The Gemini AI model-slash-chatbot isn’t just Google’s effort to catch up with OpenAI — it's an experiment that could totally change the company’s future.
What happened: Google released Gemini Ultra, the most advanced version of its generative AI model. It’s also adopting the Gemini name across its AI products, like Workspace’s Duet AI and the Bard chatbot, which is now available in Canada.
- Canada was one of the few countries that couldn’t access Bard. Google confirmed that was due to its now-resolved spat with the government over the Online News Act.
- The standard Gemini Pro model will continue to be what most people will use in the chatbot, be it in a browser or new mobile app (the latter is U.S.-only for now).
- Ultra is reserved for those willing to pay $27 a month, though that comes with the features of a Google One Premium subscription, like extra cloud storage. In the future, it will also include Gemini integrations in Gmail and Google Docs.
Why it matters: Search isn’t going away, but Gemini is a potential step away from having it be Google’s key product. And it matches how more people are finding information online without a search engine, be it TikTok videos, Reddit posts, or their favourite email newsletter.
- One way Gemini differs from OpenAI’s GPT-4 is that it’s also trained on images, audio, and code — not just text.
- Google is also positioned to put AI into a lot of everyday services, be it email, Docs, advertising, or smart homes.
Zoom out: One analysis puts Gemini Ultra’s performance on par with GPT-4, but they both have strengths. GPT-4 is better with code and writing. Gemini is better at explanations, integrating images and search results, and seems less prone to technical errors.
Bottom line: Both chatbots, however, still have issues with consistency and hallucinations.