It seems like the problem they tried to solve is "people cost more than GPU's".

It's a large an ambitious problem but they pulled it off.

Elsewhere we have “We have no moat, and neither does OpenAI” [1] and it doesn't paint a bright future for OpenAI.

If they didn't target any problem, will OpenAI continue to be worth $30 billion if/when LLMs become as accessible as, say, relational databases?

[1]: https://news.ycombinator.com/item?id=35813322

It sounds like they solved a wide problem instead of a narrow one

I also ignore this rule, in my life and work.

As do most of us.

I will accept my billion-dollar valuation now. Thanks.

I don't know about that, we've been trying to solve natural language as an interface to computing for like 30 years or more.

OpenAI is just the first case where this is so good it's viable and lo and behold it took off like fire.

That is a well defined established problem that anyone paying attention said there was money in.

I was trying to get seq2seq networks to take english and generate API calls all the way back in 2016.

Survivor bias. You can have enough money, time, and talent to tackle a category without a precise milestone plan, but it's not the wisest move if you're struggling to stay alive. "Build a survivable business first" should be the general rule.

It might be my imagination, but it seems like the team over there has some smart people but they happened to hit the ML lottery at the right time, really don't have a clue about business, and now they have to scramble to top what they did before. Brockman and a few of the ML researchers are definitely good at what they do. I don't see the company itself really turning into something profitable when faced with sustained competition.