Sam and Greg's response to OpenAI Safety researcher claims
OpenAI’s handling of AI safety, leadership turmoil, and its close ties to Microsoft has triggered skepticism about its ethics, transparency, and long‑term intentions. Commenters weigh whether fears of AGI-driven human extinction are overblown hype or a real policy priority, while highlighting nearer-term risks such as job displacement, scams, deepfakes, military use, and regulatory capture that could entrench a few dominant players. Many see a gap between the company’s rhetoric about “safety” and its aggressive commercialization, and debate whether open-source or regulation offers a better counterweight.
Perceptions of OpenAI and its Leadership
- Many say OpenAI’s reputation has eroded: from admired lab to “cringe,” cult-like culture, and reality‑TV‑style drama around leadership changes.
- Others argue most non‑tech people still see it as the “name brand” AI provider and don’t follow the drama.
- Some see the safety‑team departures as normal high‑growth‑startup churn; others see them as a major red flag that safety voices are sidelined.
- The use of a cofounder’s account and first‑name branding (“Sam and Greg”) is seen by several as manipulative image management.
Microsoft, Google, and Control
- Debate over whether Microsoft is the de facto controller of OpenAI (via funding, compute, board seat) or merely a profit‑sharing partner.
- Some credit Google with much of the underlying transformer/LLM research, with OpenAI winning on productization and RLHF.
AI Safety, AGI Risk, and “Superalignment”
- Strong split:
- One side: AGI may be near (late 2020s), extinction risk is real, and OpenAI under‑invests in alignment compared to capabilities.
- Other side: current models are glorified autocomplete, far from even “cat‑level” intelligence; doomsday talk is hype and regulatory theater.
- Some see “AI safety” as devolving into PR/”cover your ass” or regulatory capture; others distinguish serious existential‑risk work from DEI‑style ethics and corporate sanitization.
- Several liken risk‑mitigation roles to corporate compliance: either ignored, co‑opted, or turned into theater.
Current Harms vs Speculative Catastrophe
- Concrete/near‑term concerns raised:
- Scams, deepfakes, fake news, disinformation, and AI girlfriends exploiting loneliness.
- Students cheating; erosion of trust in online content and in whether a human is speaking.
- Job loss in areas like customer support; concentration of power and wealth; more effective surveillance, propaganda, and AI‑enabled warfare.
- Others counter that LLMs are still mediocre, error‑prone, and less dangerous than many existing technologies.
Openness, Regulation, and Contracts
- Criticism that “Open” AI has become closed and profit‑driven, undermining its original charter.
- Disquiet over reported exit agreements with lifetime non‑disparagement and equity clawback threats; some note leadership has since promised to roll these back.
- Views diverge on open‑source vs closed models: open models as defense against “AI aristocracy” vs fear of uncontrolled misuse.