Bard bombshell — Google AI plans just leaked ahead of I/O 2023
With just one day to go, it looks like Google’s entire I/O 2023 game plan just leaked — at least for Google Bard.
CNBC has reportedly seen internal documents from Google that outline the search giant’s AI plans for this year’s I/O conference. And it looks like AI will definitely be a key feature of I/O, with us reporting just yesterday on how we expect to see Project Magi — Google’s next-generation AI search engine — at least partially unveiled at I/O.
But Magi is noticeably absent in CNBC’s reporting. Instead, it focuses on PaLM 2, a new large language model (LLM), and upgrades to Google Bard, the AI experiment that Google launched earlier this year to compete with ChatGPT.
PaLM 2: Google’s new AI workhorse
The big headliner for Google this year appears to be PaLM 2, the next iteration of Google’s Pathways Language Model that currently powers Bard in addition to many of the search giant's other AI tools. Reportedly, the new model will support over 100 languages and has been already operating under the radar as “Unified Language Model.”
While PaLM 2 seems focused on coding, math and creative writing, it seems that it can also be fine-tuned into something more specialized. Google has already announced Med-PaLM 2, an AI tool focused on medical diagnosis, and it appears that it will run on PaLM 2 given the suffix chosen.
Big Bard, Giant Bard and Multi-Bard
Google Bard has already seen some “Experiment updates” since its launch earlier this year, but it looks like some big updates are being reserved for I/O tomorrow.
According to the CNBC report, Google has been secretly testing three versions of the AI chatbot: “Big Bard,” “Giant Bard,” and “Multi-Bard.” While much isn’t revealed about Big Bard and Giant Bard, we do have an idea of what Multi-Bard could bring — multimodal functionality.
This is multimodal functionality, and reportedly, Multi-Bard could bring it to Google Bard as soon as tomorrow.
Multimodal functionality is when an AI chatbot can handle inputs or outputs in more than just text. For example, ChatGPT cannot understand images — it can only handle text. But GPT-4 can understand what is in an image and synthesize it into a text response. This is multimodal functionality, and reportedly, Multi-Bard could bring it to Google Bard as soon as tomorrow.
Aside from this bombshell, the other Bard updates we can expect to see are increased language support, including Japanese and Korean and expanded availability.
What else to expect from AI at I/O 2023
Aside from PaLM 2 and a beefier Google Bard, Google is also expected to expand its generative AI capabilities in Workspace apps, including “template generation in Sheets and image generation in its Slides and Meet products.” This comes on the heels of similar generative AI features being introduced to Gmail and Google Docs earlier this year.
But again, this is all based on leaked documents, so some or all of it may not come to pass. So make sure to stay tuned to our Google I/O 2023 coverage tomorrow so you don’t miss any of the official announcements from Google, from Bard upgrades to a new Pixel 7a phone.