Ask HN: How to build a moat in times of “almighty” foundation models?
Many hard problems became trivial with GPT4 (if you're able to use an API).
What could be the competitive edge of a company building on this technology?
5 コメント
General availability of this tech will even the playing field. You'll still want clever people to actually make all of this work to solve specific problems. So the tech will not necessarily be the advantage if the business isn't working...
When you think about how frustrating some of the simplest personal and business workflows are, there's going to be a much higher bar for what software should do. So there's a lot of opportunity this opens up.
When you think about how frustrating some of the simplest personal and business workflows are, there's going to be a much higher bar for what software should do. So there's a lot of opportunity this opens up.
‘Workflow’ is a good keyword and the OP should look up ‘workflow engine’ and the related technology. That is, big businesses take a task like processing a loan application and break it up into little bits and use business rules engines and related technology to allocate those tasks to people and programs.
That kind of technology can put A.I. systems on wheels in that you have a defined process on guardrails that is resistant to bullshitting, hallucination and such but that can take advantage of the A.I. to do certain tasks.
That kind of technology can put A.I. systems on wheels in that you have a defined process on guardrails that is resistant to bullshitting, hallucination and such but that can take advantage of the A.I. to do certain tasks.
Two-sided markets are still king if you can make one.
Can you elaborate? How do you "make" two-sided markets (particularly when building on foundation models)?
Imagine, for instance, "craiglist that doesn't suck" because it has A.I. assistance.
Also I'd say consider incorporating foundation models into a larger workflow. The thing about the ChatGPT revolution is it has brought out a lot of really lazy people who hope ChatGPT is going to write their pitch dick for them. Not being lazy can put you in the top 90% right away.
Also I'd say consider incorporating foundation models into a larger workflow. The thing about the ChatGPT revolution is it has brought out a lot of really lazy people who hope ChatGPT is going to write their pitch dick for them. Not being lazy can put you in the top 90% right away.