At Google I/O 2023, Google unveiled its latest large language model, PaLM 2, which already powers 25 Google services including Bard, Docs, Sheets, and Gmail. PaLM 2 offers improved multilingual, reasoning, and coding capabilities, as well as versatility, with four sizes available for various use cases: Gecko, Otter, Bison, and Unicorn.
PaLM 2 is trained on more than 100 languages, improving its understanding and generation of nuanced text. “This has significantly improved its ability to understand, generate and translate nuanced text — including idioms, poems and riddles — across a wide variety of languages, a hard problem to solve”, the company shared in a blog post. The model also passes “advanced language proficiency exams at the ‘mastery' level”.
Trained on Scientific Papers
In addition to its multilingual capabilities, PaLM 2 has been trained on scientific papers, web pages, and mathematical expressions, enabling it to demonstrate “improved capabilities in logic, common sense reasoning, and mathematics”, says Google. The model has also been pre-trained on extensive codebases, making it proficient in popular programming languages like Python and JavaScript, as well as specialized languages like Prolog, Fortran, and Verilog.
Among the specialized Google services powered by PaLM 2 is Med-PaLM 2, which “achieves state-of-the-art results in medical competency, and was the first large language model to perform at ‘expert' level on U.S. Medical Licensing Exam-style questions”. Another specialized service is Sec-PaLM, which uses PaLM 2 and Google Cloud to scan and analyze potentially malicious scripts and cyber threats.
Upcoming Next-Gen Gemini Model to Be Mutlimodal
Google is also developing Gemini, a next-generation multimodal model designed for efficient tool and API integrations. Once fine-tuned and tested for safety, Gemini will be available “at various sizes and capabilities, just like PaLM 2, to ensure it can be deployed across different products, applications, and devices for everyone's benefit.”
