Thursday Thoughts on AI + Law (7/13/23)

So much happened over the past two weeks! Here's a sampling of what I found interesting:

1. AI companies raised a lot of money in the first half of the year - roughly $14 billion! I mean, Inflection raised $1.3b itself at the end of last month. Roughly half of the money apparently went to companies in SF , fueling a hiring boom .? And there were lots of deals, too, including ones that caught VCs off-guard . Pitchbook has more details .
2. NYC’s law regarding biases in hiring-related algorithms went into effect last week, and it seems to be pleasing few . If it applied to companies in Ireland, there’d be some interesting results since a large percentage of surveyed Irish firms disclosed that they did not account for algorithmic biases in their use of AI tools.
3. Benedict Evans is one of the best thinkers on tech trends and his recent post on generative AI and its impact on work is a great read.
4. ICYMI: Over 150 executives from major European companies (think: Renault Group, The HEINEKEN Company, Airbus, Siemens, etc.) signed a letter arguing that the proposed AI Act is ineffective and could negatively impact competition, which might “jeopardise Europe’s competitiveness and technological sovereignty.”
5. In the future, AI-generated content will dominate the Internet (at least, in terms of volume) and this means that misinformation may run rampant .
6. No joke: Sarah Silverman is one of the litigants suing AI developers.
7. Batting 1.000? How MLB is using AI to evaluate prospects.
8. UN Secretary-General António Guterres argued that AI must be used to benefit all people and drive sustainable development . Similarly, Volker Türk, UN High Commissioner for Human Rights, argues that AI must be grounded in respect for human rights .
9. Here is a great essay on how to think about the evolution of creative works from Gutenberg to generative AI.
10. Greylock Capital updated their memo re: competitive moats to account for the state-of-play in the AI world.
11. While many business leaders express some degree of skepticism or fear about the use of AI at work, more than 70% of surveyed executives plan to introduce the use of AI for productivity in their organizations in the next three years.
12. Meta is reportedly pushing for more openness as part of their preferred AI regulatory approach.
13. Nicholas Thompson of The Atlantic and Yuval Noah Harari had a fascinating conversation about AI.
14. The German AI Association published their thoughts on the EU’s AI Act.
15. The European Council is pushing for more development of AI regulatory sandboxes.
16. Anthropic released Claude 2 .
17. Google launched Bard in the EU and Brazil , and the Verge dug into how Google/Deepmind approach AI . The humans who helped train it say they’re overworked and overtired.
18. OpenAI made Code Interpreter more widely available , which is a huge productivity booster .
19. China has promulgated a number of different AI regulations. Here’s how that process works .
20. In Costa Rica, on the other hand, lawmakers are using ChatGPT to generate laws to regulate AI.
21. Bank regulators don’t think they need new laws to regulate AI (and might be quite right).
22. A deadlocked FEC is being urged to develop regulations on deep-fake campaign materials, as AI makes dirty politics cheap .
23. AI is improving weather forecasting .
24. Some wealth management firms could go under as a result of widespread AI adoption.
25. The WSJ published a good article on the competitors coming for Nvidia (which is currently dominating the AI chip market).
26. Will widespread AI use mean a drop in patents approved , as more ideas are viewed as ‘obvious’?
27. AI powers relevancy models that power much of our Internet experience.
28. Elon Musk loves to blame AI for things but he also just started a very heterogeneous X.AI company .
29. People need to stop trying to use automated tools to ‘detect’ AI-generated content: it doesn’t work , and it discriminates against non-native English speakers . Then again, human judges also detect false positives when judging art .
30. Adobe is apparently lobbying for a federal law to prohibit AI-powered impersonation efforts.
31. The FTC is reportedly investigating OpenAI .
32. OpenAI extended a deal with Shutterstock. And Shutterstock is offering an AI-related indemnity (bold!).
33. Like other consulting/accounting firms, KPMG is spending heavily on AI . Relatedly, McKinsey published a “must read” list of 10 articles about AI.
34. Google is helping to fuel the UK’s interest in AI .
35. What if, in the future, nearly everything you see on the Internet is created by AI ? More on that .
36. OpenAI created a team to combat rogue AI .
37. Perhaps to stay ahead of legal issues in foundational models, Google announced a competition for machine ‘unlearning.’
38. The White House met with union leaders to discuss potential AI impacts on the workforce.
39. Tech (and other) companies are now (correctly) making risk factor disclosures relating to AI in their regular SEC filings.?
40. If one author had his way, some AI-related risks would result in tech execs facing jail time.
41. SemiAnalysis published a great post outlining the vulnerabilities of large players in the AI space (but also highlighting how much money is being thrown at gaining an advantage in the growing AI economy).
42. CEPA argues that the U.S. should avoid going down a path similar to the EU’s AI Act proposal. And members of the Federalist Society share a pretty similar view , as one might expect. Japan is also expected to take a lighter approach, and is subsidizing Japanese AI firms .
43. Another cool use of generative AI: figuring out how to make environmentally friendly materials .
44. Multi-modal models keep getting more impressive.
45. Unity’s CEO thinks AI will make video games way more awesome .
46. There has been all kinds of talk over the years about ‘digital natives.’ Gen-Z may be the first ‘AI native’ generation.
47. OpenAI decided to pause its “Browse with Bing” feature after discovering that some users were abusing the feature to get around paywalls.
48. If you’re GitHub and you’re helping automate some coding, you might be able to help make that code more secure and thus improve the overall security of the entire software and internet industries. And, as McKinsey points out, you can also dramatically improve productivity in some developer tasks.
49. ChatGPT enabling 4000 tokens is cool. LongNet enabling 1 billion tokens is way cooler.
50. Cisco’s CEO discussed the security dangers of unethical deployments of AI.
51. BingChat now has voice capabilities .
52. Google scored a major win in an AI privacy-related lawsuit in the Seventh Circuit.
53. Scientists are now creating animal ‘twins’ using AI to understand how their minds might work.
54. Eric Topol points out that medical AI is advancing very quickly . One example: treating brain cancer .
55. Speaking of brains, scientists used AI to map the brain of fruit flies .
56. Google updated their privacy policy to note that they will use scraped data from the Internet to train AI models. And Meta is probably going to use Threads to train its models .
57. Experts are still debating whether CDA Section 230 applies to AI-created works.
58. The UK Parliament’s Communications and Digital Committee is launching an inquiry into the risks/benefits of large language models.
59. Here’s a great summary of the “rise of the AI engineer .” And if you’re trying to become something of a ‘prompt engineer,’ here are some good tips for how to use LLMs to expand your horizons.
60. Researchers have found a way to start reducing hallucinations in LLMs .
61. AI is helping us preserve endangered languages. It is also helping us understand long-lost tongues as well.
62. It might be the case that only the most sophisticated platforms are able to effectively combat the misinformation onslaught that might arise from widely available LLMs.
63. Axios dug into how the U.S. is trying to stay ahead of China in the AI arms race.
64. Netflix is bringing AI to the green screen with a little magenta help.
65. Canada is working with the EU and the U.S. to help shape an AI Code of Conduct.
66. Not exactly surprising but, yes, well-capitalized companies are going to have an advantage in a resource-intensive battle for AI development and deployment.