World Solve

Once deployed, a model could be subtly altered or poisoned without obvious signs, potentially changing its behavior in harmful ways. There's no simple, universal way yet to verify a model is exactly as it was originally trained.

Larger asteroids are easier to spot years ahead, but smaller ones capable of serious regional damage often go undetected until they're very close. Expanding detection capability remains a quiet but serious safety gap.

Detecting gravitational waves was a breakthrough, but modeling exactly what happens during these violent mergers pushes our computing power and physics equations to their limits.

Fast radio bursts flash briefly and then vanish, and we still don't fully understand what causes them. Some theories point to exotic stars, but no explanation is fully confirmed yet.

Right now we can only detect exoplanets indirectly, through shadows and wobbles in starlight. Direct imaging would let us actually study alien landscapes and atmospheres instead of just inferring their existence.

Certain recursively defined sequences appear statistically random for enormous stretches, yet whether hidden structure always eventually emerges remains an open and evolving question. This connects deeply to both number theory and the study of pseudorandomness.

Some sequences appear statistically uniform for extraordinarily long stretches before finally revealing subtle bias, and whether this always eventually happens remains an open question in number theory. This connects deeply to the study of pseudorandom number generation.

Statistical and mathematical tools can detect many kinds of patterns efficiently, but a fully general and guaranteed method covering every possible hidden pattern type remains an unresolved theoretical question. This connects pure mathematics with the practical challenges of modern data analysis.

If advanced artificial systems could one day be conscious, we'd face serious ethical obligations toward them, but we lack any reliable test for detecting genuine inner experience. This uncertainty already complicates debates about AI rights and welfare.

Fraudulent scholarship offers generated to look authentic can convincingly target students seeking financial aid, extracting personal or banking information under false pretenses. Detection and public awareness campaigns remain inconsistently effective.

Sophisticated fake prescriptions generated with AI assistance can be difficult for pharmacies to distinguish from legitimate ones without additional verification steps. Closing this gap remains an urgent unresolved challenge for pharmaceutical safety.

Fraudulent insurance solicitations using AI-generated communication can convincingly mimic legitimate agents, leading victims to pay for coverage that doesn't actually exist. Detecting and shutting down these operations quickly remains an unresolved consumer protection challenge.

Fraudulent warranty claims, sometimes supported by AI-generated fake documentation or images, can cost manufacturers significant money if not caught before payout. Detecting this kind of fraud at scale remains an unresolved operational challenge.

AI-generated voice calls soliciting charitable donations can convincingly target elderly individuals who may be less familiar with distinguishing genuine from fraudulent solicitations. Detection and protective measures for this vulnerable group remain underdeveloped.

Entirely fabricated conference proceedings, complete with fake papers and attendee lists, could potentially be used to pad academic resumes without detection. Verifying the legitimacy of claimed academic activities remains an unresolved challenge.

Entirely AI-generated academic theses can be difficult for reviewers to distinguish from genuine student work without specialized detection tools, and these tools remain imperfect. Protecting the integrity of academic credentials remains an ongoing unresolved challenge.

Aircraft safety relies heavily on scheduled inspections rather than continuous, real-time structural monitoring, since affordable sensors capable of this level of detail remain limited. Expanding this kind of predictive monitoring remains an unresolved aviation challenge.

Physical hardware used in sensitive systems can sometimes be tampered with before installation, and detecting this tampering reliably and affordably remains technically challenging. Closing this hardware security gap remains an unresolved priority.

Fraudulent internship offers, sometimes requesting upfront fees or personal information, can be convincingly generated to target students eager for career opportunities. Detecting and preventing this scam pattern remains an unresolved challenge for students and universities alike.

Illegal wildlife trade pushes species toward extinction while operating largely undetected across international borders. A more coordinated global tracking system could catch this trade before, not after, populations collapse.