Here are this weeks links, just sneaking into the week:
- Stack Overflow and OpenAI partner to enhance OpenAI models and share them right back. In the deal, OpenAI will benefit from StackOverflow's deep reservoir of coding knowledge for training. StackOverflow will have access to build its own AI experience. This is a logical move for both parties, but StackOverflow will still need to solve existential problems down the line, and some of its users are reacting by deleting their content. While content licensing will bring in a revenue boost for StackOverflow, and they can no doubt do the deal with other AI model builders, the big challenge in the long term is the slowdown in people asking questions. The more people who stop asking and others who stop answering, the faster the drop in the relevance of the content. StackOverflow and sites like it will need to find a way to retain that question-and-answer dynamic.
- Your dating ‘AI concierge’ will soon date hundreds of other people’s ‘concierges’ for you. This is a low-information clickbait article, but it is based on comments from the founder of Bumble, Whitney Wolfe Herd. This is the quote: “If you want to get really out there, there is a world where your [AI] dating concierge could go and date for you with another dating concierge,.” “Truly. And then you don’t have to talk to 600 people. It will scan all of San Fransisco for you and say: ‘These are the three people you really outta meet.'”. It's hard to imagine that bot-bot interactions are not happening in many use cases like this: scheduling meetings, getting reservations, first contact for sales, and (yes) seeking romantic partners. People are already automating dating interactions without being open about it. The challenge is that a platform like Bumble will have difficulty differentiating what a true exchange of revealed preferences is versus their own matchmaking algorithm. How much will you let your agent tell white lies in order to get any dates at all? How will that escalate?
- AI Warfare. At the other end of the spectrum, Michael Spencer deep dives into the use of automated systems and military activities. It's a high-level overview, but the conclusion seems self-evident: no country can ignore the fact that any possible adversary may have new capabilities (particularly in the form of autonomous drones.). The conclusion for humanity should also be obvious: this is not just a zero-sum game; it is a deeply negative sum given, and we should be looking for ways to avoid everybody, not only to avoid wasting all these resources but certainly never using these weapons.
- Meta debuts new techniques for improving inference speed in diffusion models. The high cost of inference for diffusion models still means that things like real-time image creation are out of reach, but the field keeps evolving. Meta is likely already one of the largest properties serving image generation and amongst the most motivated to push the cost down.
- Apple's iPad advert and the flattening of culture. Apple's beautifully produced advert showed a large collection of artistic and musical equipment being crushed to result in ... the new iPad Pro. Not everyone agrees it was a bad ad, but the strong reactions have an underlying theme: the idea that diverse, varied cultures were being crushed into a sleek, uniform device. This speaks to a deep-rooted fear that technology (and now AI) will diminish the human experience by smashing it into vanilla conformity. There is a very real danger to us, and it has indeed become easier to create bland facimilies of "culture." However, I don't think it's inevitable that we'll lose the old ways: new tech broadens possibilities to many more people, at one point, things like the electric guitar we also have new tech and above all hopefully we will have the prosperity to deepen and extend the old arts. The Apple ad at least highlights how important it is that we keep the whole spectrum from old to new alive. Let's not crush it in an industrial press!
Wishing you a wonderful Sunday!
