Could AI be programmed to have intuitions nor irrational behaviors?
Superintelligence: The Idea That Eats Smart People
531–540 of 607 posts
Re: Superintelligence: The Idea That Eats Smart People
#532Earlier quoted context omitted.
An uncaring blind AI wouldn't be very interesting, I would assume whoever had turned it on and was observing it would just turn it off and tweek the algorithm. But I don't think this is even possible as when the AI was in child stage and was learning, it would learn from people, so it would become like people. Or at least understand them. At some point it has to know less than a human and will learn at a rate that hu…
> I don't agree with the assumption that a factory making paperclips will make the transition to a super intelligent AI in a short time frame. Why? Because factories run by humans take a long time to transition? It might take weeks or months to make a process tweak in a human-run factory, but deploying new code happens in seconds or faster.
Yes because the deplyment of the code is the hard stage. rolls eyes
How long would it take to write -- scientists are going through this stage right now and it's taking about 50 years so far. You think a paperclip factory is going to be quicker? Nope. And any smart AI that is being used to design a better AI will take years and will be front page news (assuming non-military) just like any other tech company (funding, profit, getting good hires)
This assumption that the factory is smart in secret and buys more servers in secret and creates code somewhere in secret is just so wrong. It will be dog or dolphin smart at some point and will ask for things, it'll communicate with us.
Re: Superintelligence: The Idea That Eats Smart People
#533What the author is actually talking about is of course the mythological techno-singularity event, which is of course bullshit. Have you ever programmed any piece of software that suddenly implemented features of its own? Did your program become sentient? Laughable. These are child-like fantasies belonging in 50s sci-fi. What I do know for a fact, is that when it comes to AI fear mongers, there is not much intelligenc…
>Have you ever programmed any piece of software that suddenly implemented features of its own? I'm not sure what particular fear mongers you're talking about, but Bostrom and the like are talking about software that's specifically made to change and improve itself on a fundamental level.
This is no different than Facebook ads/news feed adapting to your interests and only showing what you like in the hope of maximizing total engagement and retention.
We all know that Google/Facebook/Bing build a little bubble for you and customize results and ads based on your behaviour.
Re: Superintelligence: The Idea That Eats Smart People
#534Earlier quoted context omitted.
Too late to edit, so I'll post just a few examples here: >The only way out of this mess is to design a moral fixed point, so that even through thousands and thousands of cycles of self-improvement the AI's value system remains stable, and its values are things like 'help people', 'don't kill anybody', 'listen to what people want'. Bostrom absolutely did not say that the only way to inhibit a cataclysmic future for hu…
Well, the thing is there is no such thing as 'machine intelligence', so it's all just an assumption on top of an assumption about a thing we don't have a very good grasp of yet. You're essentially saying that the author is wrong for saying the philosopher's stone can't transmute 100 bars of iron to 100 bars of gold, because a philosopher's stone could absolutely do that type of thing, because that's what philosopher'…
I did not say that the a machine intelligence mustn't share motivational complexity a la humans. I said that a SAI need not share such complexity. Those are two very different statements.
And to understand how/why a machine intelligence could arise without being substantially similar to a human intelligence and sharing similar motivations, well, I suggest you read the book or similar articles. In short, just because humans are the most sophisticated intelligences of which we yet know, it would be a very callous and unsubstantiated leap to believe that a machine intelligence is likely to share similar traits with humankind's intelligence. If this is unclear to you, I recommend you learn about how computer programs currently work, and how they're likely to improve to the point of becoming superintelligent.
By the way, there are many types of SAI, for example an SAI who's superintelligent portion relates to speed, or strategy, or a few other types.
Re: Superintelligence: The Idea That Eats Smart People
#535Earlier quoted context omitted.
Please see my spirited defense from 13 hours ago: https://news.ycombinator.com/item?id=13242592
I saw that, I still disliked the comment enough that I needed to write that. Also to claim that the book has no concrete theory in the same sentence as you stating you're not familiar with the book. Like, c'mon...
That isn't what I claimed.
Re: Superintelligence: The Idea That Eats Smart People
#536>Hopefully you'll leave this talk a little dumber than you started it, and be more immune to the seductions of AI that seem to bedevil smarter people. I rather like the smarter people stuff - it's kind of exciting to figure how AI could solve human problems and stuff like the OpenAI initiative seem a sensible precaution against things going wrong. The arguments against seem a little reminiscent of arguments against g…
It's a commendable initiative but the real value of AI is in its data. I think about data as memories. The algorithm learns from its memories and creates a map of abstract concepts and behaviors which it uses to make plans to maximize some goal.
Currently in the name of free services the big tech Co's collect petabytes of data about us and learn from it. The data is owned by them.
Re: Superintelligence: The Idea That Eats Smart People
#537Earlier quoted context omitted.
I don't buy the AI quasi-religious stuff. But your argument here is flawed. If protein folding can do the process in constant time, we may be able to find another process (but electronic rather than wet chem) that can also do it in constant time.
Being able to find constant time algorithms for algorithms that currently take exponential time is not at all assured.
I agree. I wasn't thinking about finding a constant time algorithm, though - more of finding an analog circuit that would mimic the behavior. After all, proteins don't actually fold by following an algorithm; they fold by responding to physical forces.
Re: Superintelligence: The Idea That Eats Smart People
#538"I live in California, which has the highest poverty rate in the United States, even though it's home to Silicon Valley. I see my rich industry doing nothing to improve the lives of everyday people and indigent people around us." This is trivially false. Over a hundred billionaires have now pledged to donate the majority of their wealth, and the list includes many tech people like Bill Gates, Larry Ellison, Mark Zuck…
Yes, and even if you ignore philanthropy, the tech industry generates enormous amounts of tax revenue, which is supposed to be spent by the government to help improve the lives of "everyday people and indigent people". A question people don't ask enough is: given that we give vast trillions of dollars to the government, most of which is spent on various kinds of social programs (health care, education, social securit…
so Apple, Google all pay full taxes in Cali? all profits booked into HQ?
wow, they're awesome.
Re: Superintelligence: The Idea That Eats Smart People
#539Moore's Law (on a single CPU core) effectively ended around 2009. Our programming idioms for parallel processing are still so crude, hardly anyone can write maintainable parallelized code for anything but toy problems. Until we can restart Moore's Law again, this is all academic. http://www.globalnerdy.com/2007/09/07/multicore-musings/
But the law says nothing about cores or clock speeds etc. just transistors and their sizes. The regular reductions in feature size on dies are still happening, as I understand it.
Size of copper atom is 0.2nm. We are already at 10nm fab processes. We can't really get too small otherwise it gets unstable and heat becomes a major issue.
On the positive side, electric flows much faster than synapses and transistors are smaller than neurons. So with today's technology if we were to pack silicon parallel cores in size of human brain we'd have a much faster brain.
Re: Superintelligence: The Idea That Eats Smart People
#540Earlier quoted context omitted.
> I don't agree with the assumption that a factory making paperclips will make the transition to a super intelligent AI in a short time frame. Why? Because factories run by humans take a long time to transition? It might take weeks or months to make a process tweak in a human-run factory, but deploying new code happens in seconds or faster.
Yes because the deplyment of the code is the hard stage. rolls eyes How long would it take to write -- scientists are going through this stage right now and it's taking about 50 years so far. You think a paperclip factory is going to be quicker? Nope. And any smart AI that is being used to design a better AI will take years and will be front page news (assuming non-military) just like any other tech company (funding,…
> How long would it take to write -- scientists are going through this stage right now and it's taking about 50 years so far. You think a paperclip factory is going to be quicker? Nope. And any smart AI that is being used to design a better AI will take years and will be front page news (assuming non-military) just like any other tech company (funding, profit, getting good hires)
It may take years or decades, but the point is that once it reaches the point that it can make itself smarter, it'd go from "barely smarter than a human" to "inconceivably smart" in a matter of milliseconds.
> This assumption that the factory is smart in secret and buys more servers in secret and creates code somewhere in secret is just so wrong. It will be dog or dolphin smart at some point and will ask for things, it'll communicate with us.
Dogs and dolphins don't communicate with us very effectively. I think it's very plausible that the AI could get to beyond-human-level smart without ever spontaneously deciding to talk to us. Are you sure there's a region where it's smart enough to talk but not smart enough to hide how smart it is?