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
The “more data wins” era is mostly over. What’s replacing it is messier and more interesting: a stack of techniques — fine-tuning, preference optimization, retrieval — each with its own cost curve, its own failure modes, and its own demands on data quality. Pick the wrong lever and you can
