AGI will make its own harness (or whatever else it needs to solve a new problem). As long as you need a human engineer to handcraft a task-specific harness/system for each new problem, AI isn't general. It's an automation tool to be wielded by software engineers. Harness-related research is important and valuable -- as a vector of better automation. But I don't think it gets us closer to general intelligence. General intelligence is when you can adapt on your own.
#### ARC Prize
@arcprize · 2h ago
Today's @symbolica harness is a clear example of what human-crafted targeting can achieve on ARC-AGI-3 public demo set You can "buy" performance with benchmark-specific prompts/strategies
Their approach could still contain useful ideas, excited to see what the community finds
6
9
114
26.1K
24 Replies
21 Retweets
235 Likes
18.8K Views 
One Sentence Summary
François Chollet argues that true AGI requires autonomous adaptation, distinguishing it from task-specific automation harnesses that rely on human engineering.
Summary
In response to a discussion about ARC-AGI benchmarks, François Chollet clarifies his definition of general intelligence. He asserts that AGI must be capable of creating its own 'harness' or problem-solving structure. If a system requires human engineers to handcraft specific setups for each new task, it remains an automation tool rather than a general intelligence system. While acknowledging the value of harness research for automation, he emphasizes that true AGI is defined by the ability to adapt autonomously.
AI Score
86
Influence Score 40
Published At Today
Language
English
Tags
AGI
ARC-AGI
General Intelligence
Automation
François Chollet