The meme is real

Just had this little gem land in my inbox. I actually laughed out loud

There is a proverb in my country: don't bring a bag to a famous tree.
It means that you have little to gain from something everyone knows about.

This is a good indication that A.I./deep learning/etc. is a fad.

this.
Also the little fact that we simply do not actually have AI nor are we working on it.
What is called AI today is just pattern matching on steroids, nothing more.
Evident by the fact that the supposed super-AI that beat Jeopardy had every single one of its follow-up project fail.

It's definitely not a fad. That's insane. What is the alternative for tasks like voice recognition, image recognition, face recognition, environment state modelling, and things such as this?

see my comment here its not ai for those tasks, its just heuristics and pattern matching.
People have been raving about the eve of AI since ELIZA, and they are just as wrong today as they were back then.

>people have been ranting and raving about these automobiles, but they seem like a fad to me. they still are just wheels with reins

thats the fallacy: We have automobiles, we do not have AI yet.
With your example it would be as if people had been raving about their awesome automobiles when all they had was horse drawn chariots in ancient Rome.

Your fallacy is that AI is nothing because it's nothing yet. It needs time, it's pretty obviously a serious and important part of the future, if you can't see it you're a moron. I work in AI infrastructure and it keeps evolving constantly

>voice recognition, image recognition, face recognition,
These were already done 10 years ago. We are now in the mass hysteria phase where people who have no idea what they're doing are over-applying a tool (statistical learning) where it has little benefit. Do you remember Alpha Go? Turns out the algo doesn't generalise even to chess. Talk about a dead end.

>algos designed around linear algebra
>expect them to work in domains which linear algebra doesn't work well
This is foolishness.

>It's definitely not a fad.
I bet 10 years from now you'll cling on to the next big thing like a brainlet.

You're a moron, these were not "done" 10 years ago, what the fuck are you talking about. We didn't even have iPhones 10 years ago and webcams were grainy. What universe are you living in?

>We didn't even have iPhones 10 years ago and webcams were grainy.
Because of hardware limitations, cretin.
(Please tell me you're trolling.)

Hardware limitations are software limitations. It's completely clear you have no comprehension regarding this topic so I'll give you a pass. You're just trying to be contrarian.

>says the nitwit who uses an Apple product as an argument
Your shitty phone is not using a single algorithm that wasn't already developed 10 years ago. The only reason it is in your phone right now is due to increase in processing power for sufficiently small architectures.
You're the clueless one. Fuck off.

>Your shitty branch of knowledge called mathematics does not use a single piece of information not derived from derivations made thousands of years ago by Euclid

Cringeworthy post. You really have no idea about anything you're talking about.
Go back to where you belong.

>he thinks all math is derived from Euclidean geometry
How to spot a brainlet in one post.

>Can't even into geometry
Sad, very sad.

Top Kek. Sasuga Appledrone. As if anyone needed even more proof that machine learning is a meme.

>Alpha Go? Turns out the algo doesn't generalize even to chess. Talk about a dead end.

it just beat the worlds best algo for chess (stockfish). why does clueless people comment on AI and infect these threads. Its so strange

>Your fallacy is that AI is nothing because it's nothing yet. It needs time
Yes, assuming we were researching true AI, but no such research is being done right now. Its just about improving our ANN and whatsnot to better match patterns. Media interprets that as AI, wrongly.

> it's pretty obviously a serious and important part of the future
because...?

> if you can't see it you're a moron.
ah, insults. Why yes, those suddenly convince me more than facts, thank you.

> I work in AI infrastructur
clearly you do not. I am sure you also work in the Navy seals and are trained in gorilla warfare.

>and it keeps evolving constantly
true but not in a way that would make it closer to true AI.

what does your brain do, if not match pattern and reason about it. pattern matching + logical inference. Add in some creativity in the mix and you have a "brain" that can do most tasks that human can do. If this is not AI then idk what is. do you expect it to be sentient?

True AI is an arbitrary concept. What would something require in order to be AI, qualia? Is human processing really all that different from pattern-matching on steroids?

>>voice recognition, image recognition, face recognition,
>These were already done 10 years ago.

LOL what the actual fuck are you talking about. Do you live in a parallel universe?

>and reason about it
thats the keyword here. "AI" today cannot do this at all.

> Add in some creativity in the mix
thats something that no one knows how to even begin to try with "AI". At least for the "reasoning about" part we have some theoretical, very limited, models from the AI research of the 80s.

>If this is not AI then idk what is. do you expect it to be sentient?
That would be indeed AI, or at least the beginning of AI. But unfortunately this is not being researched currently. The whole field of contemporary AI with its Deep learning, machine learning and unsupervised learning is exclusively dedicated to creating better pattern matching algorithms. it is not dedicated at all to being able to reason about those patterns, much less to be able to be creative.

>True AI is an arbitrary concept.
No on the contrary it is very clearly and precisely defined in the scientific literature.

>What would something require in order to be AI, qualia? Is human processing really all that different from pattern-matching on steroids?
Yes human processing is fundamentally from pattern matching. Granted it does that too and much better than most algorithms today even. But its only the basis, the lower layer of human intelligence. Humans can understand those pattern, they can reason about them, they can invent new ones.
All this is, even for Natural Intelligence, not quite well understood as every neuroscientist will assure you. So how the hell could we ever attempt to teach what we do not understand to machines? To have real AI you first have to understand NI.

We had, in fact we had them much, much earlier. One of the earlier space probes included an image recocnition program for one of the moons of Saturn IIRC. There were 4kb space left to write a photography-trigger program. The scientists managed to write one in 2kb and used the remaining space to write a image recognition program.

Most of CS is really ancient, several decades old.

Yikes. Should've at least said cantor's naive set theory, at least then you wouldn't look like a total brainlet

Things like this is why I think we should attack the biological angle first, we have some loci and alleles directly associated with intelligence, flip those on, damn the consequences, and then use AI to augment what we can already do, with direct enhancement, THEN maybe we'll have full "ai".

Vaguely related, I see our minds as inferior hardware with superb code, we use our limited capacity amazingly well, but it's just that, limited.

>very concisely defined
If you just search it on Google right now you'll get multiple conflicting definitions with varying degrees of scope.
The thing is it might be possible to understand intelligence mathematically rather than just biologically.

Why? Because someone else has a 300k job and you don't?

Im sorry, what exactly about 100 billion neurons, each connecting with up to 10000 other neurons at the same time, each equiped with not just binary (forward, dont forward) operations for incoming signals sounds "primitive" to you?

All that using less energy than you average Desktop CPU and being able to withstand tremendous damage up to and including losing one half of its mass or being split in two.

>If you just search it on Google
Yeah and if you search it in actual scientific literature....

>The thing is it might be possible to understand intelligence mathematically rather than just biologically.
That would be very interesting. I however doubt that it is possible. No mathematical research is being done in that field.

The non existence of true ai doesn't make the study of ai a meme. On the contrary, trying to find a general intelligence is largely useless. The money is in the applications of lesser AI to products. Only about a year or two into this ML boom and we're seeing massive improvements to program functionality

>thats the keyword here. "AI" today cannot do this at all.

reasoning is just logical inference. the hard part was to extract useful features from all the complex signals and patterns, sound, vision, text, speech. you can use logical inference on this. there are also of course more sophisticated ideas working on now. why would you assume there are no research papers on this. it is. You can also look into Reinforcement learning, which is a whole different approach, instead of reason about things, learn behavior through reward and error. just like animals

Creativity, look into GAN , generative adversarial networks. By feeding in data about some concept. the network learn the probability distribution of that concept. Now it can create novel instances of this concept. One huge step towards creativity and imagination. One example is create new pictures that don't exist based on some category, dog, bedroom, computer.

Sentient AI are philosophy its not serious matter and you are not approaching this subject correctly if you even think about that stuff. AI is artificial no real intelligence. its about being able to make machines do human tasks. so you can outsource more human activities to machines. its a functional goal. And there is no scientific way to measure sentient. so you will always be disappointed there.

>yeah and if you search it in actual scientific literature
You tally missed the point. The scientific literature or formulations of true ai can be found with a simple google search and are opposed to each other. Where do you think I was getting threse formulations of true AI? Gizmodo?

No mathematical research being done? I highly doubt that, considering computer science is basically 99% math in and of itself

More than a fad what most people dont get is that the innovation required to advance AI is in hardware, we need more computational powe4, not software where we been using things created decades ago

I lol'd just from the company names
>SALT
>NIPS

So? Do trend-chasing brainlets seriously think a board game player is akin to GENERAL intelligence?

>The non existence of true ai doesn't make the study of ai a meme. On the contrary, trying to find a general intelligence is largely useless
agree

>Only about a year or two into this ML boom and we're seeing massive improvements to program functionality
completely disagree. Programs are mostly just as crappy as they were before.
Take Quora, renowed for their "fantastic" use of AI.
Apparently because I am interested in ASICs (applications specific integrated cirquits) I should also be Interested in Aiscs (sposts shoes), this is not at all AI, this is just a cheap and simple word matching/find similar algorithm.

>Sentient AI are philosophy its not serious matter and you are not approaching this subject correctly if you even think about that stuff. AI is artificial no real intelligence.
My point exactly. And all those things being done under the umbrella of AI are not really AI but simply pattern matching. The things you describe: reasoning about stuff is as old as APL and Prolog and no great improvements have been made.
Meanwhile in reality, most ANN etc. are really just most primitive pattern matching, nothing more. It is not at all revolutionary, just a bit of improvement (sometimes not even that) over the current situation.

>reasoning is just logical inference
its much more than that.

>Creativity, look into GAN , generative adversarial networks
Not creativity at all, just simple evolutionary/genetic algortihms they run into dead ends very soon outside of lab conditions and human interference.

>considering computer science is basically 99% math in and of itself
no, not even close to that.
CS has become very different from most of math except the parts about logic and some parts of cryptography

These individual neurons lack processing power. It's intricately and wonderfully designed, and efficient as hell.

It is this primitive masterpiece that's allowed us to create the resource and energy surplus that we now enjoy. Lets take this bounty nature has given us and refine it.

We evolved in a climate where we were one disaster from death, one bad week groom starvation until the past few decades.

Energy consumption is of very little concern to us, *now*. Overclock this bitch see what it can do.

>We evolved
AHAHAHAHAHA

Please, at least pretend to know something about reality. It would make other people much more inclined to listen to you.

Whats next? Astrology, homeopathy, crystal healing?

you just don't know about this stuff, most of what you are saying is technically wrong, poor understanding of the field. You are more into philosophy than technology

>you just don't know about this stuff, most of what you are saying is technically wrong, poor understanding of the field
you are confusing me with you.
Everything I say is 100% backed by scientific research, contrary to you who seem to back his claims mostly by popsci and "science" articles in the mainstream media.