Booth 21-25 | AI Data Management Zone | Tokyo Big Sight
Most improvements to AI systems happen before deployment: better training data, better fine-tuning, better RLHF. Once the model is out in the world, you generally get the performance you trained
Researchers built a study to answer a question that sounds simple: can people tell whether a scientific abstract was written by an LLM or a human? The answer, for people with machine-learning expertise, was mostly: no. This result is interesting on its own. The implications for how we build annotation
