Google’s Pathway Language Model, launched in 2022, has been one of the most advanced language models in the world and was the foundation upon which Google’s AI bot, Bard, was built. However, with the AI battle getting heated with each GPT improvement, Google has now updated its PaLM model with the PaLM 2.
What is the PaLM 2?
The PaLM 2 is Google’s Language model released on May 10, 2023, as an update to the PaLM. According to Google Vice President Zoubin Ghahramani, PaLM2 is a ‘state-of-the-art’ language model which has improved multilingual, reasoning, and coding capabilities.
On multilingualism, Google states that PaLM 2 has been heavily trained on multilingual text spanning over 100 languages. This means that PaLM 2’s ability to understand, generate and translate nuanced texts across different languages has been improved, extending to idioms, riddles, and poems.
On reasoning, it has been fed a wide dataset that includes scientific papers and web pages with mathematical expressions and, thus, has significantly improved capabilities of logic, reason, common sense, and mathematics.
On coding, Google says that PaLM 2 has been trained through publicly available source code datasets. Thus, it excels in popular programming languages like Python and JavaScript but can generate code in languages like Prolog or Fortran.
What this means
Google’s PaLM 2 is set to be a direct rival of GPT-4, and according to some tests, it is already beating GPT-4. For instance, it is available in smaller sizes that are specific to different applications that do not have as much onboard processing power. Google also said that its reasoning capabilities are competitive with the GPT-4, and tests have shown this to be the case.
However, there is still no real-world application for the model as it remains under development. So, subscribe to your newsletter to be updated on PaLM2 and other tech-related news.
