Tim Scarfe and Adam Becker examine claims that accelerating computation will produce a near-term technological singularityA technological singularity is a hypothetical point when accelerating technological change becomes so rapid or transformative that later social conditions become difficult to predict.. Adam Becker argues that the evidence relies on cherry-picked historical milestonesBenchmark cherry-picking selects favorable tests, time periods or results while omitting evidence that would make an AI performance claim less impressive. and overlooks the physical limits that eventually end every exponential trend.
Tim Scarfe and Adam Becker distinguish the real capabilities of language modelsA language model learns patterns in sequences so it can interpret context and generate likely continuations such as text, code or structured responses. from claims about consciousness, agency and general intelligence. Adam Becker expects systems based on current methods to continue requiring human supervision because confident errorsA hallucination is an AI output that presents false, unsupported, or invented information as though it were reliable. arise from the same process that produces correct outputs.
Tim Scarfe and Adam Becker also question the assumptions behind recursive self-improvementRecursive AI improvement is a feedback process in which AI helps develop better AI systems that then contribute to further improvement., instrumental convergence and longtermist AI risk. They argue that treating intelligence as a single quantity detached from bodies, environments and social context makes these theories appear more certain than the evidence supports.
Tim Scarfe and Adam Becker connect these narratives to concentrated economic power. Adam Becker says technical systems cannot replace political decisions about climate, inequality and public accountability, and calls for stronger regulation of the technology industry rather than accepting elite visions of the future as inevitable.
Watch on YouTube
