AI · just now
Why LLMs alone can't predict user intent
Yobi CTO Frank Portman explains that next-token prediction, while effective for language, is not suited for forecasting human behavior. Yobi builds a 'foundation model of behavior' using transformers and graph neural networks to power millions of personalization decisions per second. The approach prioritizes consumer data privacy while delivering real-time intent predictions.
Meridian48 take
The piece usefully challenges the assumption that LLMs are a one-size-fits-all solution, but Yobi's claims about privacy and scale need independent verification.
intent-predictiongraph-neural-networks