#productcohort-diana

Hi! If you are building a product with AI, here are a few interesting quotes from a podcast with the CPO of OpenAI re: how to build for the future...

• When you're talking about databases, I bet the database you used this year is probably 5% better than the database you used two years ago, but that's not true at all with AI. It's like every two months computers can do something they've never been able to do before and you need to completely think differently about what you're doing.
• Everything we're used to with computers is about giving a computer very defined inputs. You're confident that if you do the same thing three times, you're going to get the same output three times. LLMs are completely different from that. You probably get spiritually the same answer for the same question, but you don't get the same answer every time.
• We point ourselves in a direction so that we have some rough sense of alignment. I don't for a second believe that what we write down in the roadmap documents is what we're going to actually ship three months from now, let alone six or nine. But that's okay. It's like the Eisenhower quote, "Plans are useless. Planning is helpful,". We're going to throw out the roadmap halfway because we will have learned new things. The moment of planning is helpful even if it's only partially.
• We use ensembles of models much more internally than people might think. If we have 10 different problems, we might solve them using 20 different model calls, some of which are using specialized fine-tuned models, using custom prompts for each one. You want to break the problem down into more specific tasks versus some broader set of high level tasks. And then you can use models very specifically to get very good at each individual thing. And then you have an ensemble that tackles the whole thing.