Blog
-
Autonomous GUI Agent – Theory, Promises and Limitations.
Progress of GUI agents. Drawbacks of voice and language interface. GUI as self-emergent order of cooperative game. What applications are useful and reliable enough, what are not? Connection to other inherently complex problems. Read more
-
Misleading Promises of Long Context LLM.
1. LLM still cannot reason in the long context. 2. The role of false positive rate. 3. Why does LLM performance degrade with context length? Really? 4. How to deal with the problem? 5. Why is the problem easily overlooked? Kuhnian structure of AI. Read more