3 ms·
it speaks to a general problem of ML systems. Real-world problems are open-ended and a system that cannot reason about what it sees but merely applies object cl
by Barrin92 7y ago
it speaks to a general problem of ML systems. Real-world problems are open-ended and a system that cannot reason about what it sees but merely applies object classification is completely clueless and won't reach a level of fidelity that is needed for safety.
I'm increasingly convinced that virtually every unstructured problem in the physical world is an AI-hard problem and we won't be seeing fully autonomous driving for decades.
- gdulli 7y agoWe as humans possess some skills that are so profoundly important but also so subtle that we don't even recognize them as skills. And excessive optimism about AI is a lack of recognition of how fundamental those skills are to our navigation of the world (both figuratively and literally.)