Please login or create an account. If you do not have access to this content, you will be shown a 30 second preview and licensing options.
- Presentation
The Future of AI: Agentic Systems, Research Labs, and Physical World Applications
Description
The speaker reflects on the rapid evolution of AI and argues that its future may be shaped by agentic systems: autonomous, goal-driven models that can think, plan, use tools, and coordinate with other agents with little human oversight. They describe how current AI has advanced from simple question-and-answer interactions to tool use, autonomous workflows, and multi-agent “labs” that can generate, critique, and rank ideas for research or clinical decision-making. Examples include AI-assisted literature reviews, clinical synthesis for patient care, coding, and scientific discovery, as well as robots performing increasingly dexterous physical tasks such as picking, placing, and handling lab or surgical-like actions. The talk also looks beyond software into physical-world applications, suggesting AI may increasingly affect real environments and workflows. While emphasizing the excitement and transformative potential of AI, the speaker warns about risks such as errors in autonomous systems, reduced human oversight, and changes to professional roles. They conclude by encouraging the audience to keep experimenting with AI tools, stay curious, and remain vigilant as the field continues to change quickly.
View moreConclusions
- AI is progressing so rapidly that many capabilities once thought far in the future, such as coding, writing, and expert-level task performance, are already becoming routine.
- The next major shift in AI is likely to be toward agentic systems that can act autonomously, use tools, and pursue goals with minimal human oversight.
- Multi-agent systems may outperform single models for complex research tasks because specialized agents can collaborate, reflect, rank, and refine outputs.
- In clinical and scientific workflows, agentic AI could increasingly automate literature review, guideline synthesis, and care-plan generation, especially in informatics-heavy settings.
- AI is beginning to extend from digital tasks into the physical world, including robotics that could eventually perform laboratory and procedural work.
- As AI becomes more autonomous, error handling, oversight, and safety will become central concerns because failures may be harder to trace and correct.
- Widespread AI adoption may reduce the amount of routine work done by humans, changing the roles of researchers, engineers, and clinicians rather than simply replacing them outright.
- The most prudent response to this rapidly changing field is to keep testing new models, stay curious, and remain vigilant about both their capabilities and limitations.
- Toosi et al. PET clinics, 2021.
- Grace et al., AJR, 2025.
- Truhn et al. Nature Reviews Cancer, 2026.
- Wang et al. ICLR, 2025.
- Gottweis et al. arXiv, 2025.