Is there actual AI research going on, or is it all just buzzwords that boil down to applied statistics?

Is there actual AI research going on, or is it all just buzzwords that boil down to applied statistics?

Other urls found in this thread:

newscientist.com/article/mg23331144-500-ai-learns-to-write-its-own-code-by-stealing-from-other-programs/
youtube.com/user/keeroyz
twitter.com/SFWRedditGifs

To elaborate, is there any substantial research being done on actual artificial consciousness? If so, by whom?
Can we expect true AI and related fields to encounter breakthroughs in the near future, or are we still extremely far from that?
What are the biggest obstacles the field is facing?

No-one cares about "Conscious general AI overload" except teenagers and sci-fi authors.

Actual AI research is about constructing algorithms for solving specific problems. Adding machine learning methods (i.e. linear algebra and statistics) to the problem solving is a recent trend. For example, Deep Blue (the famous chess AI) didn't use machine learning, whereas AlphaGo (the famous Go AI) used.

Then could self-modifying AIs capable of writing their own algorithms for specific tasks take us to the next level of research? There's already been research into systems that can write their own source code.

newscientist.com/article/mg23331144-500-ai-learns-to-write-its-own-code-by-stealing-from-other-programs/

>No-one cares about "Conscious general AI overload"
Why?

>artificial consciousness

as long as we don't understand how or why human consciousness happens all endeavors in this field are doomed to fail

Presumably because nobody has the slightest idea how to build one. Science progresses incrementally. It would have been completely unproductive for somebody in the 17th century to try building a cellphone, because they lacked the knowledge to have even the slightest idea where to begin.

Look at the alchemists. They just decided for unscientific reasons that it should be possible to turn lead into gold, and then went about trying to make that happen, and it was a colossal waste of time.

The difference between alchemy and AI is that we at least know that building a machine with human level intelligence is possible, since we've already got an example of such a machine. But we still don't have any idea how it works.

I'd bet that neuroscientists will understand the brain before programmers build a human-level AI. And I'm betting that the day they do, the programmers will be very surprised, because I doubt it'll resemble anything any human would think of just by attacking the problem head-on. But who knows.

Same reason no one is seriously attempting time travel, antigravity, or teleportation

>write optimising software
>turns whole multiverse into paperclips
Oops heheheh

>building a machine with human level intelligence is possible, since we've already got an example of such a machine
Lies.

Fuck off dualist filth.

What are you even implying here?

Biological immortality when? This senescence thing is super uncool.

One day after the catastrophic failure of your vital functions.

There won’t be AI for hundreds of years.

What people think AI today is is just machine learning. There’s no intelligence about it. It has to go through thousands of iterations to train the neural network to be able to identify the correct solution. Given a problem it never faced it would be clueless how to act.

The only people who think "artificial consciousness" is a problem are people who cannot even provide a definition of consciousness.

>hundreds of years.
Aren't you exaggerating?

There is no machine with human level intelligence. There are machines that can perform very narrowly defined tasks better than humans, but there is no machine with anything resembling the generalized competence of human intelligence.

Honestly I don’t think so. The brain is probably millions of orders of magnitude more complicated than computer chip.

>There’s no intelligence about it. It has to go through thousands of iterations to train the neural network to be able to identify the correct solution.

But human children, by the earliest estimates, acquire object permanence after being alive for 3 months. If a baby is awake for half of its life, that's ~3.5 million seconds of data.

It takes more than 7 months for humans to be able to crawl and actually maneuver physically in the environment. That's ~8 million seconds of data.

How much data has been pumped through the human brain before it begins to solve problems that require intelligence?

Do the same analogy but with an adult human brain, with all structures fully developed but with 0 data of information, then let it live and experience world.

Well for most tasks that are dependent on pattern recognition the machines will probably be better than humans.

That’s not what we generally view as acting intelligently.

Intelligence would be the ability to take those patterns, abstract them and apply them to other problems without ever needing to be trained for the other problem.

Neural nets might produce something similar to AI when they have the ability to interact with each other I’m just not sure we have the technology to replicate the processing ability and efficiency of the brain.

>the technology to replicate the processing ability and efficiency of the brain.
So the biggest problem comes down to hardware?

I would think it’s a major limitation at this point.

There are machines with human level intelligence. They're called humans.

Explain why you believe what your brain does is beyond the scope of statistics.

There is also a gap in theory. There is no math that elegantly describes recurrent non-linear systems with high input sensitivity.

Understanding of recurrent neural networks isn't in a good place right now either.

>Is there actual AI research going on
Certainly.

>or is it all just buzzwords that boil down to applied statistics?
While I see what you mean, you should not be so quick to dismiss applied statistics. Reducing problems to applied math is how most computing works, and it would not surprise me one bit if the human brain can be entirely summarized to "applied statistics". That's what intelligence in its most fundamental form is, after all.

How did those figures come up?

I was talking about the human brain, chumley

Applied statistics will never reach the same level as a human brain unless we manage to improve computational power by extreme amounts.

Citation needed. Why not?

>Applied statistics will never reach the same level as a human brain unless we manage to improve computational power by extreme amounts.

NO... If I make my dog 1000000 times faster, that does not mean he will ever be smarter be smarter than just a dog. Even if I give my dog the ability to remember everything it has ever experienced.. it is still just a dog.
Dogs ARE self aware and can learn... but they are dogs... they will never create a space ship.

>Dogs ARE self aware

citation needed

Lots of modern AI is focused around making tasks that computers could previously do with lots of human written rules (recognizing letters for example) and finding ways to have a computer learn how to do that. For example if you wanted to recognize letters in the early days you would hard code rules for every letter and it would work like 80% of the time. Now we can use a neural network which assigns a weight to each possible pixel in an n by m grid. This works 99.999% of the time and you just need a labeled dataset to train it. Also, check out this youtube channel, he's always explaining new ML research youtube.com/user/keeroyz

Why? Our brains are simply circuits. CPUs are simply circuits. Unless there's some magical secret sauce in your brain I disagree.

the human brain's advantages over computers are about being able to process more things in parallel and be really energy efficient

And we can't make hardware that does the same why? Literally, it's just carbon vs silicon.

we can't

else my phone would last weeks on the same charge

Why not? There's no size restriction or power restriction in the scenario we're talking about. You were saying that there's a secret sauce to the brain and because of that there are things that computers will never be better at than humanss.

why do you think someone would hide such technology? I'm not trying to imply anything, just pointing out the fact that there are no computers as efficient as the human brain on the market

We don’t completely understand how the brain works to construct a replica in silicon, or even know if it’s possible.

Oh sure. I'm not saying the technology is here, just that it's possible.

Yes but that doesn't mean that any other circuit formed via a genetic algo would be any different.

geez... neuroscientists are such brainlets

Hey OP

Check out Ayasdi and Topological Data Analysis.
They do it at Duke University, and some place in ohio.


Using topology for data science they're BTFOing traditional statistics-based machine learning. Really cool.

>not using quantum elon musk AI theory

it's like you want to be behind forever

Were talking about emergent consciousness not neuronal action potential.

It’s simply not understood.

The guy you responded to was referring to the human brain as an existing machine, not a machine that replicates the human brain you dense motherfucker

Your brain is doing "applied statistics" you edgy autist.

No it's not.

Applied statistics is actual AI research and always has been.

Then what is it doing? Magic?