Google launched its next-gen large language model, PaLM 2

Google today launched PaLM 2, its newest large language model (LLM), at its I/O developer conference.

PaLM 2 will be the foundation model for most of the new AI features that Google is announcing today at the conference and will power the updated Bard chat tool, the tech giant's competitor to OpenAI's ChatGPT. Developers can now access PaLM 2 via Google's PaLM API, Firebase, and Colab tools.

Similar to OpenAI, Google did not provide many technical details about how it trained this next-generation model, like the number of parameters.

In a press preparation in front of the present declaration, DeepMind VP Zoubin Ghahramani said that what they found in their work was that it wasn't exactly the size of model and that bigger wasn't generally better.

According to Google, the new model performs better in logic, mathematics, and common sense reasoning. Additionally, PaLM 2 now offers enhanced code writing and debugging support. Twenty programming languages were used to train the model, including well-known languages like JavaScript and Python and less well-known languages like Prolog, Verilog, and Fortran.

Google's specialized coding and debugging model, Codey, which it is also launching today as part of its code completion and generation service, among other things, is based on PaLM 2.