Mark Zuckerberg‘s vision for “personal superintelligence” is built around putting advanced AI assistants into the hands of billions of people. Ben Goertzel, often referred to as the ‘Father of AGI’, agrees that AI should not be controlled by a single organization. But argues Zuckerberg’s vision falls short of genuinely distributing power.
In an exclusive email interview with Benzinga, the SingularityNET CEO said Meta Platforms, Inc‘s (NASDAQ:META) approach promises broader access to AI while leaving control of the technology concentrated within the company that builds and operates it.
Why Goertzel Says Zuckerberg Is Mixing Up Access and Power
Goertzel agrees with Zuckerberg’s argument that concentrating superintelligence in a single lab is not a future society should embrace.
“He’s right that one lab carefully aligning one system to one value set and letting it benevolently run things on humanity’s behalf, is not a path anyone should trust,” he said.
But according to Goertzel, Zuckerberg’s argument changes in a crucial way.
“But then he does this slide, smoothly, and I suspect not entirely by accident, from ‘distributed power’ to ‘distributed access,’ and those are just not the same animal.”
In Goertzel’s view, Meta’s proposal would give billions of users access to AI assistants, but those systems would still be developed, trained, priced and governed by Meta.
“Billions of people get free agents that run on Meta’s data centers, trained in Meta’s runs, shaped by Meta’s fine-tuning, priced by Meta’s compute auction, and ultimately governed by Meta’s board—that’s the old client-server internet wearing a friendlier UI.”
Why Goertzel Thinks Meta’s Model Isn’t Truly Open
Goertzel argues that genuine decentralization requires users to have meaningful influence over the systems they use—not simply access to them.
“A balance of power among the users of a system is a completely different thing from a balance of power over the system.”
He summed up his criticism of Meta’s approach with a simple analogy.
“It’s home ownership in a company town.”
Instead, Goertzel says beneficial superintelligence should be built on open-source ecosystems with open governance, allowing companies to build products on shared infrastructure rather than own the underlying platform.
“You start with an open-source ecosystem and open governance, and you let corporations grow on top of that substrate,” he said.
The ability for users and developers to leave such ecosystems is what keeps them accountable, he argued. “Exit is always on the table, and it’s exactly the availability of exit that keeps everybody honest.”
By contrast, Goertzel says companies operating closed AI ecosystems face very different incentives. “Try to build the ‘open network’ inside a corporate walled garden instead, and the incentives just invert on you,” he said. “All the engineering effort goes into stopping people from taking their stuff and leaving, rather than into actually serving them.”
Image via Shutterstock
Login to comment