All you need to know about PaLM 2, Google’s new AI model

agencies

Friday, May 12, 2023 08:00 PM

Google announced yesterday at its annual developer conference.I/O 2023) about model (PaLM2) which is the latest large language model (LLM) developed it, which is the model in which it works chat bot (cool) coldIt will also be the cornerstone of all artificial intelligence products offered by Google, thanks to its wide capabilities and capabilities aitnews.

Prepare (PaLM2(The first and direct competitor of the model)GPT-4) provided by the company (OpenAI) during the month of March, which is currently running the paid version of ChatGPT.

He said (Zubin Ghahramani) Zoubin Ghahramani Vice President in the company (DeepMind) affiliated with Google in the post announcing the model via the Google blog: “Over the past decade we have learned a lot from the development of artificial intelligence such as how to intelligently scale neural networks, as we learned during our research that the size of the model is not the main component of language models, because the largest size is not always Better yet, research creativity is key to building great models. Recent advances have taught us how to design and train models, how to embed multimedia, the importance of having human feedback, and how to build models more efficiently than ever before.”

Here’s everything you need to know about PaLM2 Google’s new artificial intelligence model:

What is the PaLM 2 form?

?

Prepare (PaLM2) – which is the new generation of the large language model (PaLM) introduced by Google for the first time in 2022 – a language model that is faster and more efficient than the version thanks to its strong capabilities in logical thinking and mathematics, writing and correcting code, and its handling of multilingual texts.

Google says that the model (PaLM2) has been extensively trained on multilingual text, which includes more than 100 languages, and this has greatly improved its ability to understand, generate, and translate accurate text with greater accuracy than previous models.

Google also focused on making its language model better in logical thinking and mathematics, as (Qahramani) indicated that the company had trained the model on a large amount of mathematics and science texts, in addition to scientific papers andweb pages containing mathematical expressions.

It is no secret that large language models, with their focus in language, have had difficulty dealing with math questions without resorting to external tools. Google says its model (PaLM2) can easily solve mathematical puzzles, solve problems in logical ways and even render graphs without the need for any external tools.

in addition to; Characterized PaLM2 Now with improved support for code generation and debugging, the model has been trained in 20 programming languages, including popular languages ​​such as: JavaScript And Pythonand languages ​​that are not widely known such as: such as: Prolog And Fortran And Verilog.

So it forms a model PaLM2 Basis Codey Google’s template for code generation and debugging, which Google also announced yesterday as part of its code completion and generation service.

prepares a model (PaLM2) is more capable than previous models, as well as faster and more efficient, and Google offers it in a variety of sizes, making it easy to rely on for a wide range of use cases.

Google says it will provide a template (PaLM2) in four different sizes arranged from smallest to largest as follows: GeckoAnd OtterAnd BisonAnd Unicorn.

model features (Gecko) It is so lightweight that it can work on mobile devices, and it is fast enough to work with interactive applications used on mobile devices, even when it is not connected to the Internet. This diversity means that the model (PaLM2) to support entire categories of products in more ways to help more people.