Friday (?!) Thoughts on AI + Law - 6/23/23

Better late than never, right? This week's edition goes out a day late as a result of the week getting away from me :)

Hopefully there's something of interest here for your weekend reading.

1. The FT reports that large technology firms developing and deploying AI solutions are negotiating terms with publishers regarding use of their content.
2. There’s a common perception that GPT models are bad at math. If prompted appropriately, though, one paper suggests that a GPT model can (strongly) pass the requirements to graduate from MIT with a degree in Math or EECS. Others (rightfully) question the methodology here.
3. Yann LeCun has some deep thoughts on the topic of AI and AGI and he’s put together a document to explain them.
4. A Stanford-Princeton team has a set of recommendations for how to regulate generative AI.
5. The Verge has a great long piece on how AI is built and the role that different stakeholders (including labeling teams) play in the process.?
6. Bloomberg has a great visualization of how AI tools may exacerbate human biases.
7. The World Economic Forum has a thought piece out describing how policymakers can learn from the GDPR when it comes to how to regulate AI.
8. The French government has some concerns regarding the state of the AI Act’s regulation and its potential impact on AI development in Europe. As it stands, a report out of Stanford suggests that most large generative AI models would not necessarily meet the AI Act’s standards. So the AI Act might confound France’s effort to be the AI hub of Europe.
9. Speaking of France, China is trying to cozy up to France to cooperate on various fronts (including AI).
10. China has a series of existing AI regulations. The government plans to release another one by the end of the year. Foreign Affairs explains why regulating AI in the U.S. won’t hold the country back vis-a-vis China. More on that topic here.
11. Wild: China has published an allow-list of deep-fake producers who align with CCP values.
12. In the U.S., it seems like every agency under the sun wants in on regulating AI. Perhaps that’s why the EU approach (creating a new regulatory body) might be preferable. Related: Congress is considering new bipartisan proposals for AI regulation. Chuck Shumer is working hard to steer that ship, perhaps with some input from the tech community. As Brookings points out, there are (at least) three major challenges to address in any successful regulation (it’s a very good read).
13. Meanwhile, the Singapore government indicated it will be holding back on regulating AI for now.
14. When the President comes to the West Coast, he comes to talk about AI.
15. A Texas-based VC argues that the AI Act proposal suggests that Europe is “not part of the future of the world, innovation-wise.” And that might mean fewer AI billionaires being minted there.
16. That could have been much worse: the lawyer who used ChatGPT to write his brief was fined $5,000.
17. A16Z has a blog post outlining what they view as the emerging standard for “AI tech stack.”
18. Norwegian consumer groups are pressing for regulators to apply existing laws to AI as they await AI-specific regulations. Other consumer groups in Europe are joining in.
19. Evidently, hundreds of thousands of ChatGPT credentials have been leaked on the dark web.
20. U.S.-based job posts for generative AI-related jobs are up significantly.
21. The IAPP is quickly becoming the AIPP (or, as it appears, the “IAPPAI”). The rapid pivot from privacy to AI in corporate governance is just like the move from IP to privacy around a decade or so ago…same maneuver, different shiny field-of-focus.?
22. Fascinating: financial firms that pore over incredible amounts of data to discern potential trends for investing are expecting to face increasing volumes of junk/fake data originating from AI tools.
23. Axios asks whether AI can be used to help us better understand ourselves (‘us’ meaning human society more broadly).
24. Meta: Amazon’s Mechanical Turk uses humans to crowdsource tools (humans-as-AI) but those humans are now increasingly using AI tools to complete those tasks (AI-as-humans-as-AI).
25. Stanford’s HAI digs into the meaning of “fair use” of copyrighted material in an AI age.
26. The FEC is digging into how to regulate political ads that use generative AI.
27. Orcas have been … orca-nizing in the oceans. But not this one - it’s an LLM from Microsoft :)?
28. Amazon is opening up their own generative AI playground. And teaching their massive workforce how to best realize opportunities (and avoid risks) relating to generative AI tools.
29. CIO has a feature looking at how the AI explosion is impacting IT hiring.
30. WIRED published an interesting read on Microsoft’s Satya Nadella and his views on AI.
31. You’ve come a long way, Etsy: from artisan wares to AI-powered junk.
32. Dropbox is investing $50m in an AI-oriented venture fund.
33. Celebrities can use AI to create additional ‘versions’ of themselves to scale their marketing efforts.
34. Axios reports researchers are concerned that AI could greatly reduce the barriers to entry for developing pathogens.
35. Sayash Kapoor and Arvind Narayanan published an insightful essay on the threats from AI-boosted disinformation on social media.
36. Cannes Lions 2023 indexed pretty heavily on AI.
37. Professor Ethan Mollick has a blog post that flags a great point: some of the advantages accruing to employees using generative AI are a result of them using AI tools discreetly. If employers knew that much of their work was being done via AI, in other words, it might negatively impact the employees’ long term prospects.
38. The UK is launching a task force to evaluate how the government should approach AI, and it appointed a tech entrepreneur to chair the group.
39. Silly AI, the Grammys are for humans only.
40. Meta AI leaders are trying to reassure people that existential fears about AI are overblown. They’re right, I think, but calling people’s fears ‘ridiculous’ is not a preferred way to make friends or influence people.
41. A lot is happening in the UAE: first, Falcon; now, dedication of an AI task force.
42. ChatGPT is helping drive lots of work in the pharmaceutical R&D space now.
43. The FT dives into the wild world of AI in hiring processes.
44. TIME magazine honors AI companies (among others) this week. They also breathlessly reported that OpenAI and other companies have–shockingingly–tried to influence the language in the laws that may regulate them.