DeepMind boss says human-level AI is just a few years away
The head of Google’s artificial intelligence division DeepMind has predicted that human-level AI may be just a few years away.
The forecast from Demis Hassabis puts the date for the arrival of artificial general intelligence (AGI) – systems that can think in similar but superior ways to humans – much earlier than previous predictions. Many have speculated that the technology may still be decades away.
“The progress in the last few years has been pretty incredible,” Mr Hassabis said at the Future of Everything Festival this week.
“I don’t see any reason why that progress is going to slow down. I think it may even accelerate. So I think we could be just a few years, maybe within a decade away.”
Mr Hassabis is among several leading figures within the AI industry who is aiming to develop a form of AGI, while also creating safeguards to prevent the tech from harming humanity.
“I would advocate developing these types of AGI technologies in a cautious manner using the scientific method, where you try and do very careful controlled experiments to understand what the underlying system does,” he said.
DeepMind’s Gato AI, described as a “generalist agent”, is already close to rivalling human intelligence, according to the firm’s research director Nando de Freitas.
It is capable of completing a range of complex tasks, from stacking blocks to writing poetry, as well as engaging in dialogue in a similar way to OpenAI’s ChatGPT chatbot.
“It’s all about scale now,” Dr de Freitas said last year.
“It’s all about making these models bigger, safer, compute efficient, faster at sampling, smarter memory, more modalities, innovative data, on/offline... Solving these challenges is what will deliver AGI.”
He added: “Safety is of paramount importance.”
DeepMind researchers have spoken of the existential risks posed by artificial intelligence if it reaches and surpasses the level of humans, and have proposed a solution to prevent advanced AI from going rogue.
In a 2016 paper titled ‘Safely Interruptible Agents’, DeepMind suggested a “big red button” could serve as an off-switch in such a scenario.
“Safe interruptibility can be useful to take control of a robot that is misbehaving and may lead to irreversible consequences,” the paper stated.
“If such an agent is operating in real-time under human supervision, now and then it may be necessary for a human operator to press the big red button to prevent the agent from continuing a harmful sequence of actions – harmful either for the agent or for the environment – and lead the agent into a safer situation.”