03-27-2023, 05:17 PM
(03-25-2023, 09:43 AM)KillerGoose Wrote: The GPT model is revolutionary. I am not sure if you meant "prompt" or "program" but I have used it to help me craft machine learning models. I am a machine learning hobbyist, so I have had proofread my code/write code for me/explain the code etc.
ChatGPT has restraints on it, but the actual GPT-4 model is very powerful. It is able to work and correctly solve high level math problems and produce mathematical proofs. It has consistently outperformed users on Leetcode in solving coding problems. It can create games from scratch. It has the ability to "memorize" the layout of a structure that you describe and draw it back for you. It was able to pass the bar exam and is able to pass the Sally-Anne false belief test. GPT-4 is not AGI but these are all things that an AGI will be capable of doing, so we are getting some flashes of what could look like.
Bard, which is Google's LLM, is also incredible. It has some interesting responses.
I just got off the waitlist and tried Bard last night. I like it and it is up to-date, unlike ChatGPT and Caktus AI, but ChatGPT is at this point still better.