It's Not a Rocket, but Google's Gemini Is Preparing for Lift-Off
September 17, 2023
2 min 19 sec read
You're wrong if you think Google's too busy dancing with the Devil in court to be up to anything else. Gemini, the company's latest foray into AI technology, is still on track to debut by the year's end. That's tentative, of course, assuming everything stays smooth sailing in Google's vast ocean.
CEO Sundar Pichai introduced Gemini as a project under development by Google DeepMind and a collaboration between DeepMind and Google's Brain Team. This new AI model is a large language model (LLM) that Google gleefully claims will outperform the current lame lineup of AI systems available to us.
And just to sweeten the deal, it will merge the strengths of DeepMind's AlphaGo, known for mastering the intricate game Go, with language modeling capabilities, meaning AI on steroids.
Gemini is a multi-model, which means it's good at all kinds of stuff, including integrating text, images, and all sorts of other cool data to make its conversational capabilities so much better for us "real people."
Pichai hinted that future features of Gemini could include memory and planning skills, which will allow it to do things that require reasoning. That would already put it ahead of a lot of the human population, but we'll leave that story for another day.
Jeffrey Dean, Google's Chief Scientist, said that Gemini is "next-generation multimodal" and would leverage Google's AI infrastructure, Pathways, to train on diverse datasets.
A Search Engine Journal article titled Google Gemini: What We Know So Far
suggests that Gemini could "compete with AI systems like ChatGPT and OpenAI and possibly outperform them." The folks at Google are a little more confident and seem to think they have a good chance of knocking out the competition.
Demis Hassabis from DeepMind was quick to express his undying adoration of the new AI, talking about Gemini's potential like was going to be the Second Coming of AI. In a June interview with Wired, he bragged about reinforcement learning and tree search features that would give Gemini reasoning and problem-solving skills.
Let's not forget about Gemini's memory, fact-checking, and retrieval skills. It will be the Swiss Army knife of AI, with improved features to enhance accuracy and get rid of junky error-ridden content.
Hassabis stressed that Gemini builds on DeepMind's previous multimodal work, which includes an image captioning system, Flamingo, with promising results.
In his own Wired interview, Sundar Pichai, let us know that Gemini is just a pit stop on Google's AI roadmap. Not satisfied with the current rag-tag collection of chatbots, Google is looking for "incredible universal personal assistants" that will revolutionize every aspect of our lives.
Recent news suggests Gemini will soon be ready for lift-off with a beta release, but most of us will have to wait because Google is only giving early access to a select group of developers, hinting at Gemini's potential integration into services like Google Cloud Vertex AI.
Google needs to check the rearview mirror because Meta (Facebook) is also in the race to build a giant language model to compete with ChatGPT, throwing in a llama-themed model because, well, a llama is kind of cool, right?
And let's not forget all of this AI hype and excitement came right after the first AI Insight Forum, where tech CEOs chatted with the US Senate about AI's future.
The AI revolution is upon us, folks, whether we're ready or not. Pretty soon, we'll have AI doing everything for us, including making our morning coffee and drinking it for us.
Are you ready?
Want to read this in Spanish? Spanish Version >>