Earlier quoted context omitted.
Pick a camp: symbolic or sub-symbolic. If latter then do a neuroscience degree along with a computer science degree. Get an insight while experimenting with biological brain. Make a start-up that capitalizes on your new idea.
Getting 2 degrees sounds like way too much work. How about reading a plenty of books and research papers? What do you mean by symbolic or sub-symbolic?
Google to Buy Artificial Intelligence Startup DeepMind for $400M
201–209 of 209 posts
Re: Google to Buy Artificial Intelligence Startup DeepMind for $400M
#202Earlier quoted context omitted.
I actually read this paper a couple weeks ago as part of a deep learning reading group that I co-run. While several of these authors are household names in the RL community, this paper was not actually that impressive to me. The only real "deep" learning here is that they used a GPU library and stochastic gradient descent to perform Q-learning updates on a network with 3 large hidden layers. It was an interesting app…
This deep learning group you mentioned. May anyone join in?
Sorry, it's for UT Austin PhD students only.
Re: Google to Buy Artificial Intelligence Startup DeepMind for $400M
#203Earlier quoted context omitted.
> You can't even say if one human surpassed other human in intelligence, how are you gonna measure machine? It's not because something is difficult to measure that it does not exist(1). Even if you don't believe in IQ tests, you gotta admit that those people Google just hired are more intelligent than the average Joe. They sure are more intelligent than me anyway. Probably more intelligent than you as well. In the sa…
> It's not because something is difficult to measure that it does not exist(1). No, I don't say that it doesn't exists. What I say is that intelligence is not something to be expressed in one number, it doesn't work that way. Intelligence is more like NxN matrix of numbers, where each number in a matrix represent individual skill in some specific task. As for your example of Google employee and average Joe, what you…
> What I say is that intelligence is not something to be expressed in one number, it doesn't work that way.
But nobody in the singularity field says that intelligence is one-dimensional quality! It's a strawman.
It reminds me of a common accusation that "computer people" have subpar worldview because they "reason in 0s and 1s, and the world is not binary", to which I say that actually "computer people"'s view is superior because they figured that out long ago and developed proper methods to quantify and deal with uncertainty.
> Intelligence is more like NxN matrix of numbers, where each number in a matrix represent individual skill in some specific task.
This is also not a good model, because what we usually mean by intelligence are reasoning capabilities, not e.g. motor skills. You don't say about a surgeon that he is smart, because he can manipulate a blade with great precision; we say he's exceptionally skilled.
> For the simplest example the average Joe will know his house better than the Google employee who hasn't even been at Joe's house.
Put Average Joe and Google Employee a house they have never seen before and see which one will learn how to navigate faster - that's a way to measure intelligence. Not the knowledge, but the ability to process and use it.
> If someone would say "singularity will be when machines will manipulate mathematical concepts and invent/discover and prove theorems that no human mathematician alive can understand" I would agree. But if they say "machine is more intelligent than any human" I can't agree.
Saying "machine more intelligent than human" is just a shortcut for saying "machine that is able to reason about the world faster, better, with less biases than human; which will manipulate mental concepts and prove theorems out of reach for humans, as well as invent better technology, tackle human social problems better than humans do, etc. etd.".
Re: Google to Buy Artificial Intelligence Startup DeepMind for $400M
#204Earlier quoted context omitted.
I am a bit of a jack-of-all-trades. I generally bill myself as the "&" in R&D, bridging the gap between academic researchers & mathematicians and mainstream software engineering. I.e. I can do a bit of classifier design, but most of my focus tends to be on build systems, automated test systems, parameter tuning, data quality monitoring and all the nuts-and-bolts stuff that most researchers don't want to be bothered w…
What impression did they give you during the interviews? In retrospect, do you think that was their whole goal, to get acquihired by Google?
Yes, in retrospect, I think that was the goal. They were interested in me for work on their machine vision product (my area of expertise), but the spiel was very much around how they had built up one of the world's best machine learning research groups. Over the years I have seen many great academics fail to produce much commercially, so perhaps I was less impressed with that than I should have been. :-)
Re: Google to Buy Artificial Intelligence Startup DeepMind for $400M
#205Earlier quoted context omitted.
> It's not because something is difficult to measure that it does not exist(1). No, I don't say that it doesn't exists. What I say is that intelligence is not something to be expressed in one number, it doesn't work that way. Intelligence is more like NxN matrix of numbers, where each number in a matrix represent individual skill in some specific task. As for your example of Google employee and average Joe, what you…
> What I say is that intelligence is not something to be expressed in one number, it doesn't work that way. But nobody in the singularity field says that intelligence is one-dimensional quality! It's a strawman. It reminds me of a common accusation that "computer people" have subpar worldview because they "reason in 0s and 1s, and the world is not binary", to which I say that actually "computer people"'s view is supe…
> This is also not a good model, because what we usually mean by intelligence are reasoning capabilities, not e.g. motor skills.
Yes, the model is not good, I agree. But your understanding of intelligence incorrectly. You think that intelligence is a rate of learning. Or in neuroscience terms, brain plasticity. So, first, plasticity is largest at birth and gradually decreases as brain matures. This means that newborn would be "more intelligent" than 50yo man. Secondly, a fast rate of learning is not necessarily a good thing. If you ever worked with neural networks, you'd know that when training it, you can adjust at what rate the weights in artificial neurons would change. If you make fast rate, for one, network quickly overlearns, meaning he becomes too specialized and adjusted to exact cases he experienced, and secondly, it can quickly "forget" what he has learned. Slow learning rate makes it longer to learn, but also is more "stubborn" and doesn't give of on old beliefs so easily either.
On the other hand, as you said, knowledge also does not mean intelligence. If it cannot learn from it's mistakes, it is surely not intelligent. So I'd say intelligence is a combination of experience and plasticity.
> Saying "machine more intelligent than human" is just a shortcut for saying "machine that is able to reason about the world faster, better, with less biases than human; which will manipulate mental concepts and prove theorems out of reach for humans, as well as invent better technology, tackle human social problems better than humans do, etc. etd.".
So in the other words machine becoming more proficient in some very specific skill or multiple skills. Yes, that is common sense.
It's not that I disagree with the core idea of singularity, I just find it pointless and unnecessary. Some might say "it brings attention to the field", but I'm not really sure it helps AI research, since it attracts the wrong kind of people that would make actual research.
Re: Google to Buy Artificial Intelligence Startup DeepMind for $400M
#206Google ∩ Nest ∩ Boston Dynamics ∩ Ray Kurzweil ∩ Andrew Ng ∩ DeepMind = Domestic Robot?
= attempt of implementation of sci-fi sounding idea that will be rewarded even if it fails.
It's is good.
Re: Google to Buy Artificial Intelligence Startup DeepMind for $400M
#207Here's a talk the founder gave in 2010 for Singularity U about combining neuroscience and machine learning. http://vimeo.com/17513841
Coming to the thread late, slides can be found here: http://chris.ill-logic.com/systems-neuroscience/ (plus the video)
Re: Google to Buy Artificial Intelligence Startup DeepMind for $400M
#208>The DeepMind-Google ethics board, which DeepMind pushed for.. Damn, if DeepMind had to 'push' for an ethics board then that is a fairly bad sign. I am getting more worried. [0] https://www.theinformation.com/Google-beat-Facebook-For-Deep...
What's to say what is an ethical thing to do to an AI? Or that someone decades from now will be ethical? The most ethical thing would be to shut down the program.
Re: Google to Buy Artificial Intelligence Startup DeepMind for $400M
#209Earlier quoted context omitted.
and does it matter? I mean, since it's a copy, to your biological "you", it's just no different as if they put some other random person in VR and told you it was you in there. I find the 'upload your brain to a hard drive' thing silly unless they figure out how to have your biological consciousness seamlessly move to the VR, like it was taking some bus ride and ended up in this new place.
In the old mans war books they cover that by having the old and new bodies staring at each other while they move the consciousness. So one minute you're staring at a new body then you're staring at both then gradually the vision of the new body fades out and you're left staring at your old body. I think something similar is the only way you're going to get people to want to do this. Though whether or not that is poss…
Sounds like an idea. If I really had to.
I would still have a similar problem with "teleporters" (of the Star Trek kind), though.