Thursday Thoughts on AI + Law (7/20/23)

Another week, another bajillion news stories about AI and how society and governments are trying to figure out what to do about it.

As always, feel free to forward on or share with anyone else who might be interested in the topic.

1. As actors and writers strike, the LA Times correctly points out that AI is presenting a Napster-like moment for film.
2. Ensuring that his blog lives up to its name, Professor Ethan Mollick published a guide for how to get AI to do useful things.
3. The Verge takes media executives to task for how they are deploying AI for creative purposes.
4. Deepak Chopra asks what AI can do to help us understand the meaning behind creativity, sentience, and transcendence.
5. Senator Schumer will stage a series of AI education boot-camps for members of Congress in the fall/winter. If only it could happen sooner! If Congress doesn’t act, California might have to lead the way (like it did with privacy). Or the courts could make the rules…
6. LLaMa 2 is out. Nathan Lambert is pretty excited about it. And Microsoft is helping serve it to the world.
7. People are asking: is GPT-4 getting worse and, if so, is it because OpenAI might be taking a Ship of Theseus approach? It’s more complicated than it initially appears. Related: evidently, some details about how GPT-4 works were leaked.
8. Howie Xu makes a great parallel between the authors of Google’s Transformer paper and Shockley’s Traitorous Eight (if you’re not familiar with the story, it’s worth reading).
9. Some convenience-oriented apps (e.g., Uber, DoorDash) have been using AI to set contractor pay (which is causing problems for some contractors).
10. AI is a means to an end, not the end: writers have seen remarkable improvements as a result of using AI tools to enhance their drafting process.
11. Gary Gensler is warning everyone about the potential dangers of AI in financial markets.
12. The Washington Post published a copy of what is purported to be the CID the FTC sent to OpenAI. Their coverage of the topic is pretty good.
13. Modern Diplomacy argues that Indonesia has the potential to be a major player in the Sino-American AI battle.
14. There are so many different LLMs available now. How does one choose which one to use for a particular context?
15. Time might be nigh for Apple’s AI play - surely, they couldn’t keep muttering ‘machine learning, not AI’ forever.
16. Neurotech and AI, coupled together, could effectively destroy privacy. In the meantime, Wired published a list of the different ways you can protect your privacy in the current AI paradigm.
17. The UK is working hard to become an AI superpower. Mustafa Suleyman argues that, to accomplish that, the country needs to undergo a cultural transformation.
18. Careful, now. The Israeli military is reportedly embedding AI in deadly weapons systems.
19. Also bad: developers released an unethical-by-design, hacker-oriented LLM known as WormGPT.
20. Also really bad: an AI-powered dragnet to tell police who is ‘driving like a criminal.’ More on the dangers of AI for policing here.
21. Maybe bad? AI-generated tattoo art.
22. Political deepfakes are a growing threat and for some reason federal/state election officials aren’t prepared enough for it.
23. AI-written South Park is … weird.
24. Azeem Azhar published a guide for getting started with GPT Code Interpreter.?
25. Gizmodo is leaning into AI-written articles. Google is developing tools to make AI journalism easier. And OpenAI as a potential savior of the local news? Didn’t see it coming but it’s a welcome development. Meanwhile, the CJR dug into past precedent for how media companies might deal with AI developers.
26. The club continues to grow: Samsung and IBM have been sued over their AI tools.
27. The Carnegie Council is offering up a framework for how governments can approach AI governance.
28. EU leaders are openly pushing U.S. leaders to act on AI.
29. The UN Security Council is talking about AI now and recommending regulation.
30. The Canadian Centre for Cyber Security promulgated guidance on the appropriate use of generative AI.
31. The Ada Lovelace Institute issued its thoughts in response to the UK government’s proposals on AI.
32. Gary Marcus proposed ‘AI governance in a box.”
33. The NYT explains the chips war going on between the U.S. and China.
34. Some of the biggest chips are being used to power supercomputers for AI development.
35. How should ‘frontier’ AI models (i.e., those that are highly capable foundation models that might be dangerous) be regulated?
36. As the AI landscape continues to expand, older, established players are starting to feel the heat of competition.
37. The Markup dove into the topic of how to account for the environmental costs of AI development and deployment.
38. Emad Mostaque predicted that most outsourced coding jobs will disappear within two years.
39. Here are some details on the new Chinese regulations regarding generative AI. And here’s what they might teach regulators elsewhere.
40. Microsoft unveiled its pricing structure for AI-empowered M365 products.
41. I’m not sure who the target audience really is here but CIO Magazine published a ‘should you build or buy’ article about AI.
42. The NYT talks with content creators who are pulling back from creating in light of their works being ingested to train AI models. Authors are also writing to AI developers on the topic.
43. What does the AI Act mean for open source AI developers?
44. Data reviewers who helped train large language models are asking lawmakers to take steps to regulate the outsourcing of training.
45. OpenAI is reportedly concerned that its models might be able to re-identify faces.
46. Axios published a useful reminder that much of how we experience generative AI reflects human-mediated design and implementation choices.
47. Connecticut is implementing a law that will regulate how the state’s government uses AI.
48. UBS flagged that AI can identify serious ESG risks that might otherwise go undetected.
49. Catholic universities are pushing for ethical adoption of AI tools.
50. The Wisdom 2.0 crew is hosting a Wisdom and AI summit in October.
51. The trend towards using ChatGPT instead of Stack Overflow continues.
52. HuggingFace is raising at a $4 billion valuation.
53. A convergence of trends: AI and unaffordable housing. Related: A NYC lawmaker used AI tools to draft a (not very good?) law on housing.
54. AI is helping the U.S. government track supply chains (including the use of forced labor).
55. AI is helping detect potential cardiac events for firefighters, and helping scientists discover treatments for metabolic diseases.
56. Some companies are pushing back against regulation of AI in healthcare.
57. Jamaica’s government set up an AI task force.
58. The NYT dove deeper into the question of AI inventorship.
59. Two of my favorite topics: what happens if you follow AI-powered suggestions for trail running training?
60. AI played a role in the coverage of this year’s Wimbledon.
61. MITRE is advertising the results of their survey on AI governance and trust.