Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

There's No Fire Alarm for Artificial General Intelligence

intelligence.org

Post reply on HN
Print view View original

161–170 of 209 posts

Re: There's No Fire Alarm for Artificial General Intelligence

#161

Post by Tossrock »

UncleMeat wrote: post #128
Tossrock wrote: post #77

For me, the "smoke under the door" moment was Karpathy and Li's Deep Visual-Semantic Alignments for Generating Image Descriptions [1]. The almost perfectly grammatical machine-generated captions of photos were unnerving to me in a way that simple categorization was not. It somehow called to mind the image of a blank-eyed person speaking in a monotone while images flashed in front of them. What if they wake up? [1]: h…

Yet Andrej himself thinks that RNNs are not really making meaningful progress towards AGI.

Well, if I were to play the Yudkowsky's Advocate here (which in general, I'm not), I would say that this is precisely what he's talking about in the article. Because Karpathy knows how hard he's had to work, and how flawed and dumb in particular ways the current techniques are, he may overestimate the distance to AGI.

Now, generally I disagree with Yudkowsky on a lot of points, but I do think he raised some decent ones here.

Re: There's No Fire Alarm for Artificial General Intelligence

#162

Post by zardo »

randomsearch wrote: post #160
zardo wrote: post #138

Earlier quoted context omitted.

>ML in general is just applied statistics. That's not going to get you to AGI. I don't see how we can rule it out. The size of the statistical models we use are still dwarfed by the brains of intelligent animals, and we don't have any solid theory of intelligence to show how statistics comes up short as an explanation.

We can learn concepts rapidly from much less data than statistical methods require.

One shot and zero shot learning also use statistical models.

Re: There's No Fire Alarm for Artificial General Intelligence

#163

Post by gooseus »

johansch wrote: post #145
gooseus wrote: post #140

Earlier quoted context omitted.

I don't think I am... if a task is possible, but takes 100x the lifetime of the universe to accomplish, is it actually "possible"? Or is accomplishing that task the same as "magic"? It I send a message using a one-time pad, the other person knows what I sent, you can ask questions and see that communication is actually happening... so if we're not using magic to communicate, it must be possible to communicate in the…

That doesn't really make much sense to me, I'm afraid. Please bear with me, but my first guess is that you have invented a coping mechanism that allows you to deal with conflicting information in your mind. Are you by any chance religious? (I figure I should be allowed to ask such a normally speaking quite loaded question because of my previous statements, up the thread.)

No, I am not, by any chance, religious. I don't rule out the possibility of a God or God-like intelligence existing in the Universe, but that's simply as a result of the "absence of evidence is not evidence of absence" principle I hold to.

I don't really see how religion factors into this though... I feel like I'm talking about a simple concept too. If I show you an encrypted message and show you that other people can read the contents with a key, then ask you to read it without the key, why can't you? It's not magic, it's math.

Re: There's No Fire Alarm for Artificial General Intelligence

#164

Post by eli_gottlieb »

nurettin wrote: post #27
simonh wrote: post #13

Earlier quoted context omitted.

A conscious system might be unintelligent, while it’s possible to imagine a highly intelligent system with no consciousness. They’re just different things. Also pattern detection is often raised in the way you just did, but it’s realy a distraction. Pattern detection just helps recognise things, it’s not inherently related to the ability to reason about things. So you need both, but they are not the same thing either…

But where does consciousness arise? Is the ability to reason about things independent of this concept?

> But where does consciousness arise? Is the ability to reason about things independent of this concept?

"Who cares?" and "yes".

Re: There's No Fire Alarm for Artificial General Intelligence

#165

Post by eli_gottlieb »

backpropaganda wrote: post #5

The only non-speculative and relevant claim here is that the experts were wrong about Winograd Schemas. The paper Eliezer cites to prove that we've made unexpected progress in Winograd Schemas only deals with a very specific type of Winograd Schemas, and not an arbitrary one. This is awfully dishonest for someone purporting to be a skepticist. Also, the wording seems to imply that WS performance is already pretty hig…

>Alphago was not such an event. Yes, we did predict that Alphago is decades away, but that's assuming that academics will continue working on it at their pace using their limited resources. No expert was surprised with Alphago. No expert will be surprised when Starcraft or Dota is solved. It's simply a matter of compute and some tricks here and there. Why? Because these are closed systems, with good simulators available. You just need to keep playing and storing the actions in a big lookup table a la Ned Block, and you're done.

AlphaGo worked according to statistics, not lookup tables. Bit of a difference.

That said, theoreticians may not have been surprised, but there's a huuuuuge difference between what's doable in theory (sufficiently large neural nets are universal function approximators, after all), and what the resource requirements for problems we care about actually turn out to be. We should all have been fairly pleasantly surprised that AlphaGo required only a small data-center worth of graphics cards for training, and could then play on less hardware than that.

Re: There's No Fire Alarm for Artificial General Intelligence

#166

Post by afthonos »

randomsearch wrote: post #158
afthonos wrote: post #149

Earlier quoted context omitted.

You’re engaging in the time-honored tradition of dismissing progress with the term “just”. In the spirit of the article, I recommend you list and publish specific things that are too hard to achieve in the next five years. And then commit to not dismissing them post-hoc.

This is a really difficult question to answer, because humans are super-clever. Whatever I pick, humans can quickly work out ways to game the metric. Turing invents the Turing test, and rather than go for an AGI, we take shortcuts to spit out convincing sentences. I think for a start you'd have to move away from things that can be gamed through statistics on large amounts of data. For example, show a child a single o…

A very interesting opinion combo:

1. Of course we should be prepared for the existential threats of AGI and ASI. 2. BUT the threat isn't imminent, so we should prepare later.

The article (and I, mostly following its lead) is trying to encourage people to concretely answer the question "Okay, if not now, when? How will you know?"

The problem is, most people aren't answering based on a model ("If we can solve problem X, we have 50% probability of AGI within X years.") Instead, they're using the difficulty heuristic, and the insufficiently-impressed heuristic. ("This is really hard right now, and I'm not impressed by what I've seen so far. Therefore, 100 years.")

Your concerns about gaming are only a problem if the notes were to form the basis of an argument. I was suggesting you have them for yourself. (The act of publishing them is to encourage thinking about them now, not to be a gotcha later.) So you'll know what you mean, you won't be arguing with yourself over definitions, and you'll have thought hard about what looks dangerous to you. It's about being honest with yourself.

Incidentally, the child example is misleading. Children spend literally years understanding things like depth perception, object permanence, etc. A child is already a highly trained agent; that training comes from daily interaction with the environment. You show a baby an unmoving object somewhere and to my knowledge there is no evidence that the baby will identify it as a separate object, much less recognize it in a different configuration.

Re: There's No Fire Alarm for Artificial General Intelligence

#167

Post by eli_gottlieb »

pilingual wrote: post #51
apsec112 wrote: post #26

Earlier quoted context omitted.

That's interesting, but could you elaborate in more detail?

https://twitter.com/fchollet/status/906582914829246464 (Google researcher) DeepMind and OpenAI have been investigating approaches from cognitive science in recent months. In particular they seem interested in evolutionary algorithms. DL applications are still emerging though, such as the company that demonstrated using GANs to present models fitted in apparel a few days ago.

>DeepMind and OpenAI have been investigating approaches from cognitive science in recent months.

Really? Got any links. That might be exciting to read.

Re: There's No Fire Alarm for Artificial General Intelligence

#168

Post by sampo »

jonex wrote: post #58
denzil_correa wrote: post #35

Earlier quoted context omitted.

> My take is that it's worth taking pragmatic steps towards studying AI safety measures (i.e. OpenAI), but not going as far as to talk the likes of 'AI research regulation'. Sometimes, it makes more sense being cautiously optimistic (pro-active) rather than reactive. We have already gone down that reactive slope and it's better to act now before it's all too late [0]. [0] https://blogs.scientificamerican.com/roots-of…

Ignoring the topic of the linked article* I'd argue that there's examples of being too cautious as well. There's a lot of good that we could have done with GMO that is not being done because of very restrictive regulation. Ironically it means that GMO is mostly used for things that are not as obviously good because that's where there's enough profit to be made in the short term to make the research worth it . I'm a b…

If AI scientists actually believed that the general public will believe the talk about existential threats, they would be afraid of activist groups sabotaging and occasionally firebombing their laboratories. Like sometimes happens with GMO research. Clearly they are not.

Re: There's No Fire Alarm for Artificial General Intelligence

#169

Post by ciphergoth »

Balgair wrote: post #131
sago wrote: post #33

I've been around AI since the end of the last big hype in the late 80s. The recent leap in machine learning has felt rather hyped to me. I don't think AGI is near. But I find myself agreeing with this article. Strongly. And I have long suspected that, we miss a lot of the significance and opportunities in AI, because we have only one exemplar of 'higher' intelligence: a human being. AI folk are so concerned with gett…

> The superhuman fallacy really is the bane of AI Great point, I'll be stealing that one ;) One annoyance I have heard about SV is that all the companies are just trying to replace your Jewish Mom: Uber/Lyft is Mom's minivan, GrubHub/DoorDash/BlueKitchen is Mom's cooking, Google is Mom's encylopedia, Yelp is the synagogue's meeting hallway, Tinder is your Mom's yenta, etc. The examples abound in a non-B2B space. In t…

Bulverism is a poor substitute for actually thinking about the future. See Scott Alexander's excellent recent essay: http://slatestarcodex.com/2017/10/09/in-favor-of-futurism-be...

Re: There's No Fire Alarm for Artificial General Intelligence

#170

Post by ciphergoth »

pdimitar wrote: post #121

> They will believe Artificial General Intelligence is imminent: (A) When they personally see how to construct AGI using their current tools. This is what they are always saying is not currently true in order to castigate the folly of those who think AGI might be near. This struck a nerve. Too often, in many scientific disciplines, and even in informal conversations, the people who always demand 100% clear evidence u…

This is an error Eliezer has also written about: http://lesswrong.com/lw/1ph/youre_entitled_to_arguments_but_...
Post reply on HN