Demis Hassabis, co-founder and CEO of DeepMind, reveals that his team is leveraging an artificial intelligence program AlphaGo to develop an AI system called Gemini. Hassabis claims that Gemini will surpass the capabilities of OpenAI’s ChatGPT.
Although it’s still in developmental stages, DeepMind’s Gemini is a text-based, large language model akin to GPT-4, the underlying technology behind ChatGPT. Hassabis says the team intends to integrate techniques employed in AlphaGo into Gemini, bringing new capabilities like problem-solving and planning.
According to Wired, the DeepMind CEO says, “At a high level you can think of Gemini as combining some of the strengths of AlphaGo-type systems with the amazing language capabilities of the large models. We also have some new innovations that are going to be pretty interesting.”
Next-Level algorithm
Hassabis explains that it will take several months to complete the development process of Gemini. The endeavour could potentially cost tens or hundreds of millions of dollars. Sam Altman, the CEO of OpenAI, previously stated that the creation of GPT-4 cost the company over $100 million.
In April, DeepMind and Google’s AI lab, Brain, joined forces to create Google DeepMind. This merger aims to enhance Google’s AI research efforts, bringing together powerful expertise under one collaborative team.
“If you look at where we are in AI, I would argue that 80 or 90 per cent of the innovations come from one or the other,” Hassabis says. “There are brilliant things that have been done by both organisations over the last decade.”
Hassabis faces the dual responsibility of driving Google’s AI endeavours forward and addressing the uncertain and potential risks that could follow. The swift and notable advancements in language models have raised concerns among AI experts, including those involved in developing the algorithms. Worries revolve around the potential misuse or challenging control of the technology.
Google’s AI ambitions
Once launched, Gemini could also play a role in Google’s response to the competitive threat posed by ChatGPT, Meta’s LLaMA and other generative AI tools. Despite announcing multiple AI projects during its developer conference and making Bard available to users worldwide, Google is still trailing behind in the AI race.
The DeepMind CEO believes that pausing AI development is impractical, as it would be near impossible to enforce. “If done correctly, it will be the most beneficial technology for humanity ever,” he said.
Hassabis – who was also amongst other big names that signed a statement warning that AI might someday pose a risk comparable to nuclear war or a pandemic – said, “We’ve got to boldly and bravely go after those things. I think more research by the field needs to be done – very urgently – on things like evaluation tests,” he says, to determine how capable and controllable new AI models are.
Hassabis also believes that no one really knows for certain whether AI will end up becoming a danger to society. However, the CEO says that there won’t be enough time to develop safeguards if AI continues to progress at its current pace.
Isa Muhammad is a writer and video game journalist covering many aspects of entertainment media including the film industry. He's steadily writing his way to the sharp end of journalism and enjoys staying informed. If he's not reading, playing video games or catching up on his favourite TV series, then he's probably writing about them.
