3 ms·

For sure. If I were in charge of AI for the US, I'd prioritize having a known good and best-in-class LLM available not least for national security reasons; OAI

For sure. If I were in charge of AI for the US, I'd prioritize having a known good and best-in-class LLM available not least for national security reasons; OAI put someone on gov rel about a year ago, beltway insider type, and they have been selling aggressively. Feels like most of the federal procurement is going to want to go to using primes for this stuff, or if OpenAI and Anthropic can sell successfully, fine. Grok winning the Federal bid is an interesting possible outcome though. I think that, slightly de-Elon-ed, the messaging that it's been trained to be more politically neutral (I realize that this is a large step from how it's messaged) might be a real factor in the next few years in the US. Should be interesting! Fudged71 - you want to predict openai value and importance in 2029? We'll still both be on HN I'm sure. I'm going to predict it's a dominant player, and I'll go contra-Gwern, and say that it will still be known as best-in-class product delivered AI, whether or not an Anthropic or other company has best-in-class LLM tech. Basically, I think they'll make it and sustain.

fudged71 2y ago
Somehow I missed the Anduril partnership announcement. I agree with you. National Security relationships in particular creates a moat that’s hard to replicate even with superior technology. It seems possible OpenAI could maintain dominance in government/institutional markets while facing more competition in commercial segments, similar to how defense contractors operate.
- vessenes 2y ago
  Now we just need to find someone who disagrees with us and we can make a long bet. It feels strange to say but I think that the product moat looks harder than the LLM moat for the top 5 teams right now. I'm surprised I think that, but I've assessed so many L and MLM models in the last 18 months, and they keep getting better, albeit more slowly, and they keep getting smaller while they lose less quality, and tooling keeps getting better on them. At the same time, all the product infra around using, integrating, safety, API support, enterprise contracts, data security, threat analysis, all that is expensive and hard for startups in a way that spending $50mm with a cloud AI infra company is not hard. Altman's new head of product is reputed to be excellent as well, so it will be super interesting to see where this all goes.

fudged71 2y ago

Somehow I missed the Anduril partnership announcement. I agree with you. National Security relationships in particular creates a moat that’s hard to replicate even with superior technology. It seems possible OpenAI could maintain dominance in government/institutional markets while facing more competition in commercial segments, similar to how defense contractors operate.

vessenes 2y ago
Now we just need to find someone who disagrees with us and we can make a long bet. It feels strange to say but I think that the product moat looks harder than the LLM moat for the top 5 teams right now. I'm surprised I think that, but I've assessed so many L and MLM models in the last 18 months, and they keep getting better, albeit more slowly, and they keep getting smaller while they lose less quality, and tooling keeps getting better on them. At the same time, all the product infra around using, integrating, safety, API support, enterprise contracts, data security, threat analysis, all that is expensive and hard for startups in a way that spending $50mm with a cloud AI infra company is not hard. Altman's new head of product is reputed to be excellent as well, so it will be super interesting to see where this all goes.