Obenpai 2020

IN1654
Case Study
ObEN PAI:
Building a World of Personal AI Avatars
03/2020-6562
This case study was written by Lisa Duke, Research Associate, under the direction of Peter Zemsky,
Professor of Strategy, and Pavel Kireyev, Assistant Professor of Marketing, both at INSEAD. It is
intended to be used as a basis for class discussion rather than to illustrate either effective or ineffective
handling of an administrative situation.
To access INSEAD teaching materials, go to https://publishing.insead.edu/
Copyright © 2020 INSEAD
COPIES MAY NOT BE MADE WITHOUT PERMISSION. NO PART OF THIS PUBLICATION MAY BE COPIED, STORED, TRANSMITTED, TRANSLATED,
REPRODUCED OR DISTRIBUTED IN ANY FORM OR MEDIUM WHATSOEVER WITHOUT THE PERMISSION OF THE COPYRIGHT OWNER.
This document is authorized for use only in Lluis Soldevila's 20/21 - Digital Transformation (MIM) at ESADE - Barcelona from Oct 2020 to Jan 2021.
It had been five years since ObEN, a Los Angeles-based start-up, had been launched. The
company created personal AI avatars from image and voice data in two categories: consumer
avatars created by individuals on their smartphone, and celebrity-grade avatars that required more
data. The company used blockchain to secure the rights to the avatars and control their usage.
What started in 2014 as an idea in the head of Nikhil Jain, together with his partner Adam Zheng,
was now a business that was gaining traction in terms of partnerships and funding.
Thanks to a collaboration with Project PAI, they were pushing the boundaries of the technology
they used for data storage and digital rights in ways that blockchain – the basis of digital currencies
like Bitcoin – had not been used before. The latest, a wholly virtual avatar called Satoshi, who
presented news items on demand via smartphone, had been rolled out on Android and Apple
platforms (see Exhibit 4). The company was working with three major Asian media groups in
Korea, Japan and China to create avatars or PAI’s (personal artificial intelligence) of celebrities to
host TV shows and interact with consumers in ways previously unknown. ObEN was also testing
a promising project that aimed to have a demonstrable impact on patient health care.
The question facing Nikhil and his team was how to build ObEN/Project PAI into a sustainable
business. This was a brave new world – part science fiction – and the technological and public
acceptance challenges facing the team were significant: developing AI software that was free,
easy to use, and mirrored the individual; ensuring the data storage blockchain technology
protected data safety, and reducing the inherent inefficiencies of distributed networks. Perhaps
the biggest challenge was managing consumer perceptions of blockchain and AI, which in the
West were often confused with cryptocurrencies as a rapid rise in their prices had boosted global
awareness. Despite the hype around blockchain, subsequent fraud cases and fluctuations in the
value of cryptocurrencies had resulted in fears about data protection and trust. AI, too, raised
concerns about data privacy, cyber-attacks, surveillance and manipulation. Clearly, those
perceptions could not be changed overnight, nor by a single start-up, but for ObEN the immediate
concern was whether the company had a future. It would mean scaling and scaling fast – but how?
Building ObEN
Nikhil’s and Adam’s vision was to build a company that could offer anyone the ability to easily build
a PAI, using an app on their smartphone. Nikhil had young children at home, and they were used
to him reading to them, singing songs and talking with them. They missed him when he travelled,
which was often.
“I thought, what if there was a copy of me back home? A copy that looks like me, talks
in my voice, has a personality like mine, but in a digital format that my kids could
interact with when I wasn’t there. That was the inspiration and I believed that this kind
of problem exists for every parent and family member.”
Nikhil Jain, Co-Founder & CEO
The first step was a feasibility analysis. They discussed the idea with friends, entrepreneurs and
potential investors, asking whether it made sense. At a certain point, Nikhil realized they had
sufficient positive reactions – people saying they would buy it and use it – to go ahead. Investing
their own money, the pair started building a minimum viable product (MVP) – an app that could
mockup an avatar to show people what it looked and sounded like. It took eight months from
Copyright © INSEAD 1
coming up with the idea and doing the feasibility analysis to completing the MVP. During this
process they began inviting friends to invest, so that they could keep developing it.
They decided to call the new company ObEN – “oben” is German for “above” or “up” – representing
the hope that personal AI could be used on top of, or as an added layer to, existing technologies
in industries such as VR/AR, social media, entertainment and retail. The lower-case “b” was a flip
of “p” to mimic the design of an OPEN sign, signalling that the AI was always ready to assist, as
in “open for business”.
Attracting Investment
The pair continued to work on the technology as well as building a team. As a cutting-edge project
in terms of AI, it required top software talent. Their first employee, Mark Harvilla, joined as chief
engineer in January 2015. ObEN managed to secure enough small investors to keep going. It
would be two and half years before their first institutional investors, SoftBank and Tencent,
became interested.
“We were building towards the idea by building a team, and by building user groups to
show that the technology, if followed in terms of the roadmap, is feasible. That’s when
the institutional investors thought, ‘Yeah, these two guys are not just dreaming... they
are actually implementing their dreams with the help of technology and a good team
in place.’ That’s when they got interested and invested in us.”
Both Tencent and Softbank were heavily invested in the next generation of flexible technologies
that could be used to transform multiple industries, particularly AI. From ObEN’s perspective, they
both understood and supported its vision that personalized avatars could be a game-changer,
applicable to nearly every industry. As well as financially, they both supported ObEN by facilitating
partnerships and collaborations with other companies in their portfolios.
ObEN grew steadily, managing various technological challenges and gaining a total of $28.7m in
the A and B funding round (see timeline in Exhibit 1).
The Challenges of AI
A PAI is a digital copy of any individual. It looks and sounds like them and is capable of interacting
with people and performing tasks on their behalf. Through cross-language text-to-speech
technology, ObEN’s PAIs were able to speak multiple languages (even if the human version could
not) and display emotions and ‘visual speech’ – eye blinking, lip synching and head movements
that matched the words spoken. (Figure 1 shows the components of PAI, all built in-house).
Figure 1: The Components of PAI
Source: ObEN company documents
Consumer perceptions of AI, highlighted in the media, included concerns about data privacy,
cyber-attacks, surveillance and manipulation, as well as its broader implications such as job losses
and military usage, according to a survey of American respondents conducted by Oxford
University. 1 Ensuring that a PAI could not be used for malicious purposes by a third party was
critical in order to scale.
“We get asked a lot if Asia is more open to avatars than the West. The concept of
avatars has been around in Asia for quite some time now. Hatsune Miku is a Japanese
virtual idol [introduced in 2007] and is very famous across Asia. The difference
between us and Hatsune Miku is that most of the stuff in her is made by hand. What
we have done it to take the popularity of something like a Hatsune Miku and are
utilizing state-of-the-art proprietary AI technology to make the entire experience better,
faster and more economical, and making it available in mobile phones for mass
consumer usage. While Asia is more open, I feel that audiences in the US and Europe
are also getting comfortable with having avatars. I say this because there is such a big
usage of Google assistants or Siri and Alexa in everyone’s homes – they are
essentially avatars.”
The technology itself was the biggest challenge to overcome. For a PAI to evolve required lots
and lots of data and was computationally intensive. The team’s first task was to collect significant
volumes of data – visual data, speech/audio data – to support progress, testing and development.
To help with data collection, they partnered with local schools and colleges in Los Angeles. Data
would make the PAIs as real as possible.
Being able to replicate someone’s voice was much harder than replicating their facial likeness.
There were myriad languages and accents that had to be taken into consideration. Any
background noise when recording a voice could interfere with the vocal rendering. Not only was
1 Zhang, B. & Dafoe, A. (2019) Artificial Intelligence: American Attitudes and Trends, Governance AI, University of
Oxford, https://governanceai.github.io/US-Public-Opinion-Report-Jan-2019/index.html (accessed September 2019)
this a challenge for the developers, but a potential barrier to entry – they had to create something
that sounded like the consumer.
Beyond the technological requirements, ObEN faced similar challenges to other tech start-ups,
not least its location. Based in Los Angeles, attracting talent to develop the AI and think creatively
to deliver the concept was tough, particularly at the beginning when the lure of Silicon Valley for
the Big Tech AI players such as Google and Facebook was strong. Candidates had to be
convinced that they would not only get the perks, benefits and similar compensation (adjusted to
cost of living) but also get to work on technologies that were ahead of their time. However, being
small meant that potential employees would see faster delivery in terms of consumers using the
technology than at a large Silicon Valley firm. Moreover, since Nikhil and Adam believed that
ObEN should involve entertainment, the proximity with Hollywood was helpful. It still took the pair
a year to convince a company to believe that creating PAIs was a good idea.
“We were very fortunate that one of the largest Korean K-pop companies, SM
Entertainment, agreed to partner with us on a joint venture. They also invested in us
to create digital avatars of all their top K-pop celebrities.”
Even when they had overcome the initial challenges, Nikhil admitted - “It’s never ending.
Challenges happen every day. It’s like crisis management at the start of every day.”
Project PAI – Embracing Blockchain

While the AI technology was maturing, another issue arose: data ownership and protection.
Although it was now clearer that it was possible to create a PAI with a selfie and voice recording,
who ultimately owned it?
The team believed that the PAI was the individual’s own identity, not simply a mass of data. They
looked at how other companies were using personal data but could not identify one where it was
owned by the user. Given the opportunities for data misuse and unauthorized avatar usage, they
needed a preventive mechanism. Having looked into different databases, contractual and legal
mechanisms to resolve the issue, it appeared that blockchain, if used properly, might be a solution.
Basing an individual’s PAI data on a decentralized network would ‘raise the bar’ for anyone trying
to tamper with the data stored on it, albeit one drawback was that blockchain was computationally
intensive because of its decentralized structure.
Blockchain in Brief
Blockchain is a distributed ledger or database of transactions across a network of nodes (or

individual computers). It is essentially a large spreadsheet where anyone can add data or a
transaction that is then verified by every node. Once verified, the transaction data is added to each
node’s version of the ledger. Information already on the spreadsheet cannot be altered but can be
added to. Each node keeps a copy of the blockchain and the history of transactions and the
majority of nodes need to agree on the true copy to give it validity. Blockchain uses cryptography
whereby data of any size is input into a hash function and the output is a hash (a fixed-length
string of alphanumerics) that forms a unique digital fingerprint of the original data. 2 The output is
the same for any identical set of data, but for any change (however small) a new hash is created
– essentially a new fingerprint. Each blockchain transaction is run through a hash function to
create its unique identifier.
The five underlying principles of blockchain technology are: 3

• A distributed database where every party has access to the whole database and its history.
No single party controls the data and each party can verify the records of its transaction
partners.
• Transmission/communication is direct from peer-to-peer rather than through a central
node. Every node both stores and forwards information to all other nodes.
• Complete transparency. Users can remain anonymous or provide proof of identity to others.
Every user or node has a unique identifier made up of 30+ alphanumerics and can make their
real identity publicly available or use a pseudonym. Every transaction between users is visible
to anyone with access to the network.
• Records cannot be altered once in the database, because each record is linked to every
previous transaction. Algorithms are deployed to ensure that records are permanent, in
chronological order, and available to the entire network.
• Computational logic which means that transactions can be programmed by allowing users
to create algorithms and rules to automatically trigger transactions between nodes.
Innovating Blockchain
To test their hypothesis about the efficacy of blockchain to store and secure an individual’s data,
the ObEN team did further research into the blockchains of Bitcoin and Ethereum. Ethereum was
used by many developers for their applications. 4 Bitcoin was the oldest blockchain and had been
thoroughly vetted from a technical perspective. Building off Bitcoin would mean opening up ObEN
to the Bitcoin ecosystem, so they could adopt any downstream changes or updates.
They decided to adopt the PAI blockchain protocol, developed using an open-source approach by
the Project PAI Foundation. Project PAI served as an underlying blockchain technology to tie PAIs
to a particular user and track the PAI’s usage. ObEN was Project PAI’s first contributing innovator
which focused on creating AI applications that used – and commercialized – the PAI blockchain
technology. As Chief Engineer, Mark Harvilla was a core contributor to Project PAI’s open-source
technology, helping to plan and realize specific AI-related research projects.
To operate transactions on blockchain, Project PAI decided to launch an initial coin offering, which
would enable consumers to have their PAIs verified and then added to the blockchain by miners
who offered computational power to process changes to the blockchain. Miners were
authenticators who verified the validity of transactions occurring on the blockchain using a proof
of work (PoW) procedure. The most common consensus mechanism (agreement of transactions)
2 Morrissey, E. (2018). Breaking Down the Blockchain. Capital One Tech, September 12,
https://medium.com/capital-one-tech/breaking-down-the-blockchain-f4b87422481b
3 De Filippi, P. (2017). What Blockchain Means for the Sharing Economy. Harvard Business Review, March 15
4 Sharma, T. K. (2019). Top 10 Blockchain Platforms You Need To Know About. Blockchain Council, May 11,
https://www.blockchain-council.org/blockchain/top-10-blockchain-platforms-you-need-to-know-about/
was proof of work, described in a Project PAI white paper 5 as follows: “Creation of a valid block
required a miner to demonstrate proof of work by offering the solution to a mathematical problem
that is ‘hard’ to compute but ‘easy’ to verify. Generally speaking, the probability a given miner adds
a new block to the blockchain is proportional to the amount of computational power to which he or
she has access.”
When a solution was found, miners received PAI Coin (a digital currency) as compensation. The
maximum supply of PAI Coin was 2.1 billion, 6 with an initial placement of 1.47 billion. The starting
block reward was 1500 PAI Coin for miners, which would be halved every 210,000 blocks or
approximately every four years (Exhibit 2 illustrates the PAI Coin and credits system). Among the
challenges of PoW were high physical resources, usage/cost, and decreasing user engagement,
brought about by increasingly centralized hash power over time. 7
There were three layers of technology in Project PAI that were relevant to ObEN – authentication,
data storage, and computation. The data storage protocol allowed the company to secure any
kind of data associated with a PAI on the decentralized network that was, by design, tamper-proof.
The computation piece was a cutting-edge project exploring AI and machine learning-related
computations on the protocol that governed how blockchain worked.
“Think of it this way. The way Bitcoin works is a distributed network of co-called miners,
who compete with one another to verify whatever Bitcoin transactions occur. The
protocol of Bitcoin replaces a central entity such as a bank which has full control to
manage and verify the transaction if, for example, I go to Starbucks and buy a coffee
with my credit card. If I could pay with Bitcoin, then the transaction would be broadcast
to tens of thousands or more of nodes that make up the blockchain Bitcoin network.
These nodes all have to agree that the transaction is valid and that I actually have the
money that I’m trying to spend. The key point here is that the way the protocol is
designed is very computationally intensive. To do something as simple as buying a
coffee, at the Bitcoin network level, there’s really massive amounts of electricity being
expended in order to verify these transactions in this way – you could argue that is
wasteful. What we are trying to do is to reengineer the protocol so that this computation
can be directed or recycled towards something that’s useful.”
Mark Harvilla, Chief Engineer, ObEN / Project PAI
Using the blockchain at the simplest level involved sending PAI Coin – a data transaction. There
were two forms (see Figure 2). For data storage, the individual sent PAI Coin to themselves and
embedded in the transaction a link to the data to be published on the blockchain. The transaction
was accompanied by a fingerprint of the data. The individual could decide whether to keep it
private or not. Once the transaction was sent and confirmed it existed in perpetuity on the
blockchain and anyone could look up the transaction to verify that the data published was not
tampered with by comparing it against the fingerprint. The data sharing transaction was the same
thing except that the individual would be sending the transaction to someone else on the
blockchain. The transaction could be a contract, for example, for use of images. The individual
created the transaction and sent it to the recipient and embedded the link to the data they wanted
5 Harvilla, M. & Du, J. Prospective Hybrid Consensus for Project PAI

6 Compared to Bitcoin’s 21 million maximum.
7 Harvilla, M. & Du, J. Prospective Hybrid Consensus for Project PAI
to share. The sender could encrypt the data with the recipient’s public key before publishing it on
the blockchain.
Figure 2: PAI Data Storage and Data Sharing Transactions
Source: Case authors
Project PAI was a hard code fork of Bitcoin. ObEN chose it, reasoning that Bitcoin was the most
secure and oldest blockchain, with a lot of development having gone into it. There was also legal
precedent around ownership of assets, which made it attractive. One innovation created by the
engineers from Project PAI was to add a decentralized storage component that did not exist in
Bitcoin or other blockchains. This enabled users to store their data on a multitude of services in a
decentralized fashion, while keeping the ownership in the form of pointers and a timestamp. This
was invaluable for storing the PAI avatar data.
When an individual created their PAI, ObEN provided the ability to store the fundamental data on
the blockchain ledger as well as creating a Know Your Customer process using a third party that
created a form of provenance that the individual was actually the owner of the data/PAI. The third
party, such as a notary, approved that too. These were then stored on the blockchain with a time
stamp and a pointer to that data. This legal ownership was a smart contract mechanism that meant
individuals were able to approve others to utilize their PAI.
One example was a celebrity who created a PAI, such as famous Chinese actress, Dai Chunrong
(see Figure 3), who created a PAI that was promoted on the My Future TV show. They could
approve a company to use their PAI for publicity or even, as Dai Chunrong explained on the show,
her PAI could now dub for her shows in post-production (rather than Dai being present). All these
actions could be recorded on the blockchain too.
Figure 3: Dai Chunrong and her PAI on My Future TV
Source: https://medium.com/oben/avatars-of-the-future-pai-on-tv-24a5bf7e33d2
“The Right of Publicity law assures you that if your likeness is misused by someone
without your permission then you can enforce your copyright and trademark protection
and seek damages. In this case, the entire mechanism of pointers and timestamps for
your likeness becomes very effective. Also, the ability to contractually use it, coupled
with a digital payments mechanism is a useful tool for our entertainment clients.”
Another innovation that came from Project PAI was the development of a new rights management
system to utilize the AI rights of celebrities. These rights were recorded on the PAI blockchain too.
The celebrity PAIs could sing songs that the celebrity had never sung or be shown in an image.
These rights and associated contractual permissions for use were now available on the blockchain
with every action being recorded on the chain.
The Challenges of Blockchain
Using blockchain raised various challenges. One issue that Mark’s team had not yet resolved was
the limitation in scaling the technology. There was always a trade-off in terms of efficiency due to
the massively decentralized network. On the one hand, it allowed for data security and maintaining
verifiable data that was difficult to tamper with. On the other, it was inherently inefficient because
of the need to keep many copies of the same data. The network demanded considerable
computational power and resources – even the electricity required to keep the network functioning
made it wasteful.
Another issue ObEN faced was dispelling the myth that “blockchain = cryptocurrencies and
scams”. The goal was to educate consumers and companies alike that, despite the bad publicity,
blockchain was a fundamentally sound technology with amazing utility beyond cryptocurrencies.
Decentralization was more complex than a simple, efficient, centralized system controlled by a
single entity. It was a behaviour that humans had grown used to over decades.
“There’s a practical challenge to mainstream adoption: How do you get regular

consumers to appreciate the seriousness of having your private keys and being
responsible for all your data so a central entity can’t tamper with your data or steal
your data or steal your funds? They also can’t help you recover them if you make a
mistake or lose something.”
Mark Harvilla, Chief Engineer, ObEN / Project PAI
Use Cases
The original vision was to create technology that anyone could use to create a PAI, but this was
more a long-term dream than reality. So was the idea that the software would be free, and
monetization would occur through in-app purchases, because it would take time to scale up.
However, celebrity PAIs provided a business model for ObEN that would generate revenues as
well as publicity, as Nikhil put it, “Entertainment is definitely number one for us.” This was also an
area where regulation was lower and the Asian consumer base was more open to try new things.
The company was able to create joint ventures with three leading Asia-based companies.
- In June 2017, they partnered with SM Entertainment, Korea’s largest entertainment

company to create the AI Stars joint venture (https://oben.me/s-m-entertainment-and-
oben-form-joint-venture-to-create-the-worlds-first-celebrity-ai-agency/ ).
This was the world’s first celebrity AI platform to create celebrity PAIs and manage their AI
intellectual property.
- Another joint venture was formed with Japan’s largest entertainment conglomerate,
Yoshimoto, to build the PAIs of Japan’s top celebrities.
- A first test in China in 2018 was with a well-known girl band SNH48,
(https://www.globenewswire.com/news-release/2018/12/27/1678463/0/en/Oben-
Collaborates-With-SNH48-to-Create-World-s-First-PAI-Human-Music-Video.html). ObEN
introduced the PAIs of the female celebrities in the video, which featured the PAIs
interacting with their human counterparts and singing. Fans’ feedback was strong and
positive (see Exhibit 3).
- In early 2019, ObEN was selected by CCTV of China (one of the world’s largest TV
channels by audience size) to create avatars for their TV hosts
(https://www.forbes.com/sites/bernardmarr/2019/01/29/the-worlds-most-watched-tv-
show-will-be-hosted-by-artificial-intelligences/#684965c968de). It was a huge success,
with hundreds of millions of views of the New Year show featuring the PAIs (see Exhibit 3).
“Since then, our business has been growing quite a lot. This was very promising to the
team that our technology could become mass scale, starting with celebrities but then
going back to consumers and saying – ‘Hey, just as celebrities can have their own
avatar, every person in the world can also have one and can start using it for things
we might not have time or the ability to do.”
Another PAI the company showcased in 2019 was Satoshi – a fully automated and autonomous
3D avatar of a newscaster who could read content on demand in different languages. Satoshi was
named after the mythical figure(s) who originally created the Bitcoin blockchain. ObEN’s Satoshi
could be found on an app available to the public (see Exhibit 4). The IP was recorded on the
blockchain to promote the company and showcase its technology. The company believed that
experiences such as news-on-demand read by a friendly avatar like Satoshi was something that
could be scaled and rolled out across other business segments.
Meanwhile, they also started to work in healthcare on a project with MedStar Health Research
Institute and its affiliate MedStar Georgetown University Hospital. MedStar wanted to study cost-
efficient, high-engagement methods of managing and monitoring patients with heart failure post-
surgery. During the three-month study, as the patient was recovering at home, they would interact
daily with a healthcare PAI via ObEN’s PAI Health app, created for the study. Unless there was a
medical emergency that required patients to be re-hospitalized, they would not typically be seen
by a doctor. During the recovery period, hospital staff continued to monitor the patient’s symptoms
and condition to see if there were adverse signs that could require a change in medication or an
early check-up by a physician, as well as how patients were managing their lifestyle and their
medication. Instead of interacting with a human healthcare provider, patients simply interacted
with the PAI and answered daily survey questions that helped monitor their health, habits, and
symptoms.
ObEN introduced an app with an avatar based on the local demographic of patients in Washington,
D.C. who were largely African Americans (see Exhibit 5). The avatar put a set of questions to the
patient, running a survey of their healthcare and health. Based on the answers, the app might alert
the clinical nurse on call if major abnormalities were suspected. The nurse could either verify that
no further action was required, or elevate the urgency and contact a physician, ask the patient to
return to hospital, or call emergency services to collect the patient. While still in its early days, the
app saw high engagement from patients who completed the survey on a regular basis. This was
encouraging given that these patients were older, generally underserved and less tech-savvy than
other groups.
The company was also working on a customer service avatar for China’s Huachuang Securities,
whereby a PAI was created of a fictitious persona who could provide investment reports or any
kind of customer service to users. Another use case in work was in E-sports and live streaming.
A further opportunity was with CCTV to create a digital audience of individual avatars for a
broadcast host competition. Viewers would be able to video in and use ObEN’s face-tracking
technology and could interact as an avatar with the hosts during the competition. Chinese live-
streaming company Douyu used ObEN’s avatar in their livestream of Apple’s iPhone 11
announcement. During the livestream, the human host broadcast with his PAI counterpart live.
This leveraged ObEN’s AI Visual Speech tech, taking audio input during the broadcast and
automatically animating the mouth and facial movements of the avatar.
Building Competitive Advantage
ObEN was not the only company building avatars. The team believed their competitive advantage
came first from packaging the technology into a simple and user-friendly consumer experience.
Second, using proprietary technology, ObEN was one of the only companies in the world that
actually created its technologies in-house rather than having to depend on an external supplier’s
“full stack”. This, they believed, was the key to moving fast with customers and delivering avatar
experiences on mobile phones for consumers. The technology also required significantly less
data, which meant less computing power and storage.
Competitors such as Bitmoji (https://www.bitmoji.com ) offered consumers the opportunity to build

avatars with just faces, while Apple offered the opportunity to become an animal that mimicked
the user’s voice and facial expressions through its animoji app (https://support.apple.com/en-
us/HT208190 ) or as a person using Memoji
(https://support.apple.com/en-us/HT208986 ).
Soul Machines, a New Zealand-based company (https://www.soulmachines.com), created digital

avatars (headshots) mainly as customer service assistants for companies such as the Royal Bank
of Scotland and the Australia & New Zealand Banking Group. 8 Soul Machines focused on creating
the highest quality avatars rather than the overall integration and AI aspects that ObEN focused
on.
ObEN was the only full stack avatar offering that created realistic versions of a person in a simple
way. While ObEN had an advantage, the team was mindful of the giants – Google, Microsoft,
Apple, Amazon, Samsung, Tencent, Alibaba, and Facebook – watching trends in the market that
ObEN was creating. The challenge was to scale to a sizeable level before the giants entered and
displaced ObEN.
Adding digital payments was something the company was looking into, with in-app purchases
such as dressing up a PAI in different clothes or hairstyles. Other ideas were having a dance-off
competition or a duet with a favourite celebrity’s PAI. Nikhil was passionate about how far he could
push the technology into other areas such as healthcare and education. He hoped to support
teachers in developing countries by creating avatars who could educate children.
Challenging the Future

Although still only five years old, by 2019 the company had filed more than 21 AI patents in the
US and offered a full stack AI technology with PAIs that could sing and dance. It had grown into a
team that was 70% PhDs and MSc holders, with diverse expertise including: computer vision,
deep learning, linguistics, machine learning, speech synthesis and speech recognition. Its
adoption of Project PAI’s technology was pushing the boundaries of blockchain through proving
hypotheses and developing the technology beyond cryptocurrencies.
ObEN built a unique infrastructure across the US, China and Asia-based cloud computing
environment, so users could create their PAIs wherever they lived. As well as joint-ventures with
various media companies in Korea, China, and Japan, it opened an office in Japan to support its
growth there. Ultimately, the company needed to continue to refine its avatars through greater
data and machine learning algorithms to improve what it was offering to businesses and
consumers alike.
The bigger challenges were to convert its proof of concept projects into user and revenue-
generating business use cases, while continuing to scale and focus on the most promising
segments. The company had strong ties in the entertainment industry with key players in Asia,
8 Okutsu, A. (2018). New Zealand startup Soul Machines puts human face on AI. Nikkei Asian Review, October 2
while the health project had demonstrated that there was room for new segments to benefit from
avatars. However, one test project did not mean segment-wide acceptance. There was much work
to do to push this further.
Nikhil’s dream of consumers creating their own PAIs still seemed far off. While PAIs were
embraced in Asia, there was still much to do to change perceptions of blockchain in the West.
How could they achieve Western acceptance? Could they go it alone? What other options should
they explore? For their own growth, could ObEN scale beyond entertainment? Which industries
and markets should it target? What else did the team need to do?
Exhibit 1
ObEN Timeline of Major Milestones
Feb 24, 2014 ObEN Founded

Jan 6 - 9, ObEN demo’s Ben the WeChat Bot (a touchscreen robot concierge) at the Linq
2016 Hotel, in collaboration with WeChat and Caesars Entertainment for CES 2016
July 2016 HTC selects ObEN as one of only 33 companies to participate in its Vive X
accelerator inaugural class
Nov 3, 2016 ObEN closes $7.7M investment as part of Series A funding (Total Series A funding
$13.7M)
Mar 14, 2017 SoftBank Ventures Korea (now rebranded SoftBank Ventures Asia) invests in
ObEN
Jun 28, 2017 ObEN and S.M. Entertainment launch joint venture, A.I. Stars, a celebrity AI
agency to manage celebrity AI IP
Jul 18, 2017 ObEN secures $5M in strategic funding, led by Tencent
Nov 4, 2017 ObEN adopts the Project PAI blockchain to secure and verify its PAIs
Jan 9, 2018 ObEN raises $10M from K11
Mar 21, 2018 ObEN selected for Dubai Future Accelerators
April 18, 2018 ObEN debuts AI art concierge at K11 Shanghai, created with voice and image of
K11 Chairman Adrian Cheng, for spring art exhibition
Jul 9, 2018 ObEN selected for Softbank Innovation programme working with Softbank, Salin,
and Wrench on VR/AR social platform of live entertainment
Sep 25, 2018 ObEN debuts first Personal AI of celebrity Aijia from SNH48, at the Shanghai
World AI Conference
Oct 1, 2018 ObEN and Lippo Group jointly form the Pacific Blockchain Research Center for
blockchain development in South East Asia
Dec 27, 2018 ObEN debuts first PAI/human music video created in collaboration with SNH48
Jan 29, 2019 ObEN creates PAI hosts, viewed by millions during CCTV’s annual Network Spring
Festival Gala
Feb 25, 2019 ObEN and Korean VR/AR company Salin sign MOU to created new social and
entertainment VR/AR platform featuring PAI
Mar 4, 2019 ObEN and Medstar Research Institute launch study into using PAIs to interact with
chronic heart-failure patients
May 14, 2019 ObEN launches PAI News app at Consensus 2019, first platform with content
delivered by AI avatars
Mar 21, 2019 ObEN and Korean live streaming giant AfreecaTV sign MOU to collaborate on PAI
livestreaming content based on popular human broadcast jockeys
Jun 12, 2019 ObEN premieres the “Little E” AI analysts, created for Huachuang Securities, at
the 2019 Big Data Expo
Source: ObEN
Exhibit 2
PAI Coin and Credits System
Source: Project PAI Motivation white paper: Decentralized, Peer-enabled AI Identities for the AI Economy, October 2017
Exhibit 3
SNH48 Video with PAIs
CCTV’s Network Show Featuring the PAIs of Four Famous Celebrities
Source: ObEN
Exhibit 4
Satoshi – The Presenter of PAI News
Source: ObEN
Exhibit 5
MedStar Nursing PAI
Source: ObEN

Obenpai 2020

Uploaded by

Copyright:

Available Formats

Obenpai 2020

Uploaded by

Document Information

Original Title

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

Obenpai 2020

Uploaded by

Copyright:

Available Formats

IN1654

Figure 1: The Components of PAI

Project PAI – Embracing Blockchain

Blockchain is a distributed ledger or database of transactions across a network of nodes (or

The five underlying principles of blockchain technology are: 3

5 Harvilla, M. & Du, J. Prospective Hybrid Consensus for Project PAI

Figure 2: PAI Data Storage and Data Sharing Transactions

Source: Case authors

The Challenges of Blockchain

“There’s a practical challenge to mainstream adoption: How do you get regular

- In June 2017, they partnered with SM Entertainment, Korea’s largest entertainment

Building Competitive Advantage

Competitors such as Bitmoji (https://www.bitmoji.com ) offered consumers the opportunity to build

Soul Machines, a New Zealand-based company (https://www.soulmachines.com), created digital

Challenging the Future

Feb 24, 2014 ObEN Founded

CCTV’s Network Show Featuring the PAIs of Four Famous Celebrities

You might also like