“Part of the issue is that the existential risk concern has drowned out a lot of the other types of concerns.”
Margaret Mitchell’s observation that “the existential risk concern has drowned out a lot of the other types of concerns” is an argument about attention, not a denial of future danger. Public debate often concentrates on a dramatic possibility: highly capable AI escaping human control and threatening civilization. That scenario deserves research and preparation. But people already experience harms from automated systems through discrimination, surveillance, misinformation, labor exploitation, nonconsensual imagery, environmental costs, and inaccessible public services. When one risk frame dominates, institutions may postpone responsibilities that can be measured and addressed now.
The categories are not mutually exclusive. Techniques developed for current accountability—documentation, red teaming, independent audits, incident reporting, secure development, and clear responsibility—also build capacity for managing more advanced systems. Likewise, governance aimed only at present bias may be inadequate for models that gain new scientific, cyber, or autonomous capabilities. A complete agenda asks how risks connect across time and scale. It avoids forcing every concern into a contest where attention to one makes the others less legitimate. The challenge is to match evidence, safeguards, and authority to different kinds of harm.
Present harms have an important advantage for policy: affected people can describe what happened. Workers can report unrealistic productivity targets, artists can identify unauthorized uses, and communities can show how an automated decision fails in practice. Their testimony provides feedback that laboratory evaluations may miss. Centering those experiences changes the design process. It encourages impact assessments before deployment, monitoring by subgroup, accessible appeal, and remedies when damage occurs. It also reveals institutional problems that cannot be fixed by improving a model’s accuracy alone.
Mitchell’s quote calls for a portfolio of concern. Long-term safety research should continue, but it should not become a reason to ignore harms that are documented, preventable, and unequally distributed. Responsible governance can establish separate thresholds and tools for different risks while preserving a common principle: greater power requires greater evidence and accountability. The public benefits when experts speak across frameworks rather than competing for a single story about AI. A technology this broad will not produce one kind of danger, and society should not rely on one kind of safeguard. Keeping the portfolio visible also improves public trust. People can see that urgent daily harms are not being sacrificed to speculative scenarios, while serious long-range questions are not dismissed as impossible. Responsible institutions should be capable of holding both time horizons at once.
Discrimination, privacy loss, labor harms, and misinformation are documented today, while frontier capabilities may create additional severe risks.
A mature agenda evaluates both rather than making them compete for legitimacy.
Communities and workers see failures that controlled tests and aggregate benchmarks often miss.
Their experience can guide impact assessments, subgroup monitoring, appeals, and remedies.
Different risks require different controls, from consent and anti-discrimination rules to model security and frontier evaluations.
The shared standard is that more consequential systems need stronger evidence, oversight, and accountability.
Explore more "Quotes of The Day"
Continue exploring how writers, thinkers, artists, and advocates have described the choices that shape an everyday life.