An interview with Jürgen Schmidhuber, whose neural network work in the 1990s went on to be used in Google Translate and Siri, on why many AI fears are misplaced
Jürgen Schmidhuber believes AI will progress to the point where it surpasses human intelligence and will pay no attention to people
Context & Ripple Effects
Jürgen Schmidhuber has spent decades on this argument from both sides: the 1990s neural network research that ended up inside Google's neural-network rebuild of Translate and Siri made him one of the field's founding figures, while his Swiss AGI startup Nnaisense — covered in an earlier profile of him as a controversial pioneer — puts him among those actively trying to build superintelligence rather than merely predict it.
His claim that surpassing-human AI would simply pay people no attention lands mid-2023, at the peak of the existential-risk debate, where Meta's Yann LeCun has already staked out the same ground, calling experts' humanity-threatening scenarios "preposterously ridiculous" and warning that locking research away would be a mistake.
First-order effects
- Schmidhuber becomes a second founding-generation voice — alongside LeCun's string of interviews dismissing danger hype — legitimizing the position that superintelligence is plausible but existential alarm is not.
- Because his own work shipped in Google Translate and Siri, his skepticism carries product-level credibility that pure theorists lack, giving journalists and policymakers a counterweight to safety-focused testimony.
Second-order effects
- Labs and researchers arguing against precautionary regulation gain an citable authority: with Schmidhuber and LeCun aligned, the 'keep research open' camp now spans two of the most decorated careers in deep learning.
- The disagreement sharpens into a public fault line within the AI elite itself, forcing every lab leader and government advisor to pick a side between open-acceleration and containment framings.
Third-order effects
- If the pattern holds, the decisive fight over AI policy will be argued between eminent insiders rather than imposed from outside, with governments calibrating regulation around which camp's framing wins the credibility contest — an outcome the corpus leaves genuinely unresolved.
- Schmidhuber's later move into a national AI leadership role points toward sovereign states recruiting exactly these pioneer voices, embedding the open-research worldview inside state-funded AI programs.
The trend: The most senior figures who actually built modern deep learning are consolidating into a public vanguard against AI doom narratives, even as they assert superintelligence itself is coming.