Are We in the Singularity? Sam Altman Says Yes, but Researchers Disagree

The shift is clear: from extinction worrier (2023), to cautious prophet (early 2025), to confident declarer of a "gentle" singularity (mid-2025), to a flat declarer that "we are in the singularity" right now (2026) .

Demis Hassabis: 'Foothills of the Singularity' vs. 'Already Inside'

Google DeepMind CEO Demis Hassabis has used a similar metaphor but with a very different timeline:

• Google I/O 2026 (May): Hassabis told developers they were "standing in the foothills of the singularity" — a more cautious, early-stage metaphor than Altman's claim of already being inside it .
• He told Axios that AGI could arrive by 2029 as a genuine possibility, with 2030 as his central estimate .
• At a Stanford GSB fireside chat (June 2026), he described AGI as potentially "10 times bigger than the Industrial Revolution — and maybe 10 times faster" .
• In 2025, he called AGI "probably the most transformative moment in human history" and gave a 5–10 year window .

Key difference: Altman says the singularity has already arrived; Hassabis says we are at its very beginning (the "foothills") and expects AGI around 2029–2030. Both use dramatic metaphors, but Hassabis's timeline, while aggressive, is still measured in years rather than Altman's "right now."

The Skepticism from AI Researchers: No Evidence of a Singularity

Academic and technical pushback against Altman's claim has been substantial:

• arXiv paper (Feb 2026): A formal reassessment of existential-risk narratives found that "none of the required phenomena (sustained recursive self-improvement, autonomous strategic awareness, or intractable lethal misalignment) have been observed" in current AI systems. It characterized generative models as "powerful, opaque, and imperfect, but devoid of" the hallmarks of a true singularity .
• arXiv modeling study (Feb 2025): Another paper modeling the likelihood of a technological singularity concluded it "appears unlikely to arrive in the foreseeable future," citing a 2017 survey of NeurIPS and ICML authors where nearly half of AI researchers assigned a <10% probability to human-level AGI by 2028 .
• Scaling laws critique (July 2026): A recent analysis found that while scaling laws predict training loss in stable settings, real-world capability "improves in jumps rather than in smooth gains," and benchmark gains correlate weakly with meaningful capability. This undermines the smooth "takeoff" assumption in Altman's framing .
• Guardian op-ed (Feb 2026): Researcher and bot expert Joan Donovan argued the "singularity is not here" and dismissed claims of breakthrough AI agents as overblown, noting that humans have been developing conversational bots for decades without reaching true general intelligence .
• Independent forecasters: Multiple analysts place a true singularity (broadly superhuman AI with recursive self-improvement) in the 2030s–2040s, not the present .

Bottom line from skeptics: Altman's declaration that "we are in the singularity" is at odds with the evidence. No system has demonstrated sustained recursive self-improvement, autonomous strategic awareness, or broad superhuman capability. Many researchers see current AI as powerful but narrow, and view such singularity claims as marketing-driven narrative rather than technical reality.