Qiu and Warner made a compelling argument that AI would essentially replace software development as we know it, though it might take a little while. Qui gave an example of a simple email sorting program that her tools can’t create on their own quite yet, but will soon. Warner made a bravely specific prediction that the industry would shake out into 5-7 companies making general-purpose LLMs, while everyone else focussed on niches.
Qiu was firm in the view that general-purpose LLMs were not the right choice for everything, and that “we actually need quite a few specialized models that are able to do step-by-step reasoning.” Warner was almost messianic about the potential benefits of AI for humanity and dismissive of short-term obstacles and of VCs on Twitter: “We got to overthinking a lot of stuff and who the fuck cares what they say anyway?”
Qiu was sober about the challenge of making things that businesses actually need: the integrations and UI innovations and safety and security issues are a very long way from being solved. But she was similarly unconcerned about emerging monopolists: no, OpenAI won’t own everything, because enterprises have their own special needs.
Add comment