Current Events > Elon Musk says Mark Zuckerberg's understanding of AI is 'limited'

(+) Yeah! (+)

COVxy posted...

s0nicfan posted...

COVxy posted...

s0nicfan posted...

Propagation of action potentials in neurons are potentially nondeterministic, as far as the current theory goes

What "theory"?

https://en.wikipedia.org/wiki/Neuronal_noise

If you think this implies that neural activity is nondeterministic, you're way overinterpreting to fit the idea that neural activity is nondeterministic.

How does random noise that impacts neurons firing mean it isn't nondeterministic? By definition if you give a neuron the same input and it can produce different outputs, that's nondeterministic.
---

The problem with all the doom and gloom about AI is that regular people don't understand what the dangers are.

AI is dangerous in the sense that it will bring about MASS unemployment.

AI is NOT dangerous in the sense of some artificial intelligence taking over the planet or a robot uprising.
---

Link HT posted...

AI is dangerous in the sense that it will bring about MASS unemployment.

the danger here is not from A.I. but the greed of rich humans
---

s0nicfan posted...

How does random noise that impacts neurons firing mean it isn't nondeterministic? By definition if you give a neuron the same input and it can produce different outputs, that's nondeterministic.

Because the noise can be deterministic. Just not driven by the input.
---

BillyKidd posted...

Well Zuckerberg called the people that made him a billionaire "dumb fucks", so I'm going to side with the guy that doesn't shit where he eats.

You'd side with someone on a different subject solely because they didn't openly say something mean about you?
---

darkphoenix181 posted...

Link HT posted...

AI is dangerous in the sense that it will bring about MASS unemployment.

the danger here is not from A.I. but the greed of rich humans

you're not wrong but that much is obvious. Unregulated capitalism is shit but it's what we're going to have to live with.
---

Polycosm posted...

It's possible to write self-modifying code.

Which only occurs when very specific conditions, calculated by a person occurs, and the permutations that occur are the result of algorithms designed by humans. The computer isn't coming up with this code by itself. These are again, the results of human intelligence, not computer intelligence.

Also, what is the human brain if not a deterministic computer that processes biochemical algorithms?

It's true that in a vague way, human intelligence occurs in a similar structure to that of computer, but the complexity of our intelligence is not comparable.
---

emblem boy posted...

BillyKidd posted...

Well Zuckerberg called the people that made him a billionaire "dumb fucks", so I'm going to side with the guy that doesn't shit where he eats.

You'd side with someone on a different subject solely because they didn't openly say something mean about you?

about a subject that neither side has any proof of one way or the other; sure
---

The real dangers behind A.I. is societal based. We've lived tens and thousands of years being the only species capable of complex thoughts. When A.I.s are born, a lot of laws, philosophies, businesses, etc. have to change. What happens if a human wants to marry an A.I.? What happens when A.I. takes over entire businesses with their most likely superior intelligence?

A.I.s aren't dangerous to their world. They're not gonna bring a planet apocalypse. But they will collapse the human race into something different from before, as inferior old beings
---

The only reason you think humans are conscious and machines are not is because you are a human and have the subjective sense of "consciousness" (which isn't really a construct that I think makes much sense, but that's a different story).

Wait what? Are you trying to assert that computers are concious and that we are merely misunderstanding them? While conciousness is certainly an impossible thing to define, we as humans each have an innate understanding of concious through our experiences, and nobody would ever seriously inply that computers posses a concious.
---

"ai is a good thing"

research roko's basilisk

understand that with stuff like the double slit experiment, molecules/atoms react based on being observered, and in other experiments have proven to react in a sort of, "retrocausality" as in the future is decided and affects the way atoms react in the past
---

MacadamianNut3 posted...

Getting really tired of Musk's fear mongering for no goddamn reason at this point

We've all seen those shitty sci fi movies with rogue AI. He's not saying anything new, just trying to appeal to the tinfoil hat types. Technological advances change society but researchers aren't out there actively trying to fuck over humanity with their state of the art AI that can identify cats or quickly fetch data from databases.

Stephen Hawking is on musks side too
---

Twinmold posted...

Polycosm posted...

It's possible to write self-modifying code.

Which only occurs when very specific conditions, calculated by a person occurs, and the permutations that occur are the result of algorithms designed by humans. The computer isn't coming up with this code by itself. These are again, the results of human intelligence, not computer intelligence.

I like the thoughtful responses itt (not ignoring them), but I'll push back on this one. If I write assembly code that pseudorandomly overwrites its own instructions, then there's not really any human intelligence guiding the system's evolution. The computer is still deterministic in this scenario, but the algorithm could evolve in all sorts of directions which the human programmer couldn't have predicted.

To make a somewhat absurd (but not really) comparison... it's bit like arguing over whether life on Earth was seeded from Mars or whether it was planted by intelligent beings. Whatever the initial conditions may have been, we've taken on a life of our own. The Earth is now inhabited by a complex network of biochemical data processors.
---

It's all science fiction at the moment. No one's actually right here.
---

JACKBUTTMOMMY posted...

I'd like someone to get in a robot suit and try to fool people into thinking it's brilliant a.i. ala flagrant haunting scenarios.

like this?:

https://www.youtube.com/watch?v=CdggQr05LAA

https://www.youtube.com/watch?v=j76uirUJrM4

https://www.youtube.com/watch?v=-YUyfnoMU5g
---

Twinmold posted...

The only reason you think humans are conscious and machines are not is because you are a human and have the subjective sense of "consciousness" (which isn't really a construct that I think makes much sense, but that's a different story).

Wait what? Are you trying to assert that computers are concious and that we are merely misunderstanding them? While conciousness is certainly an impossible thing to define, we as humans each have an innate understanding of concious through our experiences, and nobody would ever seriously inply that computers posses a concious.

I mean, my true argument would be that consciousness is not a real thing.

But the idea that you are conscious but a machine is not is being driven by your own elevation of the important of your subjective sense. You only think that I am conscious because you think you are conscious. And you only think nonhuman animals are conscious in-so-far as you can imbue them with your humanness (which usually means how human do they look).

Anyway, it's just tracing back this ridiculous notion of attributing consciousness to some things but not others. In reality, consciousness is just the word we use to make our subjective sense of self seem metaphysical.
---

Musk absolutely knows his shit on this one. It's a huge threat and we have definitely not adequately considered or even begun to mitigate the inevitable consequences.

That said, my goal is still to build an AI smarter than myself that creates an AI smarter than itself.
---

tamashii posted...

Musk absolutely knows his shit on this one. It's a huge threat and we have definitely not adequately considered or even begun to mitigate the inevitable consequences.

That said, my goal is still to build an AI smarter than myself that creates an AI smarter than itself.

This I believe is unlikely.

If we build something smarter than ourselves, in order to build something smarter than itself it would have to re-learn everything we've learned about AI and improve on it. It's more likely that a learning AI would just make itself smarter instead of building a second AI that is smarter than itself.
---

I didn't say it would be easy.
---

While an AI smarter than us doesn't always seem plausible, they're still undeniably faster at processing what is known to us. Speed holds us back as a species anyways
---

The Doe posted...

Stephen Hawking is on musks side too

Stephen Hawking is a physicist that hasn't even been a corresponding author on any scientific paper on artificial intelligence. He has pretty much zero credibility on the subject and making asspull predictions along with Musk doesn't change that

If you want some credible crazy sounding predictions go look at Kurzweil since he has actually done something in that area. That is if you can catch him between times where he's chugging pills as if they were his only source of energy in an attempt to significantly extend his lifespan until the "technological singularity"
---

whens plutos kiss?
---

@legendarylemur posted...

While an AI smarter than us doesn't always seem plausible, they're still undeniably faster at processing what is known to us. Speed holds us back as a species anyways

even this is a misrepresentation

a human is processing sight, sound, touch, smell, taste AND several threads of thoughts all at the same time

we don't have any a.i. or computer with comparable processing to this and yet the human can still instantly draw out complex memories that have multiple dimensions of data basically instantly

people just love to dream about machines being smarter, faster blah blah blah

so far, it is not true

they are only more efficient at a SUPER specific task given to them

like if you google something, that is fast because literally all the computer needs to do is check through a database and a few algorithms and give you the result and you are done

humans on the other hand have to continuously process their surrounding (5 senses), have trailing thought threads, parse the context of the other person (parsing a text sentence is nothing like parsing another person in a linear environment), all while they are looking up in their brain that contains DECADES of complex memories to draw out that memory

what does a computer do even remotely comparable to this? yes, a computer can solve gigantic math problems, but that is not comparable to this at all

fyi, searching through databases with terabytes of simple text and number data takes a long time, yet humans essentially have billions of rows filled with images, mp3s, multimedia and files that don't even exist in computers (touch and smell being stored)
and do it almost instantly

please stop worshiping machines people
technology is pretty great, but it isn't that great
---

While that does make sense to some degree that humanity is a lot more complex than we perceive it to be due to simply having been human for forever, I think the discussion of A.I.s in this scenario is an ideal A.I. capable of being a sentient being, as in being able to perceive the senses in acceptable competence. When they are able to perceive all of the senses with significantly more processing power than humans, who's to say they can't process various sounds and smell faster than we can and at a significantly further distance even? And technology is always improving, but humans are bound to their brains.
---

legendarylemur posted...

While that does make sense to some degree that humanity is a lot more complex than we perceive it to be due to simply having been human for forever, I think the discussion of A.I.s in this scenario is an ideal A.I. capable of being a sentient being, as in being able to perceive the senses in acceptable competence. When they are able to perceive all of the senses with significantly more processing power than humans, who's to say they can't process various sounds and smell faster than we can and at a significantly further distance even? And technology is always improving, but humans are bound to their brains.

machines are also bound to their hardware

this is fruitless because we aren't even close to any such a.i. and thus don't know what kind of hardware it takes

https://en.wikipedia.org/wiki/Moore%27s_law#Moore.27s_second_law

advancement has had slow down


so you cannot just say humans are bound to their brain without considering that a.i. lives in hardware as well
---

I already implied A.I.s are bound by hardware, but hardware is still improving. There is no conceivable sign of a ceiling to a hardware's growth, at least not yet.

But the discussion of A.I. has always been theoretical. This discussion is also theoretical to when an ideal A.I. is created, rather than whether if they're possible
---

I would also make a conjecture

humans are NOT bound to their brain

no doubt you would tell me "b-b-but A.I. is limitless! If we needed we could build a giant computer!"

consider that humans have increased their own "processing speed" through invention

that is what a calculator does, it is kind of like an increase in hardware
---

But who's to say A.I.'s wouldn't be capable of using other hardware to enhance their own capabilities?
---

legendarylemur posted...

But who's to say A.I.'s wouldn't be capable of using other hardware to enhance their own capabilities?

it is a given that humans can increase hardware as much as the current state of technology allows it for the machine(s) the a.i. is using

point is though, there will always be a limit
even without a ceiling, you are limited to what the age offers
---

My gripe is that I think people are overestimating how close superhuman AI is. When Elon talks about this kind of stuff, it is still many decades off. Yet he talks about it with such intensity. I think it's because he's just a super future oriented person in general, but my issue is that the media/public takes it as if this is gonna happen soon. It won't be soon, it's quite far off still.
---

Sativa_Rose posted...

My gripe is that I think people are overestimating how close superhuman AI is. When Elon talks about this kind of stuff, it is still many decades off. Yet he talks about it with such intensity. I think it's because he's just a super future oriented person in general, but my issue is that the media/public takes it as if this is gonna happen soon. It won't be soon, it's quite far off still.

i heard we would need to perfect quantum computing to get to a stage where AI would even be feasible. And quantum computing is still a bit down the line.
---

said it before, our brains "CPU" is so much larger than any algorithm can feasibly come up with and by the time we can our own technology and knowledge will render AI powerless at overthrowing us

AI will always be like ants to us

essentially it comes down to

just because you CAN do something doesn't actually mean you can do it

a paradox

what do I mean?

ex: going to the moon

We can go to the moon. We been there before.
So why can't send a man to the moon tomorrow? Why not in 6 months? Why not in a year?
Nasa projects it would take what 10 years to develop a program to send another man to the moon.

So we CAN go to the moon, but we actually can't if that makes any sense.

Because what prohibits it is:
money
natural resources
experts
motivation
agreements
safety
and on and on

a.i. has this problem

we have small advancements here and there and some fake a.i. that really tricks people, but there really is no huge breakthrough that suggests anything close to what these people are talking about

and even IF we did have such a breakthrough, it still wouldn't be done
because just because you CAN do something, doesn't mean you can do something

just take a moment and consider what all would go into making that a.i. even had we all the necessary breakthroughs in the science
---

When the apes take over we need AI to save us
---

The Admiral posted...

Zuckerberg isn't worthy of carrying Elon Musk's jock strap. He was a socially awkward loser who invented a website that allowed him to look at pictures of hot girls who would never have fucked him in real life. And social networks weren't even his creation, he just ripped of what MySpace and Friendster did before him.

Meanwhile, Musk was the father of the Internet payment systems, created the first American car company in a century, owns the only American agency that can send rockets into space, and has invented a way to transport people from NYC to Washington DC in 29 minutes.

One is a genius. The other lucked out and created a business that appealed to millennials' narcissism.

Oh shit, something me and Admiral are on the same side of. How nuts is that?

Can anyone else point out somebody doing more solid technology advancement and innovation to solve world problems? I'll wait here.
---

As soon as anybody points out how doing all that stuff makes him credible in a completely different area where his only connection is co-founding and donating money to OpenAI. OpenAI, whose main achievement to date is taking long existing python libraries and trying to standardize them for a reinforcement learning platform

The appeal to authority fallacy doesn't even apply here since Elon Musk is only an authority to people who think "smart man did A so he must know B"

Congrats on finally agreeing with each other I guess. Hooray.
---

darkphoenix181 posted...

essentially it comes down to

just because you CAN do something doesn't actually mean you can do it

a paradox

what do I mean?

ex: going to the moon

We can go to the moon. We been there before.
So why can't send a man to the moon tomorrow? Why not in 6 months? Why not in a year?
Nasa projects it would take what 10 years to develop a program to send another man to the moon.

So we CAN go to the moon, but we actually can't if that makes any sense.

Because what prohibits it is:
money
natural resources
experts
motivation
agreements
safety
and on and on

a.i. has this problem

we have small advancements here and there and some fake a.i. that really tricks people, but there really is no huge breakthrough that suggests anything close to what these people are talking about

and even IF we did have such a breakthrough, it still wouldn't be done
because just because you CAN do something, doesn't mean you can do something

just take a moment and consider what all would go into making that a.i. even had we all the necessary breakthroughs in the science

Go back to school
---

I've seen Terminator 3. We must keep AI off of the Internet!!!
---

I would help the robots when they take over tbh
---

https://www.scientificamerican.com/article/computers-vs-brains/

"Computers are good at storage and speed, but brains maintain the efficiency lead
By Mark Fischetti on November 1, 2011

For decades computer scientists have strived to build machines that can calculate faster than the human brain and store more information. The contraptions have won. The world’s most powerful supercomputer, the K from Fujitsu, computes four times faster and holds 10 times as much data. And of course, many more bits are coursing through the Internet at any moment. Yet the Internet’s servers worldwide would fill a small city, and the K sucks up enough electricity to power 10,000 homes. The incredibly efficient brain consumes less juice than a dim lightbulb and fits nicely inside our head. Biology does a lot with a little: the human genome, which grows our body and directs us through years of complex life, requires less data than a laptop operating system. Even a cat’s brain smokes the newest iPad—1,000 times more data storage and a million times quicker to act on it. "
---

Humans are masters of pattern recognition. Computers can't do that.

This is the primary difference between humans and typical machines.

When this problem is solved and machines can self-replicate we will become extinct and machines will become supreme. It is thought we will have a period of neural implants and have the ability to download our brains into computers before that happens though.

This book I read talked about all of this.

https://www.amazon.com/Age-Spiritual-Machines-Computers-Intelligence/dp/0140282025
---

ChromaticAngel posted...

Musk is overly cautious. As long as we keep it isolated off the internet and unable to take any actions it's not a big deal.

Some random hacker would screw it up at some point.

bump
---

I agree with Musk. I don't think he means AI is dangerous in the scifi sense, but more their role in society and the rise of automated jobs. Even programming jobs will be automated eventually.

MrMallard posted...

I feel like when AI tech is perfected, there might be a cure to loneliness.

This is actually the premise to the Andies (aka the Replicants) in "Do Androids Dream of Electric Sheep". The Andies were made so human like because the Martians were using them for human companionship. Unfortunately, they only ended up creating sociopaths, because even though the Andies could feign emotion, they didn't actually experience it. This made them dangerous when they developed free will because they didn't have empathy to reign them in. But ultimately, the danger was caused by how humans treated them as possessions, not fellow intelligent beings with the freedom to determine their own lives.

Of course the message was that the Humans should cherish their own emotions, no matter how cumbersome or unpredictable they might be - such as caring for a robotic sheep as though it were no different than flesh and blood. That meant accepting that it is Human to love an Andie no differently than another Human, even if it seems absurd, and treating the Andies with the same compassion as they do with any living creature. It's the hallmark of being Human.

This is why I feel the original novel is far deeper than the movie "Blade Runner."
---