At the end of 2022, I published a post with 7 predictions for 2023. Today I’m doing the same for 2024. Here’s a quick preview to tease you:
OpenAI will release a small but very powerful LM (1-10B)
Ilya Sutskever will leave OpenAI for a competitor
Benchmark evaluations won’t be the gold standard
Open-source AI will catch up with GPT-4
Inference for LMs will become very cheap
ChatGPT will remain an unchallenged product
A non-transformer LM will match the top small transformer LMs
The structure is simple: I make a prediction and then make a few arguments as to why I believe that’s the case. Before we start, I’m curious: Let me know in the comments what’s your own top AI prediction for 2024.
I’m also doing a (hopefully) honest evaluation of how well I did with my 2023 prediction article and I have to say… pretty good to be my first time!