3 ms·

The value of these types of tests has been called into question long before: https://www.youtube.com/watch?v=pHSbvjq1nm4 https://www.youtube.com/watch?v=pHSbvjq

The value of these types of tests has been called into question long before: https://www.youtube.com/watch?v=pHSbvjq1nm4 https://www.youtube.com/watch?v=pHSbvjq1nm4 The above video had a tremendous effect on how I view intelligence and creativity. Seeing pigeons "talking" and solving problems really puts in perspective the power of the context/environment. That is why I think we already have the algorithms for AGI.(i.e. reinforcement learning is probably all you need) We just don't have the processing and/or the right context for it to develop. But I also think the result will have all the downsides of animals/humans. That is, it will be as hard to get the desired behavior from an AI as it is to get it out of an animal/human because the predicting the behavior out of the cost function+environment will remain extremely hard. In fact you can see this mentioned in the video. Getting the right repertoire even for seemingly "simple" behaviors is hard.

joe_the_user 8y ago
Indeed, that's BF Skinner and associates talking about how creative behavior arises naturally from learned behavior. And key here is that things like "the mirror test" are about testing animals for what humans naively believe are their unique traits. And this seems like it could be easily faulty if these naively human beliefs are false, which seems quite likely actually. And also I'd view the focus on consciousness as a naive belief even if psychologists and philosophers have attempted to systematize it. At the same time, Skinner in particular as "behaviorist" seems to go from a lab where he's seeing most behavior mediated by stimulus to a position that seems to imply behavior coming mechanically from stimulus.