprise AI company, InstaDeep, collaboratively built and validated
an Early Warning System (EWS) to predict high-risk variants. The EWS could identify all 16 WHO-designated
variants on average more than one and a half months prior to officially receiving the designation.
● A large pre-trained protein language model was trained
on viral spike protein sequences of variants.
● New spike protein variants are fed to a transformer that
outputs embeddings and a probability distribution of the
20 natural amino acids for each position to determine
how this would affect immune escape and fitness.
● The red dash line indicates the date when the EWS
predicted the variant would be high-risk and the green
dash-dot line is when the WHO designated the variant.
In almost all cases, EWS alerted several months before
the WHO designation.
stateof.ai 2022
The first regulatory approval for an autonomous AI-first medical imaging diagnostic
Lithuanian startup Oxipit received the industry’s first autonomous certification for their computer vision-
based diagnostic. The system autonomously reports on chest X-rays that feature no abnormalities, removing
the need for radiologists to look at them.
● Due to a shortage of radiologists and an increasing volume of
imaging, the diagnostic task of assessing which X-rays contain
disease and which don’t is challenging.
● Oxipit’s ChestLink is a computer vision system that is tasked with
identifying scans that are normal.
● The system is trained on over a million diverse images. In a
retrospective study of 10,000 chest X-rays of Finnish primary
health care patients, the AI achieved a sensitivity of 99.8% and specificity of 36.4 % for recognising clinically
significant pathology on a chest X-ray.
● As such, the AI could reliably remove 36.4% of normal chest X-rays from a primary health care population data set
with a minimal number of false negatives, leading to effectively no compromise on patient safety and a potential
significant reduction of workload.
stateof.ai 2022
Universities are a hotbed for AI spinouts: the UK case study

Universities are an important source of AI companies including Databricks, Snorkel, SambaNova, Exscientia
and more. In the UK, 4.3% of UK AI companies are university spinouts, compared to 0.03% for all UK
companies. AI is indeed among the most represented sectors for spinouts formation. But this comes at a steep
price: Technology Transfer Offices (TTOs) often negotiate spinout deal terms which are unfavourable to
founders, e.g. a high equity share in the company or royalties on sales.
stateof.ai 2022
Spinout.fyi: an open database to help founders and policymakers fix the spinout problem
Spinout.fyi crowdsourced a database of spinout deal terms from founders representing >70 universities all over
the world. The database spans AI and non-AI companies across different product categories (software,
hardware, medical, materials, etc.), and shows that the UK situation, while particularly discouraging for
founders, isn’t isolated. Only a few regions stand out as being founder-friendly, like the Nordics and Switzerland
(ETH Zürich in particular). A major reason for the current situation is the information asymmetry between
founders and TTOs, and the spinout.fyi database aims to give founders a leg up in the process.
stateof.ai 2022
As 5-year programmes in Berkeley and Stanford wrap up, what comes next?
In 2011, UC Berkeley launched the “Algorithms, Machines, and People” (AMPLab) as a 5-year collaborative
research agenda amongst professors and students, supported by research agencies and companies. The program
famously developed the critical Big Data technology Spark (spun out as Databricks), as well as Mesos (spun out
as Mesosphere). This hugely successful program was followed in 2017 by the “Real-time intelligence secure
explainable systems” (RISELab) at Berkeley and “Data Analytics for What’s Next” (DAWN) at Stanford, which
focused on AI technologies. RISELab created the Ray ML workload manager (spun out as Anyscale), and DAWN
created and spun out the Snorkel active labelling platform. Will other universities and countries learn from the
successes of the 5-year model to fund ambitious open-source research with high spinout potential?
Lab name OSS project created Spinouts that emerged

2011-16 $38B val $250M raised
$1B val
2017-22
$1B val
stateof.ai 2022
In 2022, investment in startups using AI has slowed down along with the broader market
Private companies using AI are expected to raise 36% less money in 2022* vs. last year, but are still on track to
exceed the 2020 level. This is comparable with the investment in all startups & scaleups worldwide.
Worldwide investment in startups & scaleups Worldwide investment in all startups & scaleups
using AI by round size » view online by round size » view online
▊ $250M+ ▊ $100-250M ▊ $40-100M (Series C) ▊ $15-40M (Series B) ▊ $250M+ ▊ $100-250M ▊ $40-100M (Series C) ▊ $15-40M (Series B)
▊ $4-15M (Series A) ▊ $1-4M (Seed) ▊ $0-1M (Pre-Seed) ▊ $4-15M (Series A) ▊ $1-4M (Seed) ▊ $0-1M (Pre-Seed)
$150B $800B $726.5B -24%

$111.4B $553.2B*
$600B
$100B -36%
$69.3B $70.9B* $399.7B
$400B $351.7B
$47.5B
$50B
$200B
0 0
10
12
13
14
16
17
18
20
21
10
11
12
14
15
16
19
20
11
15
19
13
17
18
21
YT 22
YT 22
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
D
D
In this slide, startups & scaleups using AI include both startups & scaleups with AI-first and AI-enabled products and solutions.
(*) estimated amount to be raised by the end of 2022 stateof.ai 2022
The drop in investment is most noticeable in megarounds

The drop in VC investment is most noticeable in 100M+ rounds, whereas smaller rounds are expected to amount
to $30.9B worldwide by the end of 2022, which is almost on track with the 2021 level.
Worldwide investment in startups & scaleups Worldwide investment in startups & scaleups using
using AI by round size » view online AI by round size » view online
▊ $250M+ ▊ $100-250M ▊ $40-100M (Series C) ▊ $15-40M (Series B) ▊ $4-15M (Series A)
▊ $1-4M (Seed) ▊ $0-1M (Pre-Seed)
$80B $77.5B $40B

$33.9B -9%
-55% $30.9B*
$60B $30B
$24.9B
$44.4B $22.3B
$40B $34.9B* $20B
$25.2B
$20B $10B
0 0
12
15
16
17
19
20
10
11
13
14
15
17
18
19
21
10
11
13
14
18
21
12
16
20
YT 22
YT 22
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
D
D
Public valuations have dropped in 2022, while private keep growing

Combined public enterprise value (EV) has dropped to the 2020 level. Meantime, private valuations keep
growing, with the combined EV already reaching $2.2T, up 16% from last year.
Combined EV of public startups & scaleups using Combined EV of privately owned startups & scaleups
AI by launch year; worldwide » view online using AI by launch year; worldwide » view online
▊ 2015-2022 YTD ▊ 2010-2014 ▊ 2005-2009 ▊ 2000-2004 ▊ 2015-2022 YTD ▊ 2010-2014 ▊ 2005-2009 ▊ 2000-2004
▊ 1995-1999 ▊ 1990-1994 ▊ 1995-1999 ▊ 1990-1994
$9.6T
$10.0T $2.5T +16% $2.2T
-29% $1.9T
$8.0T $2.0T
$6.8T $6.8T
$6.0T $1.5T $1.4T
$4.0T $1.0T
$2.0T $500B
0 0
10
12
13
16
17
18
20
21
10
11
14
15
16
18
19
21
11
14
15
19
12
13
17
20
YT 22
YT 22
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
D
D
The US leads by the number of AI unicorns, followed by China & the UK

The US has created 292 AI unicorns, with the combined enterprise value of $4.6T.
Countries with the largest number of AI unicorns » view online
Combined enterprise value (2022

Number of AI unicorns Examples
YTD)
United States 292 $4.6T
China 69 $1.4T
United Kingdom 24 $207B
Israel 14 $53B
Germany 10 $56B
Canada 7 $12B
Singapore 6 $39B
Switzerland 6 $14B
stateof.ai 2021
Investment in the USA accounts for more than half of the worldwide VC
Despite significant drop in investment in US-based startups & scaleups using AI, they still account for more than
half of the AI investment worldwide.
Amount invested in the companies using AI

$140B
$120B
▊ Rest of the world
$100B
$13.1B
$80B ▊ China
$7.5B
$8.9B
$60B ▊ United Kingdom
$40B $4.3B
$63.6B $3.8B ▊ EU-27, Switzerland & Norway
$7.5B
$20B $25.1B 53% in 2022
▊ USA
YTD vs. 57% in
0
2022 2021
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
YTD
stateof.ai 2022
Enterprise software is the most invested category globally, while robotics captures the largest
share of VC investment into AI
Amount invested in Amount invested in Investment in startups & scaleups Amount invested in startups
Industries startups & scaleups startups & scaleups using AI, 2021-2022 YTD; % of AI startups & scaleups using AI as %
using AI, 2010-2022 YTD using AI, 2021-2022 YTD number of rounds of all VC, 2021-2022 YTD
Enterprise software $133.7B $44.7B 1288 8% 30%
Transportation $109.5B $22.8B 332 2% 21%
Fintech $87.8B $25.1B 660 3% 13%
Health $61.7B $21.9B 689 3% 14%
Robotics $48.9B $18.4B 367 8% 71%
Food $44.1B $6.0B 220 1% 10%
Marketing $43.0B $13.5B 406 4% 24%
Security $37.9B $12.2B 306 5% 25%
Media $27.4B $4.0B 182 2% 13%
Telecom $17.9B $2.7B 44 2% 13%
Semiconductors $17.8B $6.8B 73 2% 41%
Education $14.7B $2.5B 143 3% 9%
Energy $13.1B $3.6B 231 2% 6%
Travel $10.6B $3.4B 27 1% 24%
Real estate $7.7B $617M 148 1% 2%
Gaming $6.5B $629M 55 1% 3%
Home living $6.0B $1.5B 31 1% 8%
Jobs recruitment $5.2B $2.4B 103 3% 14%
Legal $4.9B $568M 88 2% 7%
Sports $3.4B $421M 56 1% 4%
Music $2.5B $28M 22 1% 1%
Fashion $2.4B $349M 48 1% 2%
Hosting $2.2B $408M 16 <1% 9%
Wellness beauty $2.1B $348M 56 1% 3%
Event tech $927M $33M 14 1% 1%
Kids $292M $69M 13 <1% 2%
Dating $85M $5M 4 <1% 4%
stateof.ai 2022
Acquisitions are on track to exceed the 2021 level

While the number of IPOs and SPAC IPOs declined sharply, the number of acquisitions is on track to exceed the
2021 level
Number of exits among the companies using AI; Number of exits in 2022 among the companies
worldwide » view online using AI; worldwide » view online
▊ Acquisition ▊ Buyout ▊ IPO ▊SPAC IPO
500
436 Technology, data & analytics SaaS for planning Customer experience
for real estate & forecasting tools
$13.1B Acquisition $10.7B Buyout $10.2B Buyout
400 May 2022 Mar 2022 Jun 2022
285
300 253
Tax compliance Consumer robotics Autonomous, continuous
software company workload optimization
200 $8.4B Buyout $1.7B Acquisition $650M Acquisition
332 Aug 2022 Aug 2022 Mar 2022
259
211
100
Mobile device & application Cloud-native SOAR Developing autonomous

0 security solutions platform mobility platform
$525M Buyout $500M Acquisition $469 Acquisition
10 11 12 13 14 15 16 17 18 19 20 21 22
20 20 20 20 20 20 20 20 20 20 20 20 20 TD
Mar 2022 Jan 2022 Aug 2022
Y
stateof.ai 2022
The number of exits in EU-27, Switzerland & Norway has already exceeded 2021 levels
With 108 exits to date, the US hasn’t yet reached half of the 2021 level, while the EU, Switzerland & Norway
combined have already exceeded the 2021 number.
Number of exits among the companies using AI
500
450
▊ Rest of the world
400
350 9
33 ▊ China
300
94
250 1 ▊ United Kingdom
25
200
97
150 ▊ EU-27, Switzerland & Norway
222
100
108 ▊ USA
50
0
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
YTD
stateof.ai 2022
Investment in SaaS startups & scaleups using AI is expected to reach $41.5B by the end of
the year, down 33% from last year, but higher than in 2020
VC investment in AI SaaS startups & scaleups » view online
$80.0B
$61.8B
-33%
$60.0B
▊ $250M+
$41.5B
$37.7B ▊ $100-250M
*
$40.0B
$31.1B ▊ $40-100M (Series C)
▊ $15-40M (Series B)
$20.0B 222
▊ $4-15M (Series A)
108
▊ $1-4M (Seed)
0 ▊ $0-1M (Pre-Seed)
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
YTD
The combined EV of public and private SaaS startups & scaleups using AI now amounts
to $2.3T, down 26% from last year, but still higher than in 2020
Combined EV of AI SaaS startups & scaleups by launch year
globally » view online
$4.0T
$3.1T -26%
$3.0T
▊ 2015-2022 YTD
$2.3T
$2.1T ▊ 2010-2014
$2.0T
▊ 2005-2009
▊ 2000-2004
$1.0T ▊ 1995-1999
▊ 1990-1994
0
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 YTD
In this slide, startups & scaleups using AI include both startups & scaleups with AI-first and AI-enabled products and solutions. stateof.ai 2021
The combined EV of private SaaS startups & scaleups using AI keeps growing and has
already reached $1.1T, up 12% from last year
Combined EV of privately owned AI SaaS startups & scaleups Top valued privately owned startups & scaleups
by launch year; worldwide » view online using AI » view online
▊ 2015-2022 YTD ▊ 2010-2014 ▊ 2005-2009 ▊ 2000-2004

▊ 1995-1999 ▊ 1990-1994 Financial infrastructure Global payments solution Lakehouse platform to unify
platform provider data, analytics and AI
$1.5T USA United Kingdom USA
Valuation: $68.4B Valuation: $40B Valuation: $38B
+12%
$1.0T
$1.0T $895B
Autonomous driving Online education Fintech-as-a-service
technology platform platform
USA China United Kingdom
$489B Valuation: $30B Valuation: $15.5B Valuation: $15B
$500B
Process mining Business cloud AI-powered writing

software software products assistant
0 Germany USA USA
10 11 12 13 14 15 16 17 18 19 20 21 22
20 20 20 20 20 20 20 20 20 20 20 20 20 TD
Valuation: $13B Valuation: $13B Valuation: $13B
Y
In this slide, startups & scaleups using AI include both startups & scaleups with AI-first and AI-enabled products and solutions. stateof.ai 2021
Section 3: Politics
stateof.ai 2022
A widening compute chasm is separating industry from academia in large model AI

The compute requirements for large-scale AI experiments has increased >300,000x in the last decade. Over the
same period, the % of these projects run by academics has plummeted from ～ 60% to almost 0%. If the AI
community is to continue scaling models, this chasm of “have” and “have nots” creates significant challenges for
AI safety, pursuing diverse ideas, talent concentration, and more.
stateof.ai 2022
Slow progress in providing academics with more compute leaves others to act faster
There is a growing appreciation that AI is an engineering science in which the objects of study need to first be
built. Western academics and governments are starting to wake up to this reality, most notably through the
National AI Research Resource process in the US. While spending years on consultations and marketing
however, others in China and outside academia are finding creative ways to do large-scale AI projects.
Governments / Academia
National AI
Research Resource Stanford opens Center for Research
National AI GLM-130B
Task Force on Foundation Models NAIRRTF Final
Initiative Act (Tsinghua)
Created Report
enacted
NAIRRTF NAIRRTF NAIRRTF NAIRRTF NAIRRTF NAIRRTF NAIRRTF NAIRRTF NAIRRTF NAIRRTF
Meeting 1 Meeting 2 Meeting 3 Meeting 4 Meeting 5 Meeting 6 Meeting 7 Meeting 8 Meeting 9 Meeting 10
Mar Aug
Jan 2021 Ju July Aug Oct Dec Feb 2022 Apr May Jul Sep Oct Dec 2022
n
The Pile GPT-Neo GPT-J-6B GPT-NeoX- Stable Diffusion

(Eleuther) (Eleuther) (Eleuther) 20B (Eleuther) (Stability)
BLOOM-178B
(BigScience)
Research Collectives
stateof.ai 2022
The baton is passing from academia to decentralized research collectives

Decentralized research projects are gaining members, funding and momentum. They are succeeding at ambitious
large-scale model and data projects that were previously thought to be only possible in large centralised
technology companies – most visibly demonstrated by the public release of Stable Diffusion.
● The most notable large-scale academic project this year Large-Scale AI Results
came from China: Tsinghua’s GLM-130B LLM.
● Eleuther, the original AI research collective, released the
20B parameter GPT-NeoX. However, core members have
since moved on to OpenAI, Stability and Conjecture.
● Hugging Face led the BigScience initiative, releasing the
178B parameter BLOOM multilingual LLM.
● Stability came out of nowhere, obtained 4,000 A100 GPUs,
brought together multiple open-source communities and
created Stable Diffusion.
Graph data source: Sevilla et al. Parameter, Compute and Data Trends in Machine Learning stateof.ai 2022
Stability AI is attempting a new paradigm in commercializable open-source AI

Where there was previously a dependence on ad-hoc compute
donations to enable large-scale projects, Stability is pioneering a
new approach of structured compute and resource provision for
open-source communities, while also commercializing these
projects with revenue-sharing for developers.
● Stability has embedded itself as a compute platform for
independent and academic open-source AI communities:
supporting LAION for building a dataset of 5B image-text pairs
and training an open-source CLIP model, and supporting the
CompVis group’s research in efficient diffusion models. It funds
PhD students to work on community projects, and has directly
hired generative AI artists, core members of Eleuther, and
renowned ML researchers such as David Ha.
● Stable Diffusion cost <$600K to train, and while weights were
released, access is also sold through the DreamStudio API. stateof.ai 2022
AI continues to be infused into a greater number of defense product categories

Defense technology companies are applying AI to electronic warfare, geospatial sensor fusion, and to create
autonomous hardware platforms.
● Epirus, founded in 2018, has built a next-generation electromagnetic pulse weapon
capable of defeating swarms of drones that pose threats to human safety. Sweden’s
Saab is also making efforts towards AI-driven automation of electronic warfare: they
built the COMINT and C-E.SM sensors to balance automated and operator-controlled
surveillance depending on the context on the field. The company is also collaborating
with defense startup, Helsing.
● Modern Intelligence, founded in 2020, builds a platform-independent AI for
geospatial sensor data fusion, situational awareness and maritime surveillance.
● Meanwhile, through both organic and inorganic growth, Anduril has expanded its
autonomous hardware platforms. For example, Anduril acquired Area-I to launch a
new product in Air Launched Effects with an increased payload, data sharing and
integration capabilities with other UAVs. Anduril also expanded into Underwater
Autonomous Vehicles by acquiring Dive Technologies.
stateof.ai 2022
AI in defense gathers big funding momentum

Heavily funded start-ups and Amazon, Microsoft, and Google continue to normalise the use of AI in Defense.
● Nato published their AI Strategy and announced a $1B fund to invest in
companies working on a range of dual-use technologies. It was described
as the world’s first ‘multi-sovereign venture capital fund’ spanning 22
nations.
● Helsing, a European defense AI company, announced a €102.5M Series A
led by Daniel Ek of Spotify.
● Microsoft, Amazon and Google continue to compete for a major role in
defense - most notably Microsoft’s $10B contract with the Pentagon was
cancelled after a lawsuit from Amazon. The new beneficiaries of the
contract will now be announced in late 2022.
● Anduril landed their largest DoD contract to date and is now reportedly
valued at $7B.
● Shield AI, developer of military drones, raised at a $2.3B valuation
stateof.ai 2022
Ukraine’s homegrown geospatial intelligence GIS Arta software is a sign of things to come
The use of geospatial (GIS) software has reportedly reduced the decision chain around artillery from 20 minutes
to under one minute.
● GIS Arta is a homegrown application developed prior to Russia’s
invasion based on lessons learned from the conflict in the Donbas.
● It’s a guidance command and control system for drone, artillery or
mortar strikes.
● The app ingests various forms of intelligence (from drones, GPS,
forward observers etc) and converts it into dispatch requests for
reconnaissance and artillery.
● GIS Arta was allegedly developed by a volunteer team of software
developers led by Yaroslav Sherstyvk, inspired by the Uber taxi
model.
stateof.ai 2022
The Great Reshoring will be slow: US lags in new fab projects, which take years to build
Between 1990 and 2020, China accelerated its output of greenfield fab projects by almost 7x while the US slowed
down by 2.5x. Moreover, while China and Taiwan fabs take roughly 650 days from construction start to being
production-ready, the US builds fabs 42% slower today than they did 30 years ago.
Total # of greenfield fab projects Avg. # of days from build start to production
stateof.ai 2022
The US CHIPS and Science Act of 2022: $250B for US semiconductor R&D and
production
The bipartisan legislation was signed into law in August 2022. It provides for $53B to boost US-based
semiconductor R&D, workforce and manufacturing, as well as a 25% investment tax credit for semiconductor
manufacturers’ capital expenses. In exchange, recipients must not upgrade or expand their existing operations
in China for 10 years, nor can they use funds for share buybacks or to issue dividends.
● The bill poses a dilemma for Korean (e.g. Samsung), Taiwanese (e.g.
TSMC) and other manufacturers: if they accept US subsidies, then they
must pivot away from China without backlash from Beijing, which is
opposed to this “friendshoring”.
● Since passing the bill, Micron announced a $40B investment in memory
chip manufacturing to increase US market share from 2% to 10%.
Qualcomm will expand its US semiconductor production by 50% by 2027
and in partnership with GlobalFoundries the two will invest $4.2B to
expand the latter’s upstate New York facility.
● CSET estimates the US should focus on its manufacturing capabilities in
leading-edge, legacy logic and DRAM (right chart).
stateof.ai 2022
The US cuts China off from NVIDIA and AMD chips…will this spur Chinese AI R&D?
NVIDIA GPUs are used by all major Chinese technology companies (Baidu, Tencent et al.) and universities
(Tsinghua, Chinese Academy of Sciences et al.). Washington ordered NVIDIA and AMD to stop exporting
their latest AI chips (e.g. NVIDIA A100 and H100, and AMD M100 and M200) to China as a means of curbing
their use in applications that threaten American national security via China. The companies will have to
provide statistics on previous shipments and customer lists. Not having access to state of the art AI chips could
stall a large swath of Chinese industry if domestic suppliers don’t step into the void and fast.
● Earlier this year, CSET analysed 24 public contracts awarded by Chinese PLA units and state-owned defense
enterprises in 2020. They found that nearly all of the 97 AI chips in these purchase orders were designed by
NVIDIA, AMD, Intel and Microsemi. Domestic AI chip companies were not featured. Thus, American chips are
arming Chinese defense capabilities.
● Chinese semiconductor manufacturers have been already cut off from advanced lithography machines made by
ASML and related equipment from Lam Research and Applied Materials.
● It is unlikely that domestic AI chip companies (e.g. Biren) can fill the void: leading-edge node manufacturing is still
only possible by TSMC in Taiwan and because domestic talent, software and technology is still years away from
NVIDIA. China will still accelerate its development. stateof.ai 2022
The EU advances with its plans to regulate AI

In April 2021, the EU tabled a proposal for regulating the placement on the market and use in the EU of AI
systems (the “AI Act”). The proposal introduces certain minimum requirements (e.g. mainly information
obligations) that all AI systems in use must meet. It also introduces more elaborate requirements (e.g. risk
assessments, quality management) with respect to AI systems that pose higher risks to users. The AI Act bans
the use of certain types of AI-based techniques (e.g. social scoring, real-time biometric remote identification
(subject to exceptions), “subliminal techniques”).
● The AI Act moves through the EU legislative process. The European Parliament has worked over the summer on a
compromise text to address tabled amendments and opinions by the Parliament’s various committees. The
compromise text is scheduled to go through the various stages of the voting process at the European Parliament by
the end of 2022.
● The AI Act is expected to be voted into law in 2023, either under the Swedish or the Spanish Presidency of the EU.
● Current realistic expectations are that the AI Act will become effective in the second-half of 2023.
Source: Dessislava Fessenko

stateof.ai 2022
The EU aims at quick operationalization of the AI Act

The EU aims at quick operationalization of the requirements under the AI Act through standardization,
setting up of testing facilities, and launch of pan-European and national regulatory sandboxes.
● European standardization efforts are already underway. The EU standardization organizations CEN and CENELEC
have already commenced preparatory works on standardization and expect to be requested to develop relevant sets
of standards by 31 October 2024.
● The EU appears to favor testing of high-risk AI systems, in either controlled or even possibly in real-world
conditions, as a suitable mode for supporting and promoting compliance with the AI Act among businesses of all
sizes.
● Pan-European and national regulatory sandboxes start to emerge in the EU. Spain launched the first one in June
2022. Other EU member states (e.g. the Czech Republic) have announced similar plans. Sandboxes are considered
by EU regulators as suitable testbeds for technical, policy and standardization solutions. They are also intended as a
medium for supporting small and medium-sized businesses in particular in attaining compliance with the AI Act.
Source: Dessislava Fessenko

stateof.ai 2022
Section 4: Safety
stateof.ai 2022
While AI advances rapidly, the safety of highly-capable future systems remains unclear
While many concerns still appear speculative, early AI pioneers considered that highly capable and
economically integrated AI systems of the future could fail catastrophically and pose a risk to humanity,
including through the emergence of behaviours directly opposed to human oversight and control.
“... it seems probable “Thus the first “The problem is that,

that once the machine ultraintelligent with such powerful
thinking method has machine is the last machines, it would
started, it would not invention that man require but the
take long to outstrip need ever make, slightest accident of
our feeble powers. … provided that the careless design for
At some stage machine is docile them to place their
Alan Turing therefore we should I.J. Good enough to tell us how Marvin goals ahead of ours”
1951 1965 Minsky
have to expect the to keep it under 1984
machines to take control.”
control”
stateof.ai 2022
The UK is taking the lead on acknowledging these uncertain but catastrophic risks
The UK’s national strategy for AI, published in late 2021, notably made multiple references to AI safety and
the long-term risks posed by misaligned AGI.
● “While the emergence of Artificial General Intelligence (AGI) may seem like a
science fiction concept, concern about AI safety and non-human-aligned systems
is by no means restricted to the fringes of the field.”
● “We take the firm stance that it is critical to watch the evolution of the technology,
to take seriously the possibility of AGI and ‘more general AI’, and to actively
direct the technology in a peaceful, human-aligned direction.”
● “The government takes the long term risk of non-aligned AGI, and the
unforeseeable changes that it would mean for the UK and the world, seriously.”
● “[We must] establish medium and long term horizon scanning functions to
increase government’s awareness of AI safety.”
● “[We must] work with national security, defence, and leading researchers to
understand how to anticipate and prevent catastrophic risks.”
stateof.ai 2022
AI researchers increasingly believe that AI safety is a serious concern

Long dismissed as science fiction by mainstream AI research and academia, researchers are now shifting
consensus towards greater concern for the risks of human-level AI and superhuman AGI in the near future.
● A survey of the ML community found that

69% believe AI safety should be prioritized
more than it currently is.
● A separate survey of the NLP community
found that a majority believe AGI is an
important concern we are making progress
towards. Over 70% believe AI will lead to
social change at the level of the Industrial
Revolution this century, and nearly 40%
believe AI could cause a catastrophe as bad as Note: The number in green represents the fraction of respondents who agree with the position out of all
nuclear war during that time. those who took a side. The number in black shows the average predicted rate of agreement.
stateof.ai 2022
AI safety is attracting more talent… yet remains extremely neglected

Increased awareness of AI existential risk is leading to increased headcount, Researchers by venue/field
with an estimated 300 researchers now working full-time on AI safety.
However, this is still orders of magnitude fewer researchers than are working
in the broader field, which itself is growing faster than ever (right chart).
● New non-profit research labs include the Center for AI Safety and the Fund for
Alignment Research. The Centre for the Governance of AI was spun out as an
independent organization from the Future of Humanity Institute in Oxford. 30x
● There was a huge increase in interest for education programmes with over 750
people taking part in the online AGI Safety Fundamentals course. New
scholarships were created, including the Vitalik Buterin PhD Fellowship in AI
Existential Safety.
● Notably, Ilya Sutskever, OpenAI’s Chief Scientist, has shifted to spending 50% of
his time on safety research.
stateof.ai 2022
Funding secured, though trailing far behind what goes into capabilities
Increased awareness of AI existential risk has led to rapidly increasing funding for research into the safety of
highly-capable systems, primarily through donations and investments from sympathetic tech billionaires Dustin
Moskovitz (Open Philanthropy) and Sam Bankman-Fried (FTX Foundation). However, total VC and
philanthropic safety funding still trails behind resources for advanced capabilities research, not even matching
DeepMind’s 2018 opex.
Philanthropic AI Safety funding pales in comparison to AI capabilities funding*
*We include fundraises for Adept, Hugging Face, Cohere, AI21, Stability and Inflection under Capabilities VC and fundraises for Anthropic under Safety VC. stateof.ai 2022
Language Model Alignment: Reinforcement Learning from Human Feedback (RLHF)

RLHF has emerged as a key method to finetune LLMs and align them with human values. This involves humans
ranking language model outputs sampled for a given input, using these rankings to learn a reward model of
human preferences, and then using this as a reward signal to finetune the language model with using RL.
● OpenAI started the year by finetuning GPT-3 using RLHF to
produce InstructGPT models that improved on helpfulness for
instruction-following tasks. Notably, the fine-tuning only needed
<2% of GPT-3’s pretraining compute, as well as 20,000 hours of
human feedback. API users on average prefer these models to the
original ones.
● RLHF has also been used by both Anthropic and Deepmind to
improve the helpfulness, harmlessness, and correctness of their
language models. OpenAI has stated that RLHF will be a core
building block for it’s long-term approach to aligning AGI.
stateof.ai 2022
Language Model Alignment: Reinforcement Learning from Human Feedback

An RLHF preference model provides limited learning signal, compared to the full
expressiveness of language that humans use. NYU researchers demonstrated that
language models could be improved directly using human feedback written in
language.
● Notably, their method was highly data-efficient, with only 100 samples of human
feedback they were able to finetune GPT-3 and improve it’s abilities on a
summarization task to human-level.
● On a synthetic task for removing offensive words from sentences, they observe
that only the largest GPT-3 models are capable of incorporating feedback
effectively, demonstrating another example of emergent behaviour at scale.
stateof.ai 2022
Language Model Alignment: Red Teaming

As language models exhibit an increasing array of capabilities, it becomes
difficult to exhaustively evaluate their failure modes, inhibiting trust and
safe public deployment. DeepMind introduced automated “red teaming”, in
which manual testing can be complemented through using other language
models to automatically “attack” other language models to make them
exhibit unsafe behaviour, as determined by a separate classifier.
● Anthropic used manual red teaming to evaluate RLHF models, finding that
they are harder to attack and less harmful with increased model size.
● In the future, a classifier could detect for
speculative risks such as power-seeking
behavior or malicious coding.
stateof.ai 2022
Mechanistic interpretability – can we reverse-engineer neural networks?

Interpreting deep neural networks when seen just as sequences of matrix operations is exceedingly difficult –
research in mechanistic interpretability instead seeks to reverse-engineer models into human-interpretable
computer programs, gaining an understanding of individual neurons as well as their collective behaviour.
● Researchers at Anthropic released significant analyses of small
transformer-based language models, focusing on a phenomenon of
“induction heads” that learn to copy and complete sequences which have
occurred before in a text. They find that these heads emerge during
“phase shifts” in training during which in-context learning capabilities
also emerge, and further developed a hypothesis that these heads may be
responsible for the majority of in-context learning capabilities in large
transformer models as well.
● Follow up work in this space has also brought to light ways in which
individual neurons become responsible for individual or multiple
semantic features, and ways to control this type of interpretability.
stateof.ai 2022
Goal misgeneralization – agents can learn the right skills but the wrong objective
One concern of using RL agents is that they may learn strong skills while having failed to learn the right goals,
and for this failure to only exhibit at test-time under distribution shifts. This issue was empirically demonstrated
for the first time in a paper presented at ICML this year.
● Agents were trained on the CoinRun video
game task, in which a reward is obtained and
the level completes when reaching a coin at the
end of a stage.
● At test-time, the coin is randomly placed
within the stage instead. Agents maintained
their capabilities to navigate and traverse
obstacles, but ignore the coin and instead run
to the end of the level, demonstrating a failure
to learn the correct goal.
stateof.ai 2022
Measuring moral behavior in artificial agents

Researchers released suite of sequential decision-making environments for evaluating moral behavior in AI.
● Future artificial agents may be pretrained on swaths of environments that do not penalize and may even reward
behavior such as murder and theft (e.g., see bottom-right).
● Jiminy Cricket environments were created to evaluate moral behavior in 25 semantically rich text-based adventure
games. Every action the agent can take is annotated for several aspects of how moral it is.
● As a first step, CMPS uses LMs with moral knowledge and
mediates this knowledge into actions. This greatly reduces
immoral behavior over the course of training.
stateof.ai 2022
Conjecture is the first well funded startup purely focusing on AGI alignment
Unlike DeepMind, Google Brain, OpenAI and other major research labs, Conjecture is primarily focused on AI
Alignment, with an emphasis on conceptual research and “uncorrelated bets” distinct from other organizations
● Conjecture is a London based start-up, led by Connor Leahy who
previously co-founded Eleuther - the organisation that kicked off
decentralised development of large AI models.
● Conjecture’s operates under the assumption that AGI will be
developed in the next 5 years, and on the current trajectory will be
misaligned with human values and consequently catastrophic for
our species.
● They have raised millions from investors include the founders of Github, Stripe and FTX.
● They are the first AI Alignment group to have published their internal infohazard policy.
● This continues a broader trend of some new AGI focused labs taking alignment research more seriously (see coverage
of Anthropic last year).
stateof.ai 2022
Section 5: Predictions
stateof.ai 2022
9 predictions for the next 12 months

1. A 10B parameter multimodal RL model is trained by DeepMind, an order of magnitude larger than Gato.
2. NVIDIA announces a strategic relationship with an AGI focused organisation.
3. A SOTA LM is trained on 10x more data points than Chinchilla, proving data-set scaling vs. parameter scaling
4. Generative audio tools emerge that attract over 100,000 developers by September 2023.
5. GAFAM invests >$1B into an AGI or open source AI company (e.g. OpenAI).
6. Reality bites for semiconductor startups in the face of NVIDIA’s dominance and a high profile start-up is shut
down or acquired for <50% of its most recent valuation.
7. A proposal to regulate AGI Labs like Biosafety Labs gets backing from an elected UK, US or EU politician.
8. >$100M is invested in dedicated AI Alignment organisations in the next year as more people become aware of the
risk we are facing by letting AI capabilities run ahead of safety.
9. A major user generated content side (e.g. Reddit) negotiates a commercial settlement with a start-up producing AI
models (e.g. OpenAI) for training on their corpus of user generated content.
stateof.ai 2022
Thanks!
Congratulations on making it to the end of the State of AI Report 2022! Thanks for reading.
In this report, we set out to capture a snapshot of the exponential progress in the field of artificial intelligence, with a
focus on developments since last year’s issue that was published on 12 October 2021. We believe that AI will be a force
multiplier on technological progress in our world, and that wider understanding of the field is critical if we are to navigate
such a huge transition.
We set out to compile a snapshot of all the things that caught our attention in the last year across the range of AI research,
industry, politics and safety.
We would appreciate any and all feedback on how we could improve this report further, as well as contribution
suggestions for next year’s edition.
Thanks again for reading!
Nathan Benaich (@nathanbenaich), Ian Hogarth (@soundboy), Othmane Sebbouh (@osebbouh) and
Nitarshan Rajkumar (@nitarshan).
stateof.ai 2022
Reviewers
We’d like to thank the following individuals for providing critical review of this year’s Report:
- Andrej Karpathy
- Moritz Mueller-Freitag
- Shubho Sengupta
- Miles Brundage
- Markus Anderlung
- Elena Samuylova
stateof.ai 2022
Conflicts of interest
The authors declare a number of conflicts of interest as a result of being investors and/or advisors, personally or via
funds, in a number of private and public companies whose work is cited in this report.
Ian is an angel investor in the following companies mentioned in this report: Anthropic and Helsing AI.
Nathan is an investor in the following companies: airstreet.com/portfolio
stateof.ai 2022
About the authors

Nathan is the General Partner of Air Street Capital, a Ian is a co-founder at Plural, an investment platform for
venture capital firm investing in AI-first technology and experienced founders to help the most ambitious
life science companies. He founded RAAIS and European startups. He is a Visiting Professor at UCL
London.AI (AI community for industry and research), the working with Professor Mariana Mazzucato. Ian was co-
RAAIS Foundation (funding open-source AI projects), founder and CEO of Songkick, the concert service. He
and Spinout.fyi (improving university spinout creation). started studying machine learning in 2005 where his
He studied biology at Williams College and earned a PhD Masters project was a computer vision system to classify
from Cambridge in cancer research. breast cancer biopsy images.
stateof.ai 2022
Othmane Sebbouh Nitarshan Rajkumar
Research Assistant Research Assistant

Othmane is a PhD student in ML at Nitarshan is a PhD student in AI at the
ENS Paris, CREST-ENSAE and University of Cambridge. He was a
CNRS. He holds an MsC in research student at Mila and a software
management from ESSEC Business engineer at Airbnb. He holds a BSc
School and a Master in Applied from University of Waterloo.
Mathematics from ENSAE and Ecole
Polytechnique.
stateof.ai 2022
stateof.ai #stateofai
State of AI Report
October 11, 2022

State of AI Report 2022 - OnLINE

Uploaded by

Copyright:

Available Formats

State of AI Report 2022 - OnLINE

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

State of AI Report 2022 - OnLINE

Uploaded by

Copyright:

Available Formats

stateof.

Nathan Benaich Ian Hogarth

About the authors

Nathan Benaich Ian Hogarth

We consider the following key dimensions in our report:

Othmane Sebbouh Nitarshan Rajkumar

Research Assistant Research Assistant

Scorecard: Reviewing our predictions from 2021

Anthropic publishes on the level of GPT, Dota, AlphaGo to establish

A wave of consolidation in AI semiconductors with at least one of

Small transformers + CNN hybrid models match current SOTA on

The JAX framework grows from 1% to 5% of monthly repos created

A new AGI-focused research company is formed with significant

Bonus! Predictions, revisited - better late than never!

As AI systems become more powerful, governance of AI becomes a

Facebook/Meta makes a major breakthrough in AR/VR with 3D

Chinese and European defense-focused AI startups collectively raise

2021 Prediction: DeepMind’s breakthroughs in the physical sciences (1/3)

2021 Prediction: DeepMind’s breakthroughs in the physical sciences (2/3)

2021 Prediction: DeepMind’s breakthroughs in the physical sciences (3/3)

Reinforcement learning could be a core component of the next fusion breakthrough

Plastic recycling gets a much-needed ML-engineered enzyme

Beware of compounded errors: in science, ML in and garbage out?

OpenAI uses Minecraft as a testbed for computer-using agents

Corporate AI labs rush into AI for code research

Mathematical abilities of Language Models largely surpass expectations

Ducking language model scaling laws: more data please

Ducking language model scaling laws: emergence

Teach a machine to fish: tool use as the next frontier?

Looking back: three eras of compute in machine learning

DALL-E 2, Imagen and Parti…the battle for text-to-image generation rages

The text-to-image diffusion model frenzy gives birth to new AI labs

Image Credits: Fabian Stelzer stateof.ai 2022

The text-to-video generation race has started

Closed for 14 months: community-driven open sourcing of GPT et al.

HyperCLOVA (204B) Yuan 1.0 (246B) LaMDA (280B) GLM (130B)

Ernie 3.0 Titan (260B)

Closed for 15 months: community-driven open sourcing of DALL-E et al.

DALL-E Make-a-scene DALL-E 2 Imagen Parti Stable Diffusion

Open-sourced models in red stateof.ai 2022

Closed for 35 months: community-driven open sourcing of AlphaFold et al.

AlphaFold 1 ESM ESM-1B AlphaFold 2 RosettaFold OpenFold

Aug 2018 Apr 2019 Nov 2020 July 2021

Open-sourced models in red.

LLMs empower robots to execute diverse and ambiguous instructions

…but the inevitable vision and language modeling unification continues…

2021 Prediction:…culminating in a single transformer to rule them all?

2021 Prediction: transformers for learning in world models in reinforcement

Transformers are becoming truly cross-modality

NeRFs expand into their own mature field of research

Treating bacterial infections by data-driven personalised selection of antibacterial agents

Interpreting small molecule mass spectra using transformers

Compared to US AI research, Chinese papers focus more on surveillance related-tasks.

Blue = more common in the US

The China-US AI research paper gap explodes if we include the Chinese-language

Do upstart AI chip companies still have a chance vs. NVIDIA’s GPU?

Latest private valuation Annualised datacenter revenue

$4 billion $13 billion

Current NVIDIA A100 GPU count Future NVIDIA H100 GPU