Friday Links: AI's Impact on Marketplaces, Deepfake Scams and Copyright Claims over Song Lyrics

The biggest lesson from this week: deepfake scams have a higher chance of working if they lean into an outrageous personality

Friday Links: AI's Impact on Marketplaces, Deepfake Scams and Copyright Claims over Song Lyrics

Hi all - I’ve been working hard on a longer form article, but it’s already Friday, and it’s not entirely done. So.. weekly links first. The longer form will come a little later!

  • The AI First Marketplace. NfX shared an optimistic take on the opportunities for AI in Marketplace businesses. NfX generally has excellent content, and I like what’s in here (there’s a helpful framework for the places where AI could play a role). I’m also excited about how AI could transform marketplace businesses. However, I think the article also misses some real challenges AI will bring: 1) some parts of some markets may collapse because what used to require a human now can be done with a tool (e.g., basic logo design). 2) in places where AI tools emerge as project design or management aids (e.g., DIY project planners) and then make supply agreements with a choice of partners on the backend, they may entice buyers to bypass marketplaces entirely. It also misses another potentially massive factor: AI will ultimately become a client of some markets, and markets should plan for this with API interfaces.
  • Adept releases a multi-modal LLM as open source. Fuyu-8B can take both images and text as inputs. While that doesn’t seem like a significant change, it’s the first viable model like this as open source. The ability to interpret charts and photos will be important in general knowledge work.
  • A deepfake video of a Mr. Beast Ad went viral on TikTok. Not that deep fakes are real news, but it’s interesting how something that would typically not be believable (10,000 people getting an iPhone Pro for $2) becomes potentially credible when you put the words in the mouth of someone like Mr. Beast. He has done some genuinely whacky stuff. The more outrageous you are, the more people might believe deepfakes of you.
  • Mark Andreessen published the Techno-Optimist’s Manifesto.  It’s hard not to include this piece in the links since it’s kicked up such a storm. Plenty of people have put out a response. No doubt Mark achieved his aim, which was most likely to promote debate and open the Overton window on the span of that debate. I agree with a lot of the points in the manifesto - we genuinely should be optimistic and work on challenging problems, even if, at times, there are risks. On the other hand, growth at all costs, forever, is not something I think is either necessary or sane. We already think the world population will probably begin to decline or stagnate by about 2100; we’d better get good at enjoying pockets of stasis as a species as well.
  • Music publishers sue Amazon-backed AI company over song lyrics. AI and copyright will get more days in court with this new lawsuit. The complaint includes training, but also ‘that Claude illegally reproduces the lyrics by request, and in response to “a whole range of prompts that do not seek Publishers’ lyrics,” including “requests to write a song about a certain topic, provide chord progressions for a given musical composition, or write poetry or short fiction in the style of a certain artist or songwriter”.’ My guess is that training will eventually be determined to be part of fair use, whereas the latter… likely will not. Interesting, though, that suit is not against Bard or OpenAI. Presumably, they are structured in such a way as to reproduce fully formed lyrics when not explicitly asked for them.

Happy Friday!

Midjourney: Steampunk robots at the market