Hacker News new | past | comments | ask | show | jobs | submit login

>Primarily, because we have no idea what intelligence is, or how it works, why it exists even, etc etc. This goes for human-like intelligence, but also for any kind of intelligence. We just have no good scientific understanding of the subject.

Speaking as a machine learning and neuroscience researcher, how thoroughly have you investigated the matter to claim this? I would definitely not say that we have "no" understanding of the subject. We have a number of predominant paradigms and competing theories regarding the matter.




You shouldn't be downvoted. That portion is clearly over-the-top hyperbole, and it is fair to call that out.

We do have some understanding, even if it is somewhat limited.


How can I answer your question of "how thoroughly I have studied the matter" without sounding like I'm full of myself?

Anyway, that there are many competing theories about intelligence can mean one of three things: either they're all mostly wrong, or some of them are mostly right, or all of them are a little right. Which one do you think is the case, regarding the theories that you have in mind? And what are those?


How would you define what intelligence is, in simple terms to a layman?



What you describe sounds very much like a system that could be implemented with computers and software. Has anyone ever tried to do that?


There are existing software implementations, particularly the various free-energy principle implementations available in SPM from UCL (albeit, in Matlab), and a few others. What has tended to happen is that neuroscientists are more interested in a minimal bio-plausible proof of principle, and machine learners often don't keep entirely up to date with neuroscience after their training. As a result, a lot of the most interesting ideas in neuroscience haven't been scaled up to larger machine-learning applications yet.

Me being a dual trainee in both fields, well, I'm working on it.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: