World Solve

Researchers have shown that inaudible or disguised audio signals can sometimes trigger voice assistants to take unintended actions without the user's knowledge. Fully closing this security gap remains unresolved.

Humans can learn a new task from a handful of examples, while most AI systems still need massive amounts of training data to perform reliably. Closing this efficiency gap remains one of AI research's central unsolved challenges.

AI models often can't accurately tell when a question falls outside what they actually know, leading them to guess confidently rather than admit uncertainty. Teaching AI to recognize its own limits remains an unresolved research problem.

AI systems can generate false information with the same fluent, confident tone as accurate information, making it hard for users to tell the difference without independently verifying every claim. Reducing this so-called hallucination problem remains a central AI research challenge.

Most software vulnerabilities trace back to common coding mistakes that better language design could prevent from the start. Despite decades of research, mainstream programming still makes these mistakes easy to make.

Access to anti-ageing interventions could be skewed toward wealthy groups without deliberate policy. Equity-oriented regulation and public programmes will be required.

Human presence and warming conditions may introduce new diseases to previously isolated species and communities. Robust biosecurity protocols, surveillance tools, and response plans are required.

Supply chains for fuel, food, and equipment depend on limited routes and budgets, constraining long-term research. Optimisation tools and shared logistics platforms could improve resilience.

Research, tourism, and fishing generate emissions and local effects that are not fully tracked or integrated into global accounting. Standardised carbon-footprint and mitigation-planning tools are needed.

Research, tourism, and fishing generate emissions and local effects that are not fully tracked or integrated into global accounting. Standardised carbon-footprint and mitigation-planning tools are needed.

Access to field programmes is often restricted to well-funded institutions from certain countries. Virtual research infrastructures and inclusive collaboration models could broaden participation.

Datasets from many nations and disciplines remain siloed, hampering synthesis and reuse. Interoperable data platforms and FAIR-data policies can improve integration.

Sector-by-sector rules may miss combined pressures. Integrated cumulative-impact planning platforms could improve governance.

Researchers keep finding creative ways to bypass safety filters built into AI systems, and defenses often lag behind the newest bypass techniques. This cat-and-mouse dynamic remains unresolved despite constant patching.

Europe produces strong scientific research but struggles to grow globally dominant tech firms due to fragmented markets and risk-averse capital. Cross-border venture networks and translational research platforms can bridge the gap.

Academics are urged to create impact but navigate complex IP and funding landscapes. University-industry matchmaking platforms and impact-oriented grant tools could help.

Existing tests focus on known-answer questions rather than open research tasks. Benchmarks that evaluate models on genuine research subtasks can guide development.

Engineering constraints and ethical questions hinder development. Research into autonomous systems and governance frameworks could move concepts forward.

Most nations invest little in research, limiting innovation. Regional funding mechanisms and collaboration can boost capacity.

Mining plans advance without full understanding. Baseline research is essential.