Reddit co-founder Alexis Ohanian has accused the technology industry of being “tone deaf” in its efforts to communicate artificial intelligence risks to the public. Speaking with CNBC on Wednesday, Ohanian argued that the current discourse is clouded by misinformation and political maneuvering rather than substantive safety analysis.
Ohanian acknowledged that tech companies have done a poor job explaining the technology’s implications. However, he cautioned against fixating on sci-fi dystopias reminiscent of the “Terminator” franchise or the fictional Skynet system. Instead, he urged stakeholders to concentrate on practical, tangible threats such as agent swarms—where multiple AI systems collaborate to execute complex tasks—or the emerging danger of recursive self-improvement (RSI).
The comments follow recent warnings from researchers at major labs OpenAI and Anthropic regarding RSI. This concept involves AI systems assisting in the development of even more advanced models, potentially creating a feedback loop where human developers lose control over the evolutionary trajectory of increasingly capable systems.
Ohanian drew parallels between AI and nuclear energy, noting that both are powerful scientific advancements capable of causing severe harm yet delivering significant benefits if managed responsibly. He expressed hope for a positive trajectory but emphasized the need for thoughtful dialogue over reactive social media posts.
He cited the recent resignation of AI researcher Jacob Coxon as an example of the tension surrounding the issue. Coxon departed both Anthropic and OpenAI, accusing the companies of gambling with human life and noting that some builders believe AI could cause extinction by the end of the decade. Ohanian described such “doomer” rhetoric as discouraging and frustrating, arguing that slogans often lack factual grounding.
“I see all this misinformation flying around, but I do think in the long run, we will get it right,” Ohanian said, calling for a shift away from scoring political points toward addressing the important, real-world aspects of AI governance.
Tech companies need to stop playing politics and start being transparent. The public deserves better than vague slogans.
Recursive self-improvement sounds like sci-fi to some, but the math is real. Dismissing it as ‘doomer’ talk is dangerous.
The nuclear comparison feels apt. We regulated fission; surely we can manage this too with the right framework.
But wait, if we ignore existential risks entirely, don’t we risk being unprepared? Balance is key, not dismissal.
Finally, a voice of reason. The doom-scrolling is exhausting and honestly distracts from fixing actual governance gaps.