Image Credits:<\/strong> Google<\/p>\n<\/div>\n

\u201cWhat we found in our work is that it\u2019s not really the sort of size of model \u2014 that the larger is not always better,\u201d DeepMind VP Zoubin Ghahramani said in a press briefing ahead of today\u2019s announcement. \u201cThat\u2019s why we\u2019ve provided a family of models of different sizes. We think that actually parameter count is not really a useful way of thinking about the capabilities of models and capabilities are really to be judged by people using the models and finding out\u00a0 whether they\u2019re useful in the tests that they try to achieve with these models.\u201d<\/p>\n

Instead, the company decided to focus on its capabilities. Google says the new model is better at common sense reasoning, mathematics and logic. Indeed, as Ghahramani noted, the company trained the model on a large amount of math and science texts, as well as mathematical expressions. It\u2019s no secret that large language models \u2014 with their focus on language \u2014 have struggled with handling math questions without resorting to third-party plugins. Google, however, argues that PaLM 2 can easily solve math puzzles, reason through problems and even provide diagrams.<\/p>\n

PaLM 2 also now features improved support for writing and debugging code. The model was trained on 20 programming languages, including popular ones like JavaScript and Python, but also the likes of Prolog, Verilog and Fortran. PaLM 2 forms the basis of Codey, Google\u2019s specialized model for coding and debugging, which it is also launching today as part of its code completion and generation service, among other things.<\/p>\n

Google today also highlighted that PaLM 2 was trained on a corpus that features over 100 languages, making it, in Google\u2019s words, \u201cexcel at multilingual tasks,\u201d including more nuanced phrasing than previous models.<\/p>\n

Google talks about PaLM as a family of models, which include the likes of Codey but also Med-PaLM 2, the company\u2019s model focused on medical knowledge. There is also Sec-PaLM, a version that focuses on security use cases and a smaller PaLM 2 model that can run on smartphones, which could potentially open up PaLM to more privacy-centric use cases, though Google wouldn\u2019t commit to any timeline for this. Google says this model can process 20 tokens per second, which isn\u2019t extremely fast, but may just be acceptable for some use cases (Google wouldn\u2019t say which phone it tested this on, though).<\/p>\n

It\u2019s no secret that Google has taken a very deliberate approach to launching these AI features \u2014 something the company acknowledges. But at the same time, the standard line from Google\u2019s representatives about this is that the company wants to build these tools responsibly and with safety in mind. That, of course, is also what the company says about PaLM. Without being able to test it before today\u2019s announcement, we obviously don\u2019t know how well it performs and how it handles edge cases.<\/p>\n