Raw LLM Responses

I agree. Not an AI expert but from what I've seen: for a "long" time (\~50 years), we were stuck on CNN and RNN. I think the breakthrough in 2014 was GAN for image generation and 2016 from the AIAYN paper gave us Transformers which was a huge architectural step for natural language processing (LLM). The timing of both these revolutionary findings so close together caused a huge AI wave. But everything after that was just feeding more data. At some point, the brute force approach hits a wall and you stop getting as much gain for an exponential amount of data you feed in. People have been trying new stuff like "agentic" or whatever but they aren't really breakthroughs.