Compute Deserts yield winners in the world's first AI Olympiad; TIME's most influential people in AI; OpenAI faces its competition; UK, US and EU sign world's first AI treaty

Salesforce pivots to AI agents; AI amplifies change management issues; chatbots distort reality; is AI great at making art? Personhood Credentials will help us figure out who is human online

The Oxford Internet Institute has published a paper highlighting the uneven distribution of computing around the world, identifying three categories of countries when it comes to access to cloud AI infrastructure:

• Compute North: 17 countries, including the United States and China, that host a significant number of data centers equipped with the GPUs most relevant to AI development such as foundational model training, namely Nvidia A100s and H100s GPUs. All of these first-tier countries, apart for India, are located in the so-called Global North.

• Compute South: 13 countries that host compute infrastructure more suited for AI deployment (such as V100 GPUs for inference) rather than for development. All of these countries are situated in the Global South, save for Switzerland.

• Compute Deserts: All of the remaining countries in the world. These countries host no public cloud AI compute at all, whether for training or for deployment. For them, building AI services means relying on infrastructure located in foreign jurisdictions. The Compute Desert contains a number of rich countries, but it also contains all of the world’s lower middle-income and lower-income countries, following the International Monetary Fund’s classification.

Specifically for the US, The Information published a database this week, looking at existing and planned state-of-the-art AI data centers from companies such as Microsoft, OpenAI, Meta and xAI. In total, the facilities in that database could cost more than $50 billion to build, including $35 billion worth of AI chips made by Nvidia, and many more billions to operate.

Global compute divide between the Compute North, Compute South, and the Compute Desert, based on public cloud GPU compute

It’s with these two maps in mind that I tuned in to watch the closing ceremony of the International Olympiad in Artificial Intelligence (IOAI) which took place on August 14.

The IOAI was held for the first time this year and brought together high school kids from 34 countries and territories, with a total of 44 teams competing for the gold medal (each country could send up to two teams.) The students had to complete a scientific round at home, and then traveled to Bulgaria for a second, practical-focused round.

Countries in blue sent at least one team to the IOAI

Looking at the list of winners is fascinating because it is dominated by the countries from the Compute Desert, including my home country of Romania.

It made me wonder: ten years from now, will those countries stand a chance against the Compute North? And how many of this year’s winners will choose to stay in their Compute Desert, given they presumably want to continue studying (and eventually developing) AI?

And now, here are this week’s news:

❤️Computer loves

Our top news picks for the week - your essential reading from the world of AI

• Time: The 100 Most Influential People in AI 2024

• WSJ: Companies’ AI Bets Are Reaching Astronomical Heights. Why the C-Suite Likes Its Odds Anyway.

• The Information: Introducing the AI Data Center Database

• MIT Technology Review: How “personhood credentials” could help prove you’re a human online

• New Yorker: Why A.I. Isn’t Going to Make Art

• Evening Standard: AI is exciting for artists, the possibilities are limitless

• WSJ: The Threat to OpenAI Is Growing

• The Guardian: ‘He was in mystic delirium’: was this hermit mathematician a forgotten genius whose ideas could transform AI – or a lonely madman?

• WSJ: What’s Ahead for Artificial Intelligence

• The Atlantic: Chatbots Are Primed to Warp Reality

• TechCrunch: AI brings a whole new dimension to the challenge of organizational transformation

• Fortune: Marc Benioff has declared a ‘hard pivot’ to autonomous AI agents. Will it be enough for Salesforce to thrive in the generative AI era?

⚙️Computer does

AI in the wild: how artificial intelligence is used across industry, from the internet, social media, and retail to transportation, healthcare, banking, and more

• The Guardian: M&S using AI as personal style guru in effort to boost online sales

• The Verge: Google tests its ‘Ask Photos’ AI assistant that understands what’s in your pictures

• FT: AI breakthrough raises hopes for better cancer diagnosis

• The Verge: Zillow’s upgraded AI search will show you more homes you can’t afford

• BBC: AI cameras spot toddlers not wearing seat belts

• Business Insider: Blackstone is looking to supercharge employee productivity with AI-powered search

• Business Insider: The CEO of edtech startup Headway breaks down how the company used AI tools to improve its ad performance by 40%

🧑‍🎓Computer learns

Interesting trends and developments from various AI fields, companies and people

• Reuters: Eli Lilly partners with AI-focused Genetic Leap to develop RNA-based drugs

• The Verge: Bill Gates has a good feeling about AI

• The Verge: YouTube is making new tools to protect creators from AI copycats

• TechCrunch: German LLM maker Aleph Alpha pivots to AI support

• VentureBeat: OpenAI says it reached 1 million business users

• Business Insider: Leaked messages show early challenges for Amazon's big AI product and concern about losing customers to Microsoft

• BBC: AI's solution to the 'cocktail party problem' used in court

• Bloomberg: Altman Infrastructure Plan Aims to Spend Tens of Billions in US

• Axios: AWS VP says AI says its time for AI to get practical

• The Information: Microsoft Customers Pause on Office AI Assistant Due to Budgets, Bugs

• WSJ: Anthropic Makes Play for Business Customers

• VentureBeat: Lightspeed L.A. reaches agreement with SAG-AFTRA on AI Protections

• Fortune: 1 in 5 workers are ‘underground’ AI users—here’s why they’re keeping it a secret

• Fortune: Indonesia’s second-largest telecoms company wants to launch its own local language AI model by the end of the year

• Axios: Teachers still can't trust AI text checkers

• TechCrunch: Spotter launches AI tools to help YouTubers brainstorm video ideas, thumbnails and more

• New York Times: OpenAI, Still Haunted by Its Chaotic Past, Is Trying to Grow Up

• WSJ: Microsoft Rolled Out AI PCs That Can’t Play Top Games—and There’s No Quick Fix

• VentureBeat: 71% of leaders prefer hiring candidates with AI skills over those with the relevant industry experience

• VentureBeat: DeepMind’s GenRM improves LLM accuracy by having models verify their own outputs

• Reuters: AI likely to weigh on oil prices over the next decade, Goldman says

• FT: Huawei’s bug-ridden software hampers China’s efforts to replace Nvidia in AI

• Fortune: Honeywell’s CEO reveals how AI will give his company a competitive edge—and what won’t

• Fortune: Canva says its new AI features justify raising subscription prices by 300%

• Business Insider: Elon Musk is putting his AI chips to work — and he's catching up with Mark Zuckerberg

• MIT Technology Review: What this futuristic Olympics video says about the state of generative AI

• MIT Technology Review: Here’s how ed-tech companies are pitching AI to teachers

• TechCrunch: The case against AI art

• Forbes: Tech Firms Are Keeping Users In The Dark On AI’s Climate Costs

• TechCrunch: ‘Emotion AI’ may be the next trend for business software, and that could be problematic

• WSJ: Mickey Mouse Smoking: How AI Image Tools Are Generating New Content-Moderation Problems

• VentureBeat: AI is growing faster than companies can secure it, warn industry leaders

• VentureBeat: OpenAI gives developers more control over AI assistants

• FT: Dating apps develop AI ‘wingmen’ to generate better chat-up lines

• AP: AI may change your job but it won’t eliminate many, labor experts say

• Business Insider: It's getting harder to make big leaps at the frontier of AI. There will be huge winners and losers.

• Business Insider: AI companies that say AGI is close are using dubious definitions to make that claim, AI pioneer says

• MIT Technology Review: A new way to build neural networks could make AI more understandable

• Sifted: “We can control AI” — Brunch with DeepMind’s first policy chief, Verity Harding

• VentureBeat: Meta’s Transfusion model handles text and images in a single architecture

• The Information: Meta AI Has More Than 185 Million Weekly Active Users, Zuckerberg Says

• VentureBeat: Cohere just made Command R smarter. Here’s why businesses should care

• Washington Post: Why killer AI is such an alluring horror villain