Memos: OpenAI execs reject Elon Musk’s claims

OpenAI’s prime executives rejected a number of claims Elon Musk has made in a lawsuit — insisting in a Friday memo to employees, seen by Axios, that the corporate stays unbiased, dedicated to benefiting humanity and has but to attain synthetic normal intelligence in its merchandise.

What they’re saying: “[Musk’s] allegations — together with claims that GPT-4 is an AGI, that open-sourcing our know-how is the important thing to the mission and that we’re a de facto subsidiary of Microsoft — don’t mirror the truth of our work or mission,” chief technique officer Jason Kwon mentioned in a memo to staff seen by Axios.

CEO Sam Altman despatched a follow-up message, echoing Kwon’s statements and acknowledging that this 12 months is shaping as much as be a tough one for the corporate.

• “It was by no means going to be a cakewalk,” Altman’s mentioned in his observe, additionally seen by Axios. “The assaults will preserve coming.”

Between the traces: Kwon’s e mail additionally acknowledges the federal government inquiries that adopted final November’s ouster and rehiring of Altman.

• “Individually, I need to replace you on inquiries from authorities companies. After the occasions of final November, they requested us for data and this matches what we would anticipate given the circumstances,” Kwon wrote. “We’re a law-abiding firm, dedicated to assembly all authorized obligations and totally complying with the federal government’s requests. “

Zoom in: Kwon’s memo goals to rebut three of Musk’s principal claims.

1. That GPT-4 represents AGI: “It’s able to fixing small duties in many roles, however the ratio of labor achieved by a human to the work achieved by GPT-4 within the financial system stays staggeringly excessive,” Kwon writes. “Importantly, an AGI might be a extremely autonomous system succesful sufficient to plan novel options to longstanding challenges — GPT-4 cannot do this.”

2. That OpenAI has deserted its mission of benefitting humanity: “Our problem is each to construct AGI and to verify its impression is maximally constructive,” Kwon wrote. “We have discovered that the technique of making frontier know-how and making it broadly out there by APIs and merchandise finest serves each elements of this mission, since we’re capable of entice the required capital, make the know-how broadly usable, and likewise present guardrails as required by society and conscience.”

3. That it’s appearing as a subsidiary of Microsoft. “We resolve what to analysis and construct, learn how to run the corporate, who our merchandise serve and learn how to stay out our mission,” he mentioned. “We additionally straight compete with Microsoft to ship one of the best worth and merchandise to companies, builders and on a regular basis individuals. As we all know, OpenAI is the creator of ChatGPT and ChatGPT for Enterprise, whereas Microsoft provides Copilot and Copilot for Microsoft 365.”