| So I'd love to get your thoughts on the following. There's a tweet from Sam Altman that he wrote a few years ago and it's aged quite well as they say. He was announcing the release of ChatGPT and maybe an early iteration of ChatGPT, maybe 3 or 3.5 or something like that. Maybe even an earlier one. And the tweet went something like, it's our conversational, or first it's a conversational model or something. <br><br>And what's interesting to me about this is that I discovered AI through ChatGPT or got excited about it through that interface. And then from there worked back to more instructional workloads as then I used it as a chat interface, then began using LLMs through their API endpoints and then began using them programmatically and scripting and using them on my local computer. And now I doing much more of that than I am using them as chatbots. <br><br>I know a lot of people, I think even people who are pretty technically literate, aren't really aware that there's, that there's, AI can be used in this way. But what's interesting about that tweet I mentioned is it inferred that instructional models actually predate conversational models. In other words, that I think what he was saying was that OpenAI had developed GPT firstly for instruction following, and then they sort of refined it for conversation. <br><br>And what I'm curious to know is, is that accurate that instructional models predate conversational models and if so by sort of how long? |