Singularity General

File: muhsingularity.png (13KB, 1000x1500px) Image search: [Google]

13KB, 1000x1500px

Here OP, this the one you lookin for?

>>7975392
Yessss, that's the one. Thank you anon! Saved.

>>7975392
>tfw all you see are control systems

>>7975392
Well, but isn't that the idea of a singularity?

>>7975438
What? Asymptotic time expansion? Unless you're expecting a new big bang made of ether or energy or technomagic or the like, no. A singularity is just a region of phenomenology that we can't predict.

>>7975445
>asymptotic
It's more sigmoid, though.
The blue graph could well be a sigmoid function.

>>7975449
Listen, have you ever seen nature not make bell curves? Have you ever seen humans not make bell curves? Have you ever seen technology not make bell curves? Have you ever seen AI researchers not make bell curves? Have you ever seen a seed AI make an AI that wasn't made of bell curves?

The law of norms holds for all widescale statistical measures. You can't escape normality.

>>7975457
Just because it was that was a thousand times in the past doesn't mean it will be that way the next time.

>>7975460
I'm saying that it will necessarily be that way if you take a normative approach to it. That is, I'd be sympathetic to your argument if you were doing something to isolate the peaks rather than just deferring to arbitrary "before" and after boundaries with no calibrating measurements whatsoever. In its current form your argument is no different from, "Well, in theory, anything's possible," and as a fairy I kind of have to agree, but that doesn't mean that statistics doesn't matter either. Probabilistic reasoning is more than enough to say that it will be that way in sample bucket 1001 if it was that way in 1000 other buckets. For you to gesture at some arbitrary violation of the law of norms is by no means a valid optimization path.

>>7975457
>Listen, have you ever seen nature not make bell curves?
The Universe, ever since Big Bang, does not appear like a Bell curve.

>>7975530
1. Galaxies lump together in spacial distributions, represented by bell curves.
2. Stars are more numerous at the center of those same galaxies, represented by bell curves.
3. Stars themselves have disparate nuclear layers, with different pressure levels creating different reactions in a bell curve fashion.

I don't think you understand how statistical analysis works.

Daily reminder that we have a thread for singularity propaganda so you don't have to manicpost in all the other threads OTB.

>Kurzweil predicts the singularity to occur around 2045 whereas Vinge predicts some time before 2030

So?

>>7977918
I'd go with Vinge, personally. Not because there is even the slightest bit of accuracy in the prediction, but because predicting that the economy occurs sooner means a higher rate of visionary qualities.

>>7975457
Yes, I have. Sigmoids, for example, are very common in many fields. Check out the cumulative graph for the normal distribuition. And events are described with exponential or gamma functions.

>>7977922
I guess I'd need to define "nature" better to make that rhetoric work. The point of the blue line is that it goes well beyond the domain of the graph.

Once we have enough computational power we can use supercomputers to plan our economy, or at least to a large extent.
This kind of planning can be done on both - the local, firm level and the central level. That is, prices and other things which are very chaotic could be quickly computed on the firm level, taking into account the product stock which the firm has (the goal is to adjust a price until making it equal to demand at all times).
Big investment decisions, on the other hand, could be computed on the central level, taking into account all the inputs available.

Not to mention, such an economy could be Pareto efficient, just as a market economy is if it reaches perfect competition. This can be done by making firms produce at the minimal Average Total Cost. If one takes modern economic theory into account, such an economy could, in theory, be even more Pareto efficient and closer to equilibrium than a classic, laissez-faire capitalist market economy.

Obviously, all the economic decisions which can't yet be made by machines efficiently and can't be easily programmed into an algorithm could be made by humans. But eventually, as computational power increases, humans would take less and less place in this process of economic planning and management.

What are the benefits? The fact that the entire profits of the economy would no longer accumulate in the hands of the few. They would instead be used for things such as further investments to achieve spectacular economic growth, investment into R&D to achieve high levels of scientific and technological progress, space exploration, and so on.

Planned economies failed previously because they used a form of accounting that would never result in anything close to Pareto efficiency. But in this system, Average Total Cost would be minimized, so such an economy would approach Pareto efficiency by definition, and the kind of inefficient accounting that Soviet-type planned economies had would be absent.

>>7977946
Also, before someone says that we don't have the computational power for this and so on:
It can start off with humans still making most of the decisions. But humans can be gradually substituted for computers as computational power grows and as we write new algorithms.

And this can at least be tried in a small economic zone at first, as an experiment - just like capitalist was introduced in a small, special economic zone in China, for example. A zone small enough that there would be enough computational power and the human managers wouldn't be overwhelmed neither.
As computational power grows, this zone could expand, gradually replacing the old laissez-faire economy with a more efficient one.
This kind of economy still follows the laws of the market to a large extent (and it follows contemporary theoretic economic models even more) , it simply does so more efficiently than laissez-faire economies which we have now.

File: 1 BResEFDGME899cCp0j8xeQ.gif (865KB, 700x994px) Image search: [Google]

865KB, 700x994px

>>7975392
>he thinks human intelligence is the peak and thus the end point of intelligence
>he thinks we are even remotely close to what is physically possible in technology (halfway there, according to the graph)

>mfw Hassabis solves intelligence
>mfw I upload myself to the cloud
>mfw my life will consist of living in a simulated utopia for eternity

How about a bit of optimism.

>>7977971
Except that Moore's law will hit a wall sometime during 2020-2022

File: Calculations.png (147KB, 802x540px) Image search: [Google]

147KB, 802x540px

>>7978309
Exponential growth in computing power started long before Moore and will keep going.

>>7976801
>I don't think you understand how statistical analysis works.
I don't think your reading comprehension is adequate. I talked about the *universe* not *parts of* the universe.

File: image.jpg (49KB, 488x372px) Image search: [Google]

49KB, 488x372px

Have you heard of DREADDS?

>http://pdspit3.mml.unc.edu/projects/dreadd/wiki/WikiStart

You can design a nervous system with DREADDs. Such an artificial nervous system can be used in organic implants, or used to interface with non-organic implants.

My plan is to use DREADDs to mark my neurons, then inject molecules that seek out the DREADDs and cover the neurons with a molecule that won't be dissolved with sulfuric acid. Then, design a virus that'll produce an inclusion body inside the neurons which will interact with other inclusion bodies in neighboring cells like a nervous system.

Then, I'd toss my body into a vat of sulfuric acid and oxygen until all that was left was the neurons and inclusion bodies. Finally, I'd move my remains to a 3D printer, and lay down a new body made of graphene.

I have dreams of posting my pictures of my lab and new body, but I doubt anything like a forum will exist by then, on top of the fact that no one will care.

>>7977946
>enough computational power
You mean like a mass of seven billion sapient entities, all capable of achieving the same basic tactical objectives, with varying degrees of delivery?
>computed on the central level, taking into account all the inputs available
And which manner of sapience oversees this, in your vision of the future?
>as computational power increases, humans would take less and less place in this process of economic planning and management
Isn't that basically the same as saying that we're already Pareto efficient for our time period?
>the entire profits of the economy would no longer accumulate in the hands of the few
Says who? I see just as much potential for centralized cash flows in your centralized economy as there might have ever been in a "covertly" centralized economy.
>>7977949
>A zone small enough that there would be enough computational power and the human managers wouldn't be overwhelmed
Those are called cities and mayors, respectively.
>>7977971
>tfw can download a copy of you through illicit channels
>>7978339
...Until it hits the limit of what's physically possible.
>>7978350
Then what visualization method are you using? I'm not going to argue an incomplete claim.
>>7978415
>the neurons and inclusion bodies
Sounds painful. Are you a masochist by any chance?

>>7974892

singularityfags please leave

>>7978569
>...Until it hits the limit of what's physically possible.

which is much, much, much,...,much more than what we need for superhuman AI

>>7978573
at least he posted a qt gif, what do you have to offer?

>>7975366
Oh bother. We can continue in this thread if you like.
>>7978583
>superhuman AI
Well I guess I did start the thread so I kind of have to tolerate this argument. I'll bite.

What is your basis for predicting the computational limits required for superhuman AI?

>>7978658
Does AI's computational power have to be superhuman or do it's results have to be superhuman?

Humans aren't really made for any of the intellectual challanges that modern life and science pose. A lot of human computational power is wasted on our biology and even more is wasted because our brain is not suited for modern society. So even if AI's computational limit never reaches human level (unlikely), it could still be superhuman at literally every meaningful task.

>>7978677
(N)either. You're the one that brought it up.

I don't value modern society. We can as well say that modern society and human society don't have a shared fitness function. AI shouldn't be made for the sake of maintaining anti-human societies.

>>7978658

>What is your basis for predicting the computational limits required for superhuman AI?

http://arxiv.org/pdf/quant-ph/9908043.pdf?origin=publication_detail&sa=U&ei=Si1nU_2BBui88AHupoGgCQ&ved=0CDQQFjAE&usg=AFQjCNFyeJuo-A9rrIxHSPPNGv3mKVjYDQ

my guess is the human brain doesn't go that far.

>>7977971
Why haven't we done full brain emulation of insects yet?

>>7978713
>that far
Which page on how far? As far as I can tell this paper is about Moore's law and I don't have any bit of confidence that reading it will tell me about any theoretical model of digital cognition.

File: PYHCl2y.jpg (260KB, 1196x1200px) Image search: [Google]

260KB, 1196x1200px

>>7978727
Why would we?
>implying biological brains are the end-all of intelligence
>implying brain emulation is needed for AGI

Also look up OpenWorm, they are working on something like that, but are not quite there.

>>7978745
>If, as seems highly unlikely, it is possible to extrapolate the exponential progress of Moore’s law into the future, then it will only take two hundred and fifty years to make up the forty orders of magnitude in performance between current computers that perform 1010 operations per second on 1010 bits and our one kilogram ultimate laptop that performs 1051 operations per second on 1031 bits.

kurzweil says the brain will be matched somewhere around 2025.

there are 225 years of moore's law between kurzweil estimate and their theoretical ultimate laptop.

>>7978759
>kurzweil says the brain will be matched
Why? How does bitwise computation map to the abstract general intelligence the human brain has? Is there any theory behind this or are the predictions entirely arbitrary?

>>7978569

>Sounds painful. Are you a masochist by any chance?

My ability to feel pain is decided by whether or not my neurons have various kinds of TRP receptors. These receptors can be shut off by attaching permanent antagonists to them.

There's something called antihypergesia, and it can be induced by cannabinoids. So I'm 100% sure there'll be no pain.

>>7978915
>cannabinoids
Those only work when your brain is working. That is, in transitioning to another type of neural network, you can't rely on that existing stimulus modes at all. It might work for 30% of the process and start becoming excruciatingly painful after your limbs (regret regulators) have dissolved. I'd recommend thinking in terms of multiple Theseus steps if I were you.

>>7975457
>Listen, have you ever seen nature not make bell curves?

Jesus christ

Normal distributions are a feature of collections of independent events. For strongly interdependent systems like human civilization, there is no reason to expect normal distributions. Also, you're proposing that a probability distribution is the realistic alternative to a time series.

>>7976801
Bell curves do not represent things clustering in the literal physical middle of something. I don't think you understand how statistical analysis works.

>>7979012
>strongly interdependent systems
>like human civilization
That's a bit of a stretch but I'll roll with it since it's a premise of the technological singularity fantasy.
>>7979012
>you're proposing that a probability distribution is the realistic alternative to a time series
I am actually, that's fully intentional. The reason is that I don't consider human intelligence lacking in any general capacity. To me, we're just as smart as we were 200,000 years ago. Since nobody in that time period has achieved a tactical singularity of information, we can assume that intelligence is already evenly distributed in an independent fashion. Or else partially independent, if nothing else.

Basically I don't believe in superintelligence. The questions in the OP are there precisely because I expect to be just as bored post-singularity as I am now, supposedly "pre"-singularity. The tech won't make a bit of difference in how we communicate.

>>7979041
>human civilization isn't strongly interdependent

So you'll be fine if all the farms fail? You'd know just as much without books and the Internet? You'd be able to travel just as far without vehicles?

>I am actually, that's fully intentional.

Then you're just doing free association and shouldn't be casting aspersions on that other anon's grasp of statistics.

File: kim.jpg (76KB, 1024x765px) Image search: [Google]

76KB, 1024x765px

Kim is best waifu.

>>7979050
>you'll be fine if all the farms fail
Technically yes, but I realize not everyone is in the same position. The point is more that people can react to that situation intelligently and it doesn't generally compose a mass extinction. Malnutrition doesn't become lethal for months if you have water and farms aren't generally where we get our water. Dehydration kills a lot faster than hunger does.

I'm not saying it's an entirely independent or unconnected system, but the degree to which it's still capable of independent individual functioning means that it's still moderately accurate to use the bell curve to model its systemic traits.

>>7979097
>means that it's still moderately accurate to use the bell curve to model its systemic traits.

By definition, the more systemic the trait, the less independent the events.

Still doesn't make it remotely appropriate to swap a bell curve in for a curve representing any process in time.

>>7979083
Alright that's a stronger form.
>know just as much without books and the Internet
Conceivably! It depends on how resources would get distributed in a world without those systems. I might well have become one of the people that did end up studying under the grandmaster of science from the preceding generation. It is precisely the fact that we only need one active intelligence to make the entire chain of knowledge work that leads me to make the generalizations I have about the nature of intelligence. The only thing that does pierce the law of norms is the very thing that's able to conceive of the law of norms in the first place. It is precisely that trait that convinces me that we tend toward possessing general intelligence. The internet is just a convenient distribution system for something that was only able to exist because we worked as individuals to build it in the first place.
>able to travel just as far without vehicles
I assume you mean automobiles. I'm not sure that travel plays much of a role in terms of knowledge advancement, but if the internet weren't here and I had to hoof it or walk to find all the knowledge that makes science relevant, I think I would. What this argument serves to illustrate is precisely the fact that we have systems that are relevant to advancing science and systems that definitely are not relevant.

>>7979122
>I might well have become one of the people that did end up studying under the grandmaster of science from the preceding generation.

So you think societies with oral traditions produce just as much knowledge as societies with writing?

>I assume you mean automobiles.

Any vehicle too complex to personally build will do.

>I'm not sure that travel plays much of a role in terms of knowledge advancement

Knowledge is not the only way you depend on others, is the point

>>7979108

>>7979127
[to add, saying you 'might well' have found your way to a sage is still not an argument that your knowledge is independent of your relations to the rest of civilization]

I couldn't read all of this thread

Acting like "life" never follows exponential growth is completely wrong. If you are using the argument that the pattern of a bell curve is more likely than exponential growth you are just wrong.

Almost everything related to humans and life in general follow exponential growth if conditions are right.

Meaning, as long as we don't go into nuclear war or have a super volcano blow we will probably go exponential and reach singularity.

>>7979108
>a curve representing any process in time
Lots of resources only exist for the duration which processes them. It's often less about the causal dynamics of the system than the fact that the system only functions when its resources are present and accessible.
>>7979127
>societies with oral traditions produce just as much knowledge as societies with writing
Yes, actually. I realize it's not going to be a very popular stance, especially in a singularity thread, but I don't think that writing things down produces a greater frequency of rational thought. Population density is the main driver of advancing intelligence, and access to written works can only substitute for a certain ratio of "artificial" population density.
>the only way you depend on others
Right, but knowledge is the one that doesn't follow the bell curve/law of norms. All other types of resources are primarily irrelevant to the advancement of knowledge.

>>7979166
> Lots of resources only exist for the duration which processes them

Again, a probability distribution just is not interchangeable with a time series. If you swap one for the other, you probably shouldn't comment on either.

> I don't think that writing things down produces a greater frequency of rational thought

Knowledge and thought are not the same thing.

>>7978748
>Why would we?
I came here from the main page without any idea about computers just to tell you that that's the most retarded excuse I've ever heard

there are trillions of inventions that make a person just think "what is even the purpose?" things like pillows with the form of a butt, dildos shaped like the head of obama or computer programs solely dedicated to shitposting

for fucks sake, this exists https://www.youtube.com/watch?v=HtIx5F3516c

there doesn't need to be a reason for something to be created

>>7979158
>Acting like "life" never follows exponential growth
Life never follows exponential growth. Cancer dies out just as much as anything else does when the fuel runs out. You ALWAYS hit a carrying capacity simply because you exist in a closed environment. You need x to keep increasing indefinitely for y to continue its exponential growth, and x is limited.

Literally, superexponential growth cannot continue at an indefinite pace.
>if conditions are right
...Which means having resources at your disposal.

It only ever appears exponential at the beginning of the bell curve.
>>7979177
Not thought, rational thought. Rational thought is the type of thought that produces knowledge. It's not guaranteed to do so, but it is a requirement.

If anything, I'm using multiple separate time series to construct a probability distribution. No swapping here.
>>7979189
>for fucks sake, this exists
Not anymore it doesn't. Normally I'd ask why, but...

>>7979254

>If anything, I'm using multiple separate time series to construct a probability distribution. No swapping here.

Nah, you literally told other anon that a bell curve was a better representation than a sigmoid curve representing technology over time.
And that literally everything is normally distributed.

>>7976801
>>7975457

>>7979297
Yes, because resource distribution dictates the outcomes. The time series is there, but it'll only be there when the resources, which follow a distribution curve, are actually there.

>>7979328
Probability distributions are not about resources getting to places.

>>7979343
Resources getting to places ARE about probability distributions though.

A priori, at least.

>>7975457
evolution doesn't make a bell curve.

>>7979410
Evolution makes nothing BUT bell curves. Speciation is possible precisely because the peak of a gene pool's bell curve is selected against.

>>7979410
What a silly statement. Check out the natural variation in size among any species of animal. You'll find bell curves galore.

>>7978991

>Those only work when your brain is working

The definition of a working brain includes neurons covered in TRP receptors. Anasthetics work by inhibiting the sodium channels that TRP receptors control.

To feel pain, you need active TRP receptors and sodium channels. Covered in permanent antagonists, they would never relay a pain signal again. Once the neuron rotted away, leaving the inclusion body, the physical structure that produces pain would cease to exist.

>That is, in transitioning to another type of neural network, you can't rely on that existing stimulus modes at all

True - I have no idea what to expect. But I might be able to transform small areas of my brain first, before going all the way.

>It might work for 30% of the process and start becoming excruciatingly painful after your limbs (regret regulators) have dissolved

I've felt exactly what you're talking about. They never told me that my esophagus would feel like leather, and my stomach would feel like a breezy hole that can never be filled.

>I'd recommend thinking in terms of multiple Theseus steps if I were you

What's the difference between multiple steps taken over a day, or those same steps taken over a year?