Departing OpenAI leader says no company is ready for AGI

Miles Brundage, OpenAI’s elder advisor for the readiness of AGI (aka human-level artificial intelligence), delivered a stark informing arsenic helium announced his departure connected Wednesday: nary 1 is prepared for artificial wide intelligence, including OpenAI itself.

“Neither OpenAI nor immoderate different frontier laboratory is acceptable [for AGI], and the satellite is besides not ready,” wrote Brundage, who spent six years helping to signifier the company’s AI information initiatives. “To beryllium clear, I don’t deliberation this is simply a arguable connection among OpenAI’s leadership, and notably, that’s a antithetic question from whether the institution and the satellite are on way to beryllium acceptable astatine the applicable time.”

His exit marks the latest successful a bid of high-profile departures from OpenAI’s information teams. Jan Leike, a salient researcher, left aft claiming that “safety civilization and processes person taken a backseat to shiny products.” Cofounder Ilya Sutskever besides departed to motorboat his ain AI startup focused connected harmless AGI development.

The dissolution of Brundage’s “AGI Readiness” team, coming conscionable months aft the institution disbanded its “Superalignment” team dedicated to semipermanent AI hazard mitigation, highlights mounting tensions betwixt OpenAI’s archetypal ngo and its commercialized ambitions. The institution reportedly faces pressure to modulation from a nonprofit to a for-profit nationalist payment corp wrong 2 years — oregon hazard returning funds from its caller $6.6 cardinal concern round. This displacement toward commercialization has agelong acrophobic Brundage, who expressed reservations backmost successful 2019 when OpenAI archetypal established its for-profit division.

In explaining his departure, Brundage cited expanding constraints connected his probe and work state astatine the high-profile company. He emphasized the request for autarkic voices successful AI argumentation discussions, escaped from manufacture biases and conflicts of interest. Having advised OpenAI’s enactment connected interior preparedness, helium believes helium tin present marque a greater interaction connected planetary AI governance from extracurricular of the organization.

This departure whitethorn besides bespeak a deeper taste disagreement wrong OpenAI. Many researchers joined to beforehand AI research and present find themselves successful an progressively product-driven environment. Internal assets allocation has go a flashpoint — reports indicate that Leike’s squad was denied computing powerfulness for information probe earlier its eventual dissolution.

Despite these frictions, Brundage noted that OpenAI has offered to enactment his aboriginal enactment with funding, API credits, and aboriginal exemplary access, with nary strings attached.