Google hopes to rival OpenAI’s GPT-4 artificial intelligence soon with the launch of it’s own project ‘Gemini’, an upcoming model that seeks to radically transform how Bard and other AI initiatives at the company operate at their core. According to The Information, Google has begun allowing a select group of companies to test a basic chat version of Gemini behind the scenes. With an aim to make its efforts more capable, the model is set to launch very soon.
Gemini stands out due to its modular nature as shown off by Sundar Pichai this past Google I/O. Rather than being a singular AI model, it’s a collection of various models, each tailored toward a different size and type of task. These range from powering chatbots, to summarizing text, advanced code generation and more.
A potential edge for Gemini over its competitors is its data access. As stated by The Information, it could win out over OpenAI’s hugely popular and culture shifting tool because it may have direct access to Google services. We most recently saw this with Google’s Bard update, which allowed you to plug into Gmail, Google Maps, and even YouTube to receive more intelligent and appropriate contextual responses.
It’s rumored that Google also wants to take on Microsoft’s Copilot AI, especially as it pertains to analyzing and interpreting large user-provided datasets like graphs, tables, and more. According to reports, we may also be looking at seamless interaction with Gemini via voice commands. This wouldn’t surprise me one bit, especially as literally yesterday, OpenAI announced that ChatGPT could now see, hear and even speak.
It’s pretty mind blowing, and I’d encourage you to watch the tweet videos below to gain context on exactly what we may be dealing with when Gemini launches. It’s also worth mentioning that this very well may be plugged into Bard (and other Workspace products) in place of the PaLM2 model, which itself will in turn be replacing the heart and soul of Google Assistant before long!
At this time, it’s looking like businesses can expect streamlined access to Gemini via Google Cloud platform, but details are still a mystery. The Information mentions that Google may be deploying tens of thousands of its top-tier TPU AI chips to train Gemini, so the future is going to be nuts.