📊 AI: add details about investment figures limitations to our data pages

[veronikasamborska1994]

hi @paarriagadap! v small PR. I am just adding some details on our investment figures and wanted a second opinion on whether these are clear enough.

[owidbot]

Quick links (staging server):

Site	Admin	Wizard

Login: ssh owid@staging-site-ai-datapages

chart-diff: ✅

No charts for review.

data-diff: ❌ Found differences

+ Dataset garden/antibiotics/2024-10-23/tracss
+ + Table tracss
+   + Column laws_antimicrobials_terrestrial_2_8_2
+   + Column law_antimicrobials_aquatic_2_8_3
+   + Column law_antimicrobials_terrestrial_growth_promotion_2_8_5
+   + Column monitoring_consumption_human_3_2
+   + Column amr_surveillance_human_3_3
+   + Column monitoring_sales_use_animals_4_5_a
+   + Column amu_data_submission_whoah_4_5_b
+   + Column woah_reporting_options_4_6
+   + Column surveillance_amr_terrestrial_4_7
+   + Column surveillance_amr_aquatic_4_8
+   + Column surveillance_amr_food_5_3
+   + Column monitoring_amr_water_6_3
= Dataset garden/artificial_intelligence/2024-07-16/cset
  = Table cset
    ~ Column disclosed_investment (changed metadata)
+       +     - The dataset includes only certain types of equity investments (e.g., venture capital, private equity, mergers, and acquisitions) focused on privately held, AI-focused companies. It excludes non-equity financing like debt and grants, as well as publicly traded companies, thereby excluding major public companies, including Big Tech firms.
+       +     - CSET's uses a broad approach to identifying AI-related companies (using keyword and industry tag criteria) which means it may classify some companies as AI-focused that are not traditionally seen as such, while potentially missing some that might fit under other definitions.
+       +     - Due to the lack of disclosure for many investment values, CSET’s estimated metrics are calculated using median investment values for similar transactions, which provides approximations but introduces uncertainties.
+       +     - Assigning nationality based on headquarters location simplifies analysis but can miss nuances like cross-border corporate structures or varied investor origins.
+       +     - One-time events like major acquisitions can skew yearly figures, and economic conditions (e.g., interest rates, market sentiment) can affect investment trends independently of AI-specific factors.
+       +     - The methodology doesn’t comprehensively capture activities like those of public AI-related companies (e.g., NVIDIA, TSMC), corporate R&D in AI, government or public sector investments, and other significant areas like data centers, hardware deployment, and research costs. These omissions mean that the dataset might miss large portions of AI investment across broader categories.
+       +     - Given the focus on traditional corporate finance deals, the dataset's scope is limited, and the actual scale of global AI investments, especially those outside of private equity transactions, may differ substantially from what’s represented.
-       -     World aggregate does not include data for Micronesia, Tonga, Samoa, Kiribati, Fiji, Papua New Guinea, Palau, Tuvalu, Gibraltar, Jersey, Kosovo, Moldova, Isle of Man, Andorra, Montenegro, San Marino, Liechtenstein, Monaco, Vatican City, Afghanistan, Kyrgyzstan, Laos, Hong Kong, Bhutan, Brunei Darussalam, Maldives, Syria, North Korea, Myanmar, Timor-Leste, Nepal, Turkmenistan, Palestine, Yemen, Kuwait, Cape Verde, Equatorial Guinea, Swaziland, Namibia, Central African Republic (the), Angola, Ethiopia, Niger, Benin, Gabon, Gambia, Rwanda, Burkina Faso, São Tomé and Príncipe, Burundi, Guinea, Guinea-Bissau, Cameroon, Sierra Leone, Lesotho, Somalia, Chad, Liberia, Libya, South Sudan, Congo, Sudan, Malawi, Togo, Mali, Djibouti, Mauritania, Eritrea, Mozambique, Comoros, Antigua and Barbuda, Bolivia, Suriname, Nicaragua, Bahamas, Saint Vincent and the Grenadines, Grenada, Guyana, Haiti, Honduras, Cuba, Turks and Caicos Islands, Saint Lucia, and Dominica.
+       +     - World aggregate does not include data for Micronesia, Tonga, Samoa, Kiribati, Fiji, Papua New Guinea, Palau, Tuvalu, Gibraltar, Jersey, Kosovo, Moldova, Isle of Man, Andorra, Montenegro, San Marino, Liechtenstein, Monaco, Vatican City, Afghanistan, Kyrgyzstan, Laos, Hong Kong, Bhutan, Brunei Darussalam, Maldives, Syria, North Korea, Myanmar, Timor-Leste, Nepal, Turkmenistan, Palestine, Yemen, Kuwait, Cape Verde, Equatorial Guinea, Swaziland, Namibia, Central African Republic (the), Angola, Ethiopia, Niger, Benin, Gabon, Gambia, Rwanda, Burkina Faso, São Tomé and Príncipe, Burundi, Guinea, Guinea-Bissau, Cameroon, Sierra Leone, Lesotho, Somalia, Chad, Liberia, Libya, South Sudan, Congo, Sudan, Malawi, Togo, Mali, Djibouti, Mauritania, Eritrea, Mozambique, Comoros, Antigua and Barbuda, Bolivia, Suriname, Nicaragua, Bahamas, Saint Vincent and the Grenadines, Grenada, Guyana, Haiti, Honduras, Cuba, Turks and Caicos Islands, Saint Lucia, and Dominica.
        ?    ++
    ~ Column disclosed_investment_summary (changed metadata)
+       +     - The dataset includes only certain types of equity investments (e.g., venture capital, private equity, mergers, and acquisitions) focused on privately held, AI-focused companies. It excludes non-equity financing like debt and grants, as well as publicly traded companies, thereby excluding major public companies, including Big Tech firms.
+       +     - CSET's uses a broad approach to identifying AI-related companies (using keyword and industry tag criteria) which means it may classify some companies as AI-focused that are not traditionally seen as such, while potentially missing some that might fit under other definitions.
+       +     - Due to the lack of disclosure for many investment values, CSET’s estimated metrics are calculated using median investment values for similar transactions, which provides approximations but introduces uncertainties.
+       +     - Assigning nationality based on headquarters location simplifies analysis but can miss nuances like cross-border corporate structures or varied investor origins.
+       +     - One-time events like major acquisitions can skew yearly figures, and economic conditions (e.g., interest rates, market sentiment) can affect investment trends independently of AI-specific factors.
+       +     - The methodology doesn’t comprehensively capture activities like those of public AI-related companies (e.g., NVIDIA, TSMC), corporate R&D in AI, government or public sector investments, and other significant areas like data centers, hardware deployment, and research costs. These omissions mean that the dataset might miss large portions of AI investment across broader categories.
+       +     - Given the focus on traditional corporate finance deals, the dataset's scope is limited, and the actual scale of global AI investments, especially those outside of private equity transactions, may differ substantially from what’s represented.
-       -     World aggregate does not include data for Micronesia, Tonga, Samoa, Kiribati, Fiji, Papua New Guinea, Palau, Tuvalu, Gibraltar, Jersey, Kosovo, Moldova, Isle of Man, Andorra, Montenegro, San Marino, Liechtenstein, Monaco, Vatican City, Afghanistan, Kyrgyzstan, Laos, Hong Kong, Bhutan, Brunei Darussalam, Maldives, Syria, North Korea, Myanmar, Timor-Leste, Nepal, Turkmenistan, Palestine, Yemen, Kuwait, Cape Verde, Equatorial Guinea, Swaziland, Namibia, Central African Republic (the), Angola, Ethiopia, Niger, Benin, Gabon, Gambia, Rwanda, Burkina Faso, São Tomé and Príncipe, Burundi, Guinea, Guinea-Bissau, Cameroon, Sierra Leone, Lesotho, Somalia, Chad, Liberia, Libya, South Sudan, Congo, Sudan, Malawi, Togo, Mali, Djibouti, Mauritania, Eritrea, Mozambique, Comoros, Antigua and Barbuda, Bolivia, Suriname, Nicaragua, Bahamas, Saint Vincent and the Grenadines, Grenada, Guyana, Haiti, Honduras, Cuba, Turks and Caicos Islands, Saint Lucia, and Dominica.
+       +     - World aggregate does not include data for Micronesia, Tonga, Samoa, Kiribati, Fiji, Papua New Guinea, Palau, Tuvalu, Gibraltar, Jersey, Kosovo, Moldova, Isle of Man, Andorra, Montenegro, San Marino, Liechtenstein, Monaco, Vatican City, Afghanistan, Kyrgyzstan, Laos, Hong Kong, Bhutan, Brunei Darussalam, Maldives, Syria, North Korea, Myanmar, Timor-Leste, Nepal, Turkmenistan, Palestine, Yemen, Kuwait, Cape Verde, Equatorial Guinea, Swaziland, Namibia, Central African Republic (the), Angola, Ethiopia, Niger, Benin, Gabon, Gambia, Rwanda, Burkina Faso, São Tomé and Príncipe, Burundi, Guinea, Guinea-Bissau, Cameroon, Sierra Leone, Lesotho, Somalia, Chad, Liberia, Libya, South Sudan, Congo, Sudan, Malawi, Togo, Mali, Djibouti, Mauritania, Eritrea, Mozambique, Comoros, Antigua and Barbuda, Bolivia, Suriname, Nicaragua, Bahamas, Saint Vincent and the Grenadines, Grenada, Guyana, Haiti, Honduras, Cuba, Turks and Caicos Islands, Saint Lucia, and Dominica.
        ?    ++
    ~ Column estimated_investment_summary (changed metadata)
+       +     - The dataset includes only certain types of equity investments (e.g., venture capital, private equity, mergers, and acquisitions) focused on privately held, AI-focused companies. It excludes non-equity financing like debt and grants, as well as publicly traded companies, thereby excluding major public companies, including Big Tech firms.
+       +     - CSET's uses a broad approach to identifying AI-related companies (using keyword and industry tag criteria) which means it may classify some companies as AI-focused that are not traditionally seen as such, while potentially missing some that might fit under other definitions.
+       +     - Due to the lack of disclosure for many investment values, CSET’s estimated metrics are calculated using median investment values for similar transactions, which provides approximations but introduces uncertainties.
+       +     - Assigning nationality based on headquarters location simplifies analysis but can miss nuances like cross-border corporate structures or varied investor origins.
+       +     - One-time events like major acquisitions can skew yearly figures, and economic conditions (e.g., interest rates, market sentiment) can affect investment trends independently of AI-specific factors.
+       +     - The methodology doesn’t comprehensively capture activities like those of public AI-related companies (e.g., NVIDIA, TSMC), corporate R&D in AI, government or public sector investments, and other significant areas like data centers, hardware deployment, and research costs. These omissions mean that the dataset might miss large portions of AI investment across broader categories.
+       +     - Given the focus on traditional corporate finance deals, the dataset's scope is limited, and the actual scale of global AI investments, especially those outside of private equity transactions, may differ substantially from what’s represented.
-       -     World aggregate does not include data for Micronesia, Tonga, Samoa, Kiribati, Fiji, Papua New Guinea, Palau, Tuvalu, Gibraltar, Jersey, Kosovo, Moldova, Isle of Man, Andorra, Montenegro, San Marino, Liechtenstein, Monaco, Vatican City, Afghanistan, Kyrgyzstan, Laos, Hong Kong, Bhutan, Brunei Darussalam, Maldives, Syria, North Korea, Myanmar, Timor-Leste, Nepal, Turkmenistan, Palestine, Yemen, Kuwait, Cape Verde, Equatorial Guinea, Swaziland, Namibia, Central African Republic (the), Angola, Ethiopia, Niger, Benin, Gabon, Gambia, Rwanda, Burkina Faso, São Tomé and Príncipe, Burundi, Guinea, Guinea-Bissau, Cameroon, Sierra Leone, Lesotho, Somalia, Chad, Liberia, Libya, South Sudan, Congo, Sudan, Malawi, Togo, Mali, Djibouti, Mauritania, Eritrea, Mozambique, Comoros, Antigua and Barbuda, Bolivia, Suriname, Nicaragua, Bahamas, Saint Vincent and the Grenadines, Grenada, Guyana, Haiti, Honduras, Cuba, Turks and Caicos Islands, Saint Lucia, and Dominica.
+       +     - World aggregate does not include data for Micronesia, Tonga, Samoa, Kiribati, Fiji, Papua New Guinea, Palau, Tuvalu, Gibraltar, Jersey, Kosovo, Moldova, Isle of Man, Andorra, Montenegro, San Marino, Liechtenstein, Monaco, Vatican City, Afghanistan, Kyrgyzstan, Laos, Hong Kong, Bhutan, Brunei Darussalam, Maldives, Syria, North Korea, Myanmar, Timor-Leste, Nepal, Turkmenistan, Palestine, Yemen, Kuwait, Cape Verde, Equatorial Guinea, Swaziland, Namibia, Central African Republic (the), Angola, Ethiopia, Niger, Benin, Gabon, Gambia, Rwanda, Burkina Faso, São Tomé and Príncipe, Burundi, Guinea, Guinea-Bissau, Cameroon, Sierra Leone, Lesotho, Somalia, Chad, Liberia, Libya, South Sudan, Congo, Sudan, Malawi, Togo, Mali, Djibouti, Mauritania, Eritrea, Mozambique, Comoros, Antigua and Barbuda, Bolivia, Suriname, Nicaragua, Bahamas, Saint Vincent and the Grenadines, Grenada, Guyana, Haiti, Honduras, Cuba, Turks and Caicos Islands, Saint Lucia, and Dominica.
        ?    ++
    ~ Column investment_estimated (changed metadata)
+       +     - The dataset includes only certain types of equity investments (e.g., venture capital, private equity, mergers, and acquisitions) focused on privately held, AI-focused companies. It excludes non-equity financing like debt and grants, as well as publicly traded companies, thereby excluding major public companies, including Big Tech firms.
+       +     - CSET's uses a broad approach to identifying AI-related companies (using keyword and industry tag criteria) which means it may classify some companies as AI-focused that are not traditionally seen as such, while potentially missing some that might fit under other definitions.
+       +     - Due to the lack of disclosure for many investment values, CSET’s estimated metrics are calculated using median investment values for similar transactions, which provides approximations but introduces uncertainties.
+       +     - Assigning nationality based on headquarters location simplifies analysis but can miss nuances like cross-border corporate structures or varied investor origins.
+       +     - One-time events like major acquisitions can skew yearly figures, and economic conditions (e.g., interest rates, market sentiment) can affect investment trends independently of AI-specific factors.
+       +     - The methodology doesn’t comprehensively capture activities like those of public AI-related companies (e.g., NVIDIA, TSMC), corporate R&D in AI, government or public sector investments, and other significant areas like data centers, hardware deployment, and research costs. These omissions mean that the dataset might miss large portions of AI investment across broader categories.
+       +     - Given the focus on traditional corporate finance deals, the dataset's scope is limited, and the actual scale of global AI investments, especially those outside of private equity transactions, may differ substantially from what’s represented.
-       -     World aggregate does not include data for Micronesia, Tonga, Samoa, Kiribati, Fiji, Papua New Guinea, Palau, Tuvalu, Gibraltar, Jersey, Kosovo, Moldova, Isle of Man, Andorra, Montenegro, San Marino, Liechtenstein, Monaco, Vatican City, Afghanistan, Kyrgyzstan, Laos, Hong Kong, Bhutan, Brunei Darussalam, Maldives, Syria, North Korea, Myanmar, Timor-Leste, Nepal, Turkmenistan, Palestine, Yemen, Kuwait, Cape Verde, Equatorial Guinea, Swaziland, Namibia, Central African Republic (the), Angola, Ethiopia, Niger, Benin, Gabon, Gambia, Rwanda, Burkina Faso, São Tomé and Príncipe, Burundi, Guinea, Guinea-Bissau, Cameroon, Sierra Leone, Lesotho, Somalia, Chad, Liberia, Libya, South Sudan, Congo, Sudan, Malawi, Togo, Mali, Djibouti, Mauritania, Eritrea, Mozambique, Comoros, Antigua and Barbuda, Bolivia, Suriname, Nicaragua, Bahamas, Saint Vincent and the Grenadines, Grenada, Guyana, Haiti, Honduras, Cuba, Turks and Caicos Islands, Saint Lucia, and Dominica.
+       +     - World aggregate does not include data for Micronesia, Tonga, Samoa, Kiribati, Fiji, Papua New Guinea, Palau, Tuvalu, Gibraltar, Jersey, Kosovo, Moldova, Isle of Man, Andorra, Montenegro, San Marino, Liechtenstein, Monaco, Vatican City, Afghanistan, Kyrgyzstan, Laos, Hong Kong, Bhutan, Brunei Darussalam, Maldives, Syria, North Korea, Myanmar, Timor-Leste, Nepal, Turkmenistan, Palestine, Yemen, Kuwait, Cape Verde, Equatorial Guinea, Swaziland, Namibia, Central African Republic (the), Angola, Ethiopia, Niger, Benin, Gabon, Gambia, Rwanda, Burkina Faso, São Tomé and Príncipe, Burundi, Guinea, Guinea-Bissau, Cameroon, Sierra Leone, Lesotho, Somalia, Chad, Liberia, Libya, South Sudan, Congo, Sudan, Malawi, Togo, Mali, Djibouti, Mauritania, Eritrea, Mozambique, Comoros, Antigua and Barbuda, Bolivia, Suriname, Nicaragua, Bahamas, Saint Vincent and the Grenadines, Grenada, Guyana, Haiti, Honduras, Cuba, Turks and Caicos Islands, Saint Lucia, and Dominica.
        ?    ++


Legend: +New  ~Modified  -Removed  =Identical  Details
Hint: Run this locally with etl diff REMOTE data/ --include yourdataset --verbose --snippet

Automatically updated datasets matching weekly_wildfires|excess_mortality|covid|fluid|flunet|country_profile|garden/ihme_gbd/2019/gbd_risk are not included

Edited: 2024-10-28 16:39:30 UTC
Execution time: 23.79 seconds

[paarriagadap]

Hi @veronikasamborska1994! I left my comments below. It's clear in general, but maybe some of these corrections can help

[paarriagadap]

I wonder if readers would understand what CSET is. Maybe you can use "the source" or something similar

[paarriagadap]

Maybe you can give examples of these major public firms

[paarriagadap]

Something like this sounds better:

The origin of the investments is based on headquarters location. This simplifies analysis but can miss nuances like cross-border corporate structures or varied investor origins.

[paarriagadap]

I think this one is somewhat related to the first one. Maybe you can merge them and put that long text first?

[paarriagadap]

I think this one can also be combined with the others I say, and you can write something along the lines of

The data doesn't include x y z, so it may differ substantially from what is represented.

[paarriagadap]

Also don't forget to replicate these changes in the other grapher steps

[paarriagadap]

To follow the format, it would be better to have A private investment is

[paarriagadap]

Maybe the parenthesis is not necessary if you put this right below the description of the categories

[paarriagadap]

This should probably be the last bullet point

[paarriagadap]

I would put this point first, as it refers especifically to the data

[paarriagadap]

But I actually I see there is not such disaggregation of the data, so you should probably remove it

[paarriagadap]

This data does not show these categories