ChatGPTis perhaps the most recognized generative AI name in the industry right now. While the language model it employs is already leaps and bounds ahead of the competition, Google is still playing catchup, with its own large language model (LLM) in the works internally. But if this new report is to be believed,Google’s GPT-4 competitor Geminiis quickly covering the ground and is almost ready for prime time.
Bardis currently the face of Google’s generative AI efforts —the chatbot can already do a lot of stuffwith the right command. It is based on the company’s latest large language model PaLM 2, which itself is a step up from its first-gen version announced in 2022. However, evenGoogle’s newest LLM is behind GPT-4in one key area: PaLM 2 isn’t multimodal, which limits its capabilities beyond text-based input and output. That’s where Google Gemini comes into the picture, which is being trained internally right now. What sets it apart from PaLM 2 is its ability to learn from a variety of sources and generate different kinds of data, including text, images, code, and more.
According to a recent report byThe Information(paywalled), Google Gemini has hit a major milestone in the lead-up to its wider release (viaReuters). Google has started giving a few developers access to Gemini. This early trial run should help them develop further applications and use cases with Gemini’s API running under the hood. These developers will likely bring deep Gemini integrations to their services over time, much likewhat’s been the case with Bard. This gives us hope that Google will have its own version of generative AI that is as sophisticated and capable as GPT-4 sooner rather than later.
The report adds that these third-party developers have access to a rather limited version of Gemini, while Google is still training the full-fledged language model, which is said to be on par with GPT-4. There is little information available right now on the different package sizes of Gemini. Even at this year’s I/O,Google only briefly touched upon Geminiand teased that the multimodal results it was already seeing were impressive. Only time can tell how soon Google will be able to catch up to OpenAI, the Microsoft-backed maker of ChatGPT.
Google offers its generative AI tools to its Cloud customers throughVertex AI, and Gemini is expected to be offered as part of the same package. Other business users can access Duet AI, Google’s generative AI tool built into various Workspace apps, for $30 a month on top of their current plan. However, it’s worth noting that Duet AI relies on Google’s existing AI models, so its capabilities are limited to text-based queries, but there are stillsome innovative ways to use it to the fullest.
