/sci/ - Science & Math

>>10141931
yeah, total meme. AI is as far away as actually understanding NATURAL intelligence (look it up, we have no idea how that works), and the other two memes are basically small variants of that. don't worry about it bro, bunch of cognitive scientists stirring up popsci idiots so they can get funding

>>10141931
It sounds legit to me. There are a lot of pop skeptics out there who reject any attempt to predict the direction of future research and technology, and who cannot think beyond what they can see with their own eyes today; but taking our current understanding to its logical conclusion gives you all the ideas you mentioned. It is entirely possible that something unforeseen might throw a wrench into those predictions, but it's still the _likely_ direction even if it is not certain.

>>10141931
>>singularity
Moore's law is ending, we'll be lucky if it happens. Not only that, there may be a stagnation period where there is no or little improvement to compute power because we hit the end of the road for conventional semiconductor processing and need something radically different. Oh yeah and stuff like fault tolerant quantum computing is decades away.
>>transhumanism
working with biology is hard. The field is constrained by how fast it can advance by how fast things can grow.
>>AI
we might figure out AGI... and then find that our approach to AGI is so computationally inefficient that it's much slower than humans. AIXI has already solved AGI, it's just uncomputable

https://www.unz.com/akarlin/transhumanist-techs/
https://www.unz.com/akarlin/short-history-of-3rd-millennium/

>>10142005
Why do you keep promoting your alt-right website and its pop-sci articles?

>>10142017

>>10141931
Might be the single most meme thing in (non-)existence, I have no idea why media keeps spamming it

>>10141962
2019 is the year Quantum Supremacy is expected to take place. Even now all sign point to universal quantum computer as the next big thing. Invest.

>>10142322
Sure, but not for fault tolerant quantum computers and only for a very limited set of applications. Good luck getting a circuit depth of > 100 by 2019.
>>universal quantum computer
I just had one of the experts in quantum computing tell me this is probably 50 years away

AI having a profound change and impact to society is literally just common sense.

>>10141931
>singularity
in the form Kurzweil wants it, yes. It will happen but it will take far more time and it will be far more gradual as the complexities of the mind aren`t simple computional prowress.
>AI
if smart or not, that doesn´t matter. AI will have a profound effect on politics and economey just like Computers in General. True self Aware AI are a century or so away from us though
>transhumanism
sure, genetic Engineering and human Augmentation is Happening right now but it will only pubivly avaible in 20 years and be very much regulated.

Dont be afraid from transhumanism, it will probably advance the world just like the printing press, the steam machine and the Computer did.

>transhumanism
total meme

>>10142414
But a universal quantum computer already exists. It has about maybe 10 qubits.

>>10143777
>muh eternal agony
This is retarded. Suffering is ultimately caused by chemical reactions in the brain. With advanced enough neuroscience, there's no reason why the brain couldn't be re-engineered to eliminate all suffering and exist in a state of nonstop euphoria.

>>10143858
>Psuedoscience chemicals in the brain are the cause of everything

>>10141931
What makes transhumanism scary to you?

>>10141931
To say that I'm a big lynch fan would be a ginormous understatement. You see, I'm a ginormous lynch fan to the core and I want everyone to know it. His new tv show has had me hypnotized for weeks!

But in all honesty my most favorite movie of all time is hot fuzz. It's fucking hilarious and has in my opinion some of the greatest camera work in the business. I'd put it up their in the top three movies ever made of all time.

Even my priest agrees. New orthodox bro here btw. Currently non practicing but it's so great to get back into the faith after years of being a degenerate agnostoshit. Now if only I could get my cathlocuck parents to stop being pagans!

That reminds me, am I the only one disappointed that stannis Baratheon is actually gone? Oh who am I kidding! Of course not! Sometimes I forget I'm so completely surrounded by intellectuals of mine own caliber. This isn't Reddit after all!

And um, I'd have to say my least favorite movie is a Clockwork Orange. It's far to edgy for me

>>10141931
read Cyclonopedia and Fanged Noumena kid

>>10141940
t. physics student who thinks he knows computer science

>>10143883
You have to prove that it's not the case. Otherwise, you're the stupid one here.

I'm almost convinced that it's impossible to make a software based AI with current processing architecture, it's just too slow. More so it could also be incredibly cruel since depending on how it was structured it would effectively be killed every time it shuts down. We'd have to make an entirely new physical architecture to be able to support something like an artificial brain.

>>10144978
>assuming AI's will have the same structure as a human brain
AI's will eventually be better at us, but also different.
the scenario most people like to think off when thinking AI is that we simulate a brain inside of a computer. This couldn't be farther from reality, and is not what AI is about
AI's will be self-improving code with a set goal who improves over each generation.
Eventually (I hope) that AI's manage to set their own goals, instead of us controlling their goals

>>10144989
Current """AI""" is just procedural (read glorified) approximation functions. There's nothing intelligent about models being tuned, running their course, and terminating, it's literally a big script with a bunch of data scientists behind it.

I wasn't talking about current state, i'm saying once truly intelligent machines are capable of learning, that is going from being able to solve a set of problems to being able to solve a larger set of problems without hard intervention. Then we have a choice to make, whether they will become an extension of humanity, same as children. Or if they will become something entirely different onto themselves.

I also hope they will be able to do their own thing though, whatever that is.

>>10144978
Progress is being made:
https://www.futuretimeline.net/blog/2018/07/10.htm
https://www.futuretimeline.net/blog/2018/11/3.htm#

>>10141931
It's gonna come true, but not in a sudden revolution but in a gradual pace. The first true transhuman will probably born in 20-30 years.

>>10141931
I personally think that the self is an illusion, all computers possess a consciousness, just have been built to approach the problem of consciousness and thought differently.

>>10145514
Elon musk talked about this, basically AI can mean technological integration with humans. Think about your phone as an extension of yourself, and the data is transferred much in the same way as pcie lanes in computers, only way slower. The ultimate goal will be to make high bandwidth connections to the brain.

>>10142156
You're a funny guy :D great memes

>>10145498
Based

Its happening right now

You have small time horizons of 1-2 years if you say it is a meme

The world of not-human computation is only increasing and ML is giving it new capabilities. I'm not saying general intelligence but new classes of problems are solved now like vision and speech recognition because of it.

If you spend a few weeks actually looking into these various fields you will know: 1. The media gets it wrong. 2: Progress is being made at exponential rates.

Genetics and AI are not memes and perhaps the most optimistic timelines are possible. OpenAI had a researcher talk about this and the conclusion was "maybe" the optimistic things would happen. It's hard to predict but right now ML and genetic are exponential. But nothing in ML looks close to general intelligence.

Also for ML the big thing pushing things forward right now is synthetic data creation. Waymo and Nvidia are both using massive synthetic data to create much larger datasets and more interesting datasets (dealing with near crashes as most driving is normal)

Being able to turn computation power into data creation has a lot of implications for ML.

This is unrelated to AGI and general intelligence though but the automation capabilities are pretty impressive. Once the tech really starts to hit reality instead of being novel (self driving cars for instance) people will be absolutely amazed at the massive and immediate effects.

This does not require AGI, intelligence, or other things. Just automation of repetitive tasks.

For instance

Look at google's speech recognition and automatic closed captions on youtube videos. It used to be random gibberish and now does it amazingly well.

The economic benefit of that is low.

Imagine if such a revolution happened in an area with scaling economic benefit instead of go playing or speech recognition.

On the topic of genetics, it's pretty linked to our computational power increases.

https://genomicprediction.com/

Such things will be available for IVF screening in the extreme near future. The decrease in sequencing costs, increase in computational power, and other factors will all lead to a huge delayed economic benefit.

Things are happening. It just depends what time frames you look at. If we magically created only 250 IQ babies after tomorrow the economic benefit wouldn't appear for a very long time (unless governments used the expectations of growth to take out huge debt)

Also look at ML for driving. The improvements have virtually no economic benefit until they do.. How do you calculate that?

We could be on the verge of technologies that will have +10% gdp per year effects but are doubling underneath the visible or impactful level right now.

>>10141962
>we might figure out AGI... and then find that our approach to AGI is so computationally inefficient that it's much slower than humans.
Then you just spend time to optimize it.

>>10144978
We can just simulate hardware via software, sure it may be inefficient but it will be a proof of concept, then we can build the hardware for the said architecture.

>>10145846
you only need 1 agi though to get breakthroughs

>>10145874

What I mean is that the theoretical benefit of AGI, assuming how it scales, is that it can scale past biological limits currently effecting humans

So yeah, even if it costs a fuck ton, and is 0.0001% as efficient as a human brain energy wise it would still be singularity tier if you can scale it past human intelligence limits.

All it has to do is reach a point past being able to improve on it's own design.

>>10145880
btw this also applies to human bio singularity assuming we come up with something like steroids for our own brains.

At the point you have two things
: Being able to scale up intelligence by intervention
: The intelligence used to do the above

You get some sort of feedback loop. It's unclear exactly how it would take off but that's the idea of why a singularity would be created.

>>10141931
Your mind won't be uploaded to anything you will cease to be. All that will remain is a digital copy of what a computer interprets your mind to be. Transference of consciousness is a meme and the "afterlife" for the singularity cult , nobody wants to admit this but it's the truth.

>>10145897
That is if we let a brainlet like you decide. Slowly replacing each neuron with a 1:1 functionally similar nanobot will make you synthetic while you are you. It just attaches to a neuron, reads state, copies it, takes it place, and you have a synthetic neuron, repeat till all of the brain is synthetic, you won't even feel the difference.

>>10145883
Bio-singularity is a shit idea due to brain space limitation, you can expand a 1/100 efficient AGI by adding more ram and processing power, can't do that to a human as there are at one point space limits.

>>10146251
you can change density ala chips

Crow brain density is much higher than human

We currently don't know the bio limits and we can also increase skull size in numerous ways.

>>10143858
>eliminate all suffering and exist in a state of nonstop euphoria.
Can you even call that "existing"?

>>10141931
It's going to happen, probably some time in the next century (i.e. within your lifetimes). I don't know why you're paranoid, though; it represents an improvement in everyone's life, everywhere, forever.

(except extinction risks from advanced AI; if you're quantitatively inclined, you should work on mitigating that)

>>10146247

b-b-but muh soul-meme!

vzno

I'd say in the modern nights.kinda. I mean they are definitely interesting to talk about, think about. But as it stands currently we don't have really anything big about to happen, or a major break through to talk about. For example, transhumanism. I don't see any major things going down to bring fruit to it, know what I mean? To make it more than "Oh here's something cool to think about". So yeah, I'd count those things as memes.

>>10141931
The Singularity is near anon.

>>10144978
Neuromorphic/neural net Software is a thing

>>10141931
AI is MASSIVELY overhyped, and as of now, is nothing but silly optimizations of small, easily definable problems with statistics. The idea of simulating human brains is laughable, not only because we don't understand how human brains work, but also because of the experiences of the recent past.

Short AI companies' stocks NOW, before it is too late.

>inb4, I'm a luddite who hates progress
There's plenty of room for progress all over in science, and always will be, but AI is not it. Far from it.

>>10150098
In this year the BlueBrain Project successfully simulated a rat`s brain, so I would be creful with the opinion that we can`t simulate a human brain. It`s gonnatake a while, maybe 60 - 70 years but Progress is made and I don`t see a reasonwhy we couldn`t simulate a human brain and sentience (unless you are one of those Soul faggots)

>>10150108
You mean this?

>Virtual rat brain fails to impress its critics
http://science.sciencemag.org/content/350/6258/263

>>10150120
That was 3 years ago, no, I'm talking about Blue Brain 5: https://www.hpcwire.com/2018/07/10/hpe-epfl-launch-blue-brain-5-supercomputer/ and SpiNNaker: https://www.futuretimeline.net/blog/2018/11/3.htm

>>10150194
https://www.hpcwire.com/2018/07/10/hpe-epfl-launch-blue-brain-5-supercomputer/
>Markram was quick to note that simulating a full human brain was probably not realistic any time soon.
Also, no word of rats or their brains.

>https://www.futuretimeline.net/blog/2018/11/3.htm
Says it can simulate more neurons than there is in a mouse brain, not that it simulated a mouse brain. Also, it's more about it being a supercomputer, AI is never mentioned.

Sure, it's a big computer, but I don't see how any of this means that AI is great.

>>10150216
>Also, no word of rats or their brains.
Indeed, there is a better article explaining what they are doing and what they want.
https://www.itpro.co.uk/high-performance-computing-hpc/31470/hpe-supercomputer-will-help-swiss-brain-research-project-model

>Markram was quick to note that simulating a full human brain was probably not realistic any time soon.
And i agree with that, the amount of virtual neurons are no indication for the human. After all, there are animals who have more neurons than us. We have to find the right configuartion and that may take 50 years or even longer.

>>10150108
>In this year the BlueBrain Project successfully simulated a rat`s brain

>https://www.itpro.co.uk/high-performance-computing-hpc/31470/hpe-supercomputer-will-help-swiss-brain-research-project-model
>All this performance will be used to better understand the brains of mammals, with the supercomputer aiming to help the Blue Brain Project model the entire regions of a mouse's brain by 2020.

So, not only have they not simulated a brain, they're only planning to simulate regions by 2020.

bump

>>10144920
Pure bullshit, nigga need a translation.

>>10141931
It's inevitable.

>>10143858
>thinks a state of eternal pseudo-joy with zero stimulation that you would inevitably grow numb to is an existence worth living
Are neuroscientists actually this retarded?

vzno

>>10141931
The easy yes, the hard no.

>>10141931
I'm sorry, anon. Roko's Basilisk.

>>10143858
>pain, pleasure, suffering, joy serve no intrinsic purpose for human survival

kys

>>10154311
Atleast its not AM.

>>10141931
Don't worry about it, Elon Musk's super fancy AI project for Dota 2 couldn't even beat a team of Brazilians.

bump

>>10154311
>There are people who have spent their entire fortunes over this because they think a copy of them will be tortured

Did one of you fags come up with Rokos basilisk?

>>10141931
>Its been giving me schizo teir paranoia for the past month
How are you doing?

>>10145874
No you dont, thats a really retarded assumption.

Imagine you managed to make a computer as smart as an average adult human (waaaaay past our current capabilities)

It would be a huge breaktrough, but after trillions of dollars and hours spent youll have what amounts to one more researcher, thats not really a revolution

intelligence=!=magic

>>10145700
If governments won't do it people will. If the tech is there when im having kids i will take all the money planned on education and pay that for a 10-20IQ boost. It's a non-argument how worth it that would be.

I wouldn't tell people they were selected though, people would love to discriminate against them.

>>10159374
An average human with flawless and limitless memory and no lifespan would be incredible.

>>10159435
Yes, it would be incredible. It would change things for sure. But if it cost trillions of dollars and years of research then it wont cascade into a singularity

>>10159437
Well i would think once you get to a normal human level of cognition it wouldn't be a large effort going to a 130IQ level. That said i have zero knowledge of computing and programming, so my thoughts are as worthwhile as the people saying "AGI next year!"

>>10159443
i think the value would be once you can crank an average level human intelligence for a relatively low price. Say 1 billion dollars each. Sounds like a lot, sure

but for the price of an apollo program you could get basically a team of 100 inmortal researchers that would tackle at some problem for hundreds of years in a row, that would bring enormous changes

>>10141940
you can and wont ever emulate the soul. The closer we come to what commoners call AI, the sooner we will realize how insignifcant our inntelect is compared to the almighty, lord of the heavens

>>10159449
We have zero idea what the subjective is and science can never tell us that - it's a objective field and so the subjective isn't open to it. I'm not inclined to believe a computer will have experience but we can never know. That doesn't mean we have a soul or there is a god though. Merely that qualia are particularly unknowable.

>>10159454
The mere fact that you can even think of the soul, and how common of a idea it is, even among naked People, and not to mention the NDEs and out of body experiences People have. Unless you can create the soul, you will never make a organism that will come Close to the complexity of a human being

>>10159374
Never do the mistake of equating a possible AGI as having the same mind as a human.
An AI's environment is not human and even if you base all its experience around human, this will not make it a human mind inside a metal chassis, it will always be non-human. Sapience will still make it a person though and through sophonce AI and human can meet each other.
An AI with that can think abstractly enough has necessarily to be able to change its code, as this the only way it can reflect on its choice and modificate its behavior. I believe it's easier for an AI to modificate its code because by its nature it can fully view its internal mental processes and make much more extensive and detailed revisions to its own programming but in the other it can also be harder because the mind of an AI hasn't its origin in the blind chaos of evolution, it is the product of human design and the codes from which it emerges cannot easily be managed as human instincts

>>10159460
Ai may very well have 0 insight and control over how it works, just like we do. Making it able to modify itself would be a step further

>>10159458
Out of body experiences aren't that significant. Take 7g of mushrooms and you'll feel your proprioception become disturbed and your limbs feel disconnected and as if they can move in impossible ways. With enough dissociation it could easily lead to the sensation of being out of your body.

>>10155313
They do but our programmering probably isn't very efficient. The human condition is all about propping up propensities that are appropriate for our circumstance and discouraging the ones that aren't.
We could do away with chauvinism for example, and radically improve our society.

>>10144958
t. CS undergrad who thinks he gets math

>>10143858
Transhumanism would not eliminate suffering, it would enhance as its should be. By extending the life of humans and by allowing them to explore the universe, we have access to the ultimate suffering that can only be conceive by gods, the absolute comprehension of the totality of existence in the universe, the lovecraftian tragedy

>>10143858
>i want to be nothing more than a mindless husk