Biden's AI Executive Order
On Monday, President Biden signed a sweeping Executive Order on AI and launched AI.gov, a new site promoting policies and AI recruitment.
The bottom line:
Key themes include limiting large model training, adding more AI talent, and directing government agencies to think about AI.
Most AI companies will not be affected by this EO (yet). Foundation model developers will be impacted, along with infrastructure-as-a-service platforms and federal contractors.
We will undoubtedly see much more regulation on the back of this EO. But it's too early to say whether the government is stifling innovation and/or adequately accounting for AI risks.
Dive deeper:
Sunak's AI Safety Summit
The UK hosted its AI Safety Summit on Wednesday and Thursday, with a substantial guest list spanning academics, think tanks, government officials, and AI companies.
All the announcements I could find:
The Bletchley Declaration, an agreement of 28 countries to collaborate on mitigating “frontier AI risk.”
The UK AI Safety Institute (and the announcement of a US version), led by Ian Hogarth and Yoshua Bengio.
A £225 million government investment into Isambard-AI, a UK supercomputer.
An agreement from leading AI companies to allow governments to test their latest models before public release.
News that the next AI Safety Summit will be held six months from now in South Korea, and France will host the third summit in a year.
A "fireside chat" between the Prime Minister and Elon Musk, for some reason.
And while not yet an official announcement, Sunak is reportedly working on an OpenAI-powered chatbot for UK citizens to pay taxes an access pensions.
Elsewhere in AI geopolitics:
Voice cloning tools are allowing Indian politicians to send personalized messages ahead of elections.
New US rules may force Nvidia to cancel over $5 billion in GPU orders from Chinese companies.
The UN announced a new advisory body to address AI governance issues.
And the G7 is expected to announce a code of conduct for companies developing advanced AI.
Fearmongering
Playing out alongside all of the regulations has been a fierce debate among AI experts. Both Yann LeCun and Andrew Ng have accused doomsayers of having ulterior motives. Some of the accused, like Demis Hassabis, are pushing back.
Why it matters:
The debate is whether AI CEOs actually believe AI poses an existential risk, or if that narrative is an effort to get governments to clamp down on AI - and to cement their current leads.
To quote Ben Thompson, "if you accept the premise that regulation locks in incumbents, then it sure is notable that the early AI winners seem the most invested in generating alarm."
And with so much recent government action, many fear that sweeping regulations will stifle innovation, strange open-source, and put the future of AI in the hands of a few.
Elsewhere in AI anxiety:
A look at Justine Bateman's fight against generative AI in Hollywood.
Viral deepfakes of the Israel-Hamas war are causing people to doubt real images and video.
And Scarlett Johansson takes legal action against an AI ad that cloned her voice without her permission.
Things happen
A rogue AI company is putting GPUs into international waters. New Jersey high school students create deepfake porn of their classmates. LinkedIn's GPT-4-powered "job coach" available to Premium users. The first new Beatles song since 1995, made with AI. The latest AlphaFold model is more useful for drug discovery. AI Seinfeld is broken, maybe forever. Judge pares back copyright lawsuit against Midjourney and Stable Diffusion. Alibaba's Tongyi Qianwen 2 LLM, with industry-specific variants. A profile of Ilya Sutskever, Chief Scientist at OpenAI. Microsoft 365 Copilot is here (for enterprise users only). Sally Ignore Previous Instructions. Brave's new AI assistant, Leo. Copying Angry Birds with nothing but AI. Google commits up to $2B in Anthropic funding. ChatGPT can now work with your PDFs. Phind is (supposedly) better than GPT-4 at coding. AI detectors are destroying writers' livelihoods. The word of the year is AI.