3 ms·
Obviously a clickbait headline, few people think current tech will lead to AGI, but while we all wait on a new breakthrough in AI, we can build much more impres
by Eridrus 9y ago
Obviously a clickbait headline, few people think current tech will lead to AGI, but while we all wait on a new breakthrough in AI, we can build much more impressive systems with what we have now than even a few years ago.
I went to a talk from people who had built more scalable training algorithms for Restricted Boltzmann Mahcines (which Hinton was a pioneer in) a few months ago and asked them why they had chosen to use RBMs. Besides the waffly answers about how generative models were more interpretable than discriminative models, the real answer seems to have been that this was a research program that had started when RBMs were still in vogue, and their funding was tied to it, and once it ran out they were going to switch to investigating Generative Adversarial Networks like everyone else.
- dnautics 9y agoI don't think that's correct. There are a lot of people who should know better that think that the current ml technology will lead to agi.
- red75prime 9y agoLet me guess. You think so, because we don't know what the general intelligence is, but the current approaches of ML are surely can't be extended to produce it.
- dnautics 9y agoWe have some simple models of not even general intelligence that ml can't do. To steal from hofstadter:. Let's say I have a function that maps two strings. I'll give you an example function. "abc" -> "abd". What does "efg" map to? And how do you train current generation of ml to solve those problems in one shot like you just did? I don't want to knock machine learning. There's a very huge set of very interesting problems that it does solve that were basically intractable till now and we haven't even scratched the surface of applications for. But the existing techniques have understandable restrictions to the scope of soluble problem. I would also not be surprised if the best models included current ml strategies in some way, for example, in supervisor threads or initial processing of complicated input data into semantic-ish vector valued forms.
- aivijay 9y agoJust ML is never going to lead to an AGI. There are lots of other things that need to be first understood about intelligence, common sense reasoning, cognition and other things which we don't have a clue of (though some think they know it all), to build an AGI.
- mindcrime 9y agoJust ML is never going to lead to an AGI. I think there's a chance it could in principle. Or at least, I think it could yield something close. If human intelligence is largely rooted in a neural network with billions of neurons and trillions of synapses, then it seems reasonable to think that, in principle, a sufficiently deep/wide ANN could show "intelligence". But here's the rub. Doing that would be silly. We already know how to make a neural network with human like intelligence using billions of neurons and trillions of synapses... it just takes a man, a woman, and about 9 months. :-) Also consider the power consumption of an ANN of that scale, given the current state of things. So yeah, to achieve anything truly AGI like, which doesn't need it's own nuclear power plant, I suspect we're going to need other approaches in addition to today's ML/DL stuff.
- dnautics 9y agoi think it's very unlikely. Keep in mind the genesis of the current round of AI algorithms have largely been enabled by 1) the easy access to parallel computing (e.g. GPU, clustering solutions), and 2) large amounts of ingestable data. The truly spectacular AGI will emerge when you can do "one shot" learning. With the exception of zero-shot translation (which is a phenomenal result) we haven't seen very much along those lines.
- dilemma 9y agoAmong those who believe in AGI are Tesla, Google, Baidu, Microsoft, Apple, etc.