Google made the next big step forward in artificial intelligence (AI) on Wednesday when it released project Gemini, an AI model that has been trained to act like a person. This is likely to fuel more debate about the pros and cons of AI.
In the beginning, less advanced forms of Gemini called “Nano” and “Pro” will be added right away to Google’s AI-powered chatbot Bard and its Pixel 8 Pro smartphone.
Google says that Bard will become smarter and better at planning things after Gemini helps him. Google says that Gemini will be able to quickly summarize recordings made on the Pixel 8 Pro and send automatic answers on messaging apps, starting with WhatsApp.
The biggest steps forward for Gemini won’t happen until the beginning of next year, when its Ultra model is used to release “Bard Advanced,” a more powerful version of the robot that will be first only available to test users.
At first, the AI will only work in English around the world. However, Google execs told reporters at a briefing that the technology will have no trouble adding support for other languages in the future.
Based on a demo of Gemini for a group of reporters, Google’s “Bard Advanced” might be able to do AI multitasking like never before by recognizing and understanding presentations with text, pictures, and video all at the same time.
Gemini will also finally be added to Google’s main search engine, but the exact date of that change hasn’t been announced yet.
Demis Hassabis, CEO of Google DeepMind, the AI company behind Gemini, said, “This is a big step forward in the development of AI and the beginning of a new era for us at Google.” Google beat out other bidders, including Meta, the parent company of Facebook, to buy London-based DeepMind almost ten years ago. Since then, it has combined it with its “Brain” business to focus on Gemini’s growth.
Google says that the technology is very good at solving problems in math and physics. This gives AI supporters hope that it could lead to science breakthroughs that make life better for people.
On the other hand, some people who are against AI worry that the technology will one day be smarter than humans. This could mean the loss of millions of jobs and even worse actions, like spreading false information or starting the deployment of nuclear weapons.
In a blog post, Google CEO Sundar Pichai said, “We’re going into this work with courage and responsibility.” “That means setting high goals for our research and going after the skills that will help people and society the most, while also building in safety features and working together with governments and experts to deal with risks as AI gets smarter.”
The arrival of Gemini is set to step up the competition in AI that has been heating up over the past year between Microsoft and OpenAI, a startup in San Francisco.
OpenAI was already working on its most advanced AI model, GPT-4, with Microsoft’s money and computing power when it released the free ChatGPT tool at the end of last year. That robot that was powered by AI became famous all over the world, which made people talk about the business potential of generative AI and put pressure on Google to fire Bard.
Bard came out around the same time that OpenAI released GPT-4 in March. Since then, the company has been adding new features for both consumers and businesses. In November, they revealed a feature that lets the chatbot analyze images. It has been fighting for business with Anthropic and Microsoft, which has invested billions of dollars in OpenAI and has exclusive rights to its technology.
So far, the partnership has been good for Microsoft. The company’s market value has grown by more than 50% so far this year, mostly because buyers think AI will be very valuable for the tech industry. Alphabet, the company that owns Google, has been riding the same wave. So far this year, its market value has grown by more than $500 billion, or about 45%. Even though Gemini has been getting a lot of attention lately, Alphabet’s stock went down a little on Wednesday.
Concerns have been made that OpenAI has moved away from its original goal to protect humanity as technology advances. This is because Microsoft has become more involved with OpenAI over the past year and OpenAI has become more aggressive in trying to sell its products.
These worries got worse last month when OpenAI’s board fired CEO Sam Altman without warning over trust problems that were not made public. After criticism that could have destroyed the company and sent a lot of talented AI engineers to work for Microsoft, OpenAI put Altman back in charge as CEO and changed the board’s makeup.
Now that Gemini is out, OpenAI may have to show that its system is still smarter than Google’s.
Eli Collins, vice president of product at Google DeepMind, said of Gemini, “I am in awe of what it can do.”
Google wouldn’t share Gemini’s parameter count in a virtual news conference. This is one way to measure how complicated a model is, but it’s not the only one. A white paper released on Wednesday said that the most powerful version of Gemini did better than GPT-4 on multiple-choice tests, grade-school math, and other tests. However, it also said that getting AI models to reach higher levels of thinking is still hard.
Some computer scientists think that big language models can only do so much. These models work by guessing the next word in a sentence over and over again, and they often make mistakes called hallucinations.
“With Gemini, we made a lot of progress in something called factuality. In that way, Gemini is our best model. Collins said, “But I would still call it a research problem that hasn’t been solved.”