Google first introduces LaMDA
Google introduces LaMDA for open-domain dialogue at I/O 2021; the model remains an internal research system without public weights or a general-purpose API.
What happened
Event details
At Google I/O 2021, Google introduced LaMDA, a Transformer-based model family trained specifically for open-domain dialogue. The announcement emphasized sensible and specific responses while identifying factuality, safety and bias as unresolved research issues. This was a research disclosure, not the launch of a public model service. A January 2022 paper later described models with up to 137B parameters, pretrained on about 1.56 trillion words, and experiments using retrieval, translation and calculation tools to improve grounding.
Assessment
Why it matters
LaMDA applied large language models to open-domain dialogue and placed response quality, safety and grounding in one research program; it also became an important predecessor to Google's later conversational products.
Availability
Access notes
At the time, Google disclosed the research but did not release model weights or a general-purpose public API.