Former Salesforce exec says artificial general intelligence should be able to set its own goals, and that's still a long way off

  • AI experts hold varying views on the concept and timeline of artificial general intelligence, or AGI.

  • AI expert Richard Socher said enabling AI to set its own goals would be a significant milestone.

  • But Socher says we're still years away from that.

If you've been following developments in AI, you may have come across the term artificial general intelligence, or AGI.

AGI is a still hypothetical form of AI that exhibits complex, human-like qualities. But it's a nebulous concept, and researchers don't agree on what it entails or when we'll get there.

ChatGPT maker OpenAI has said it's working toward creating AGI and that when it's achieved, it "could help us elevate humanity by increasing abundance, turbocharging the global economy, and aiding in the discovery of new scientific knowledge that changes the limits of possibility."

Researchers at OpenAI's partner, Microsoft, argued in a paper last year that OpenAI's most advanced AI model, GPT-4, "could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system" because it can "solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology."

But other AI experts say we need to apply a broader — and more rigorous — definition of AGI.

"There are two definitions of AGI. There's a simple economic one, which is 80% of the jobs will be automated with AI, and then we can call it AGI," Richard Socher, former Salesforce chief scientist and founder of the AI-powered search engine You.com, told Business Insider by email. If you restrict that notion to "digital jobs," he said we could probably get there in the coming three to five years.

But if you start thinking about AGI as a form of intelligence that can "learn like humans" and "visually have the same motor intelligence, and visual intelligence, language intelligence, and logical intelligence as some of the most logical people, then I think we're much further out." Socher said it could take as little as 10 years but as much as 200 years to get there — so there's no clear timeline.

In the coming 12 to 24 months, though, one milestone that Socher thinks we could hit is enabling AI with the power to act — like making a purchase.

"For example, you ask, 'What's the price of this yoga mat on Amazon?' You get the price, and then you ask, 'Can you buy it?'" In this case the AI system would not be responding to the query by text, but with a sequence of clicks, Socher explained.

But if we really want to move the needle toward AGI, Socher said humans might need to let go of the reins a little and build AI that can set its own goals.

"I think it's an important part of intelligence to not just robotically, mechanically, do the same thing over and over that you're told to do. I think we would not call an entity very intelligent if all it can do is exactly what is programmed as its goal," he told BI. Until then, "it's still a tool."

Read the original article on Business Insider