Ben Goertzel: Humanity is "quite close" to developing human-level AI
The next step beyond today's generative AI products like OpenAI's ChatGPT and Google's 'Bard' is to create systems that can reason, learn, think, and invent like a human.
"We are quite close to the transition to AGI," says Ben Goertzel, a cognitive scientist and developer of "Sophia the Robot," an anthropomorphized robot AI.
In an interview with ZDNet, Goertzel said that these new AI systems would be able to handle the same intellectual challenges thrown at a human university student. This system could learn recursively, and analyze its own code and hardware, effectively solving the engineering problem of self-improvement. This, in Goertzel's view, could lead to a rapid escalation of capability from the human-level to superintelligence. With the ability to re-engineer itself, introspect, and experiment with countless copies of itself, this AI could achieve a kind of self-enhancement unattainable for the human brain.
For some, this prospect is thrilling, a heralding of a new age in intelligence and innovation. For others, it's a cause for concern, raising ethical and safety questions that are harder to grapple with than the more immediate worries of job obsolescence in fields like journalism or accounting. Nevertheless, this is the goal that many in the AI field are diligently working towards, with the advent of generative AI offering tantalizing hints that we are on the cusp of this transition.
Goertzel estimates we may be a mere 3 to 10 years away from this remarkable juncture. Whether this prediction holds true or not, one thing is clear: the journey toward AGI is laden with complex challenges and profound implications for our future.