Our great sponsors
-
ParlAI
Discontinued A framework for training and evaluating AI models on a variety of openly available dialogue datasets.
-
InfluxDB
Power Real-Time Data Analytics at Scale. Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality.
Some notes from my perspective of PLATO-XL: * The model comparisons sometimes seem deliberately disingenuous. * I love that we’re measuring a ‘hallucination’ score now(!) and this model has a lower/better score when compared against other (older, of course) chatbots. * Why not compare with the latest BlenderBot 2.0, open sourced by FAIR, and available with metrics? https://github.com/facebookresearch/ParlAI/tree/main/projects/blenderbot2 * Why not also compare with GPT-3 Curie 6.7B? It would seem to make sense as this is a very popular model, though not explicitly chatbot... * They have failed to mention the largest SOTA chatbot model, Google’s LaMDA (Language Model for Dialogue Applications). Although not publicly released (and maybe even the paper was too late), the recent paper from Google shows that LaMDA has over 100 billion parameters, or ~10 times the size of PLATO-XL…