According to Bloomberg, Google has invested nearly half a billion dollars into AI startup Anthropic. Neither Google nor Anthropic were willing to comment on the investment, but both companies made separate announcements about Anthropic using Google for its cloud computing services. When announcing that deal, Google stated that it would lend computing power and AI chips to Anthropic for training and deploying AI products, so there is some public evidence of a deeper partnership.
Currently, Anthropic has a language model assistant named Claude, which should be similar to ChatGPT. While it has yet to be released to the public, Anthropic says that expanded access could be coming in the near future.
It will be interesting to see if this deal resembles the $10 billion deal Microsoft recently made with OpenAI. That deal — and previous deals — have given Microsoft exclusive access to the GPT-3 source code that powers ChatGPT, giving it a significant competitive advantage regarding chatbot AI. Google will need to work hard to claw back to level footing with Microsoft.
Google AI: What else is coming
This deal with Anthropic isn’t the only AI chatbot project that Google is involved in. In recent weeks they’ve already announced MusicLM, an AI tool that generates new music from text prompts, and Apprentice Bard, an in-house AI chatbot developed by Google that is directly aimed at supplanting ChatGPT.
Google describes MusicLM as “a model generating high-fidelity music from text descriptions”. In practice, this AI tool works by users providing text prompts with details such as genre, tempo, etc., and the AI spits out a song based on that prompt. The sound quality is 24kHz according to Google, which is just a bit above analog formats like vinyl and cassette tapes but below what you’d expect from Spotify or Apple Music. Still, when I played it through my Creative T15 Wireless speakers it sounded relatively good.
MusicLM can even create vocals for the songs it creates, though the one sample I heard didn’t blow me away lyrically.
Apprentice Bard AI, meanwhile, operates similarly to ChatGPT, where the user holds a conversation with the chatbot, providing questions and receiving responses. The key difference is that Apprentice Bard uses LaMDA (Language Model for Dialogue Application) as opposed to ChatGPT’s proprietary GPT-3.5 technology. Another key difference is that ChatGPT was trained until 2021, whereas Apprentice Bard’s training is still ongoing. This means that Bard may have more up-to-date knowledge than ChatGPT.
So far, Google is internally impressed with Apprentice Bard. It is reported to even suggest that it would offer the chatbot AI a job as a level 3 engineer. We may not have to wait too much longer to see its performance for ourselves, as Google is expected to hold an event in Paris on February 8th, streamed live on YouTube.