Google, IBM & Meta Certificates — All 10,000+ Courses at 40% Off
One annual plan covers every course and certificate on Coursera. 40% off for a limited time.
Get Full Access
Watch a 58-minute MIT seminar where Research Scientist Tim Dettmers explores the capabilities and challenges of open-source agent systems. Discover how AI systems that autonomously plan and execute actions, particularly in coding tasks like SWE-bench, can achieve performance comparable to those using closed-source models like GPT-4. Learn about the critical factors affecting agent system performance, with emphasis on implementation strategies rather than model selection. Examine the current limitations in agent system generalization and evaluation methodologies, highlighting the importance of addressing these challenges for scientific advancement. Gain insights from Dettmers, an accomplished researcher at the Allen Institute for AI and CMU Assistant Professor, known for developing accessible foundation models and creating the widely-used bitsandbytes library for efficient AI systems.
