Google scraps AI project with Pentagon after employees revolt

[Sauron]

11 minutes ago, jde3 said:

Sure man.. Machines don't learn at least not like we do.

That's why there's a whole field of research called "machine learning" dedicated to devise ways to make them learn in a way that suits a machine. The basic concept is actually fairly simple, I recommend you look it up - neural networks "learn" in a pretty straight forward way, and it is proven to work. It's not my field of expertise but, for what it's worth, as a software engineer I believe it's perfectly legitimate.

11 minutes ago, jde3 said:

They basically follow instructions and the "algos" we all think are so great are pretty busted to be honest. It's a trend and a marketing term, nothing more.

The main difference between what we refer to as "AI" today and regular algorythms is that an algorythm is fixed and coded manually by the machine's programmer, whereas a neural network actively alters itself to produce a better result. All the programmer has to do is program the neural network with the right parameters (which are what the reserch is mostly about) and the machine will do the rest. Suddenly, you're no longer telling the machine how to make a sandwich - you're giving it the means of learning that by itself. The name itself may be just a marketing term, but the technology is real and it works and it's unlike anything that came before it.

[Jito463]

Frankly, if Google doesn't do this, then someone else will.  The technology has merits, such as faster facial recognition systems to help identify enemy targets.  Enemies of this technology just see this as a means to an ED-209, which is ridiculous.  _{^{Everyone knows the ED-209 series was flawed.}}

16 hours ago, Cole5 said:

We would be better off than we are now

Nope, not gonna bite.

[LAwLz]

21 minutes ago, jde3 said:

Sure man.. Machines don't learn at least not like we do. They basically follow instructions and the "algos" we all think are so great are pretty busted to be honest. It's a trend and a marketing term, nothing more. Skip logic.

Please explain to me how the "algos we all think are great are pretty busted". 

I am also eager to hear why you don't think neural nets are actually learning, and what your definition of learning is. 

[Cole5]

6 minutes ago, Jito463 said:

Frankly, if Google doesn't do this, then someone else will.  The technology has merits, such as faster facial recognition systems to help identify enemy targets.  Enemies of this technology just see this as a means to an ED-209, which is ridiculous.  _{^{Everyone knows the ED-209 series was flawed.}}

Nope, not gonna bite.

 

What? No bait only my honest thoughts. an AI wont get into a dick waving contest with a midget 

[leadeater]

7 minutes ago, Jito463 said:

Everyone knows the ED-209 series was flawed

Well it was 80's technology, basically all technology from that era was flawed lol.

[MadyTehWolfie]

1 hour ago, M.Yurizaki said:

While I'm not sure where this technology would go, keep in mind that a lot of technology you use (in fact, the very one you use to access this website), came from defense industry funding.

 

Our hope is we find a peaceful use for this technology, that's how we evolve as a civilization.

Everything we have started out as millitary technologies. Computers, phones, landlines, all of it.

[jde3]

9 minutes ago, LAwLz said:

Please explain to me how the "algos we all think are great are pretty busted". 

I am also eager to hear why you don't think neural nets are actually learning, and what your definition of learning is. 

Look at what they do. Lets say you buy a product on Amazon. Amazon will determine you like that product and send you ad's for similar products.. however the problem with that is you already have it. You no longer need it again. If you just bought a new TV, you don't need another TV. Take a look at facebook for example. If I have a friend that posts a lot of bullshit I don't care about but occasionally respond to and read and comment on their posts Facebook will determine that I like that person and I want to see their posts more than the people that don't post very often that ARE my really close friends that I DO want to see messages from hiding them lower in my feed. Busted algos, there are many more examples.

[Jito463]

9 minutes ago, MadyTehWolfie said:

Everything we have started out as millitary technologies. Computers, phones, landlines, all of it.

The internet.

[Jito463]

10 minutes ago, Cole5 said:

What? No bait only my honest thoughts.

I assumed you were making a Skynet reference about the end of humanity.

[leadeater]

11 minutes ago, jde3 said:

Look at what they do. Lets say you buy a product on Amazon. Amazon will determine you like that product and send you ad's for similar products.. however the problem with that is you already have it. You no longer need it again. If you just bought a new TV, you don't need another TV. Take a look at facebook for example. If I have a friend that posts a lot of bullshit I don't care about but occasionally respond to and read and comment on their posts Facebook will determine that I like that person and I want to see their posts more than the people that don't post very often that ARE my really close friends that I DO want to see messages from hiding them lower in my feed. Busted algos, there are many more examples.

That's not really a neural network though, that is using a form of machine learning but it doesn't come under the research field of machine learning or AI. Those also don't truly self learn either and always have human intervention and correction.

 

AI is the stuff that learns how to play Chess or Starcraft 2 better than the best humans on earth by playing, over, and over, and over, and over learning from each experience and outcome then self adjusting. A strategy it finds today may not work tomorrow or in 2 years but that's fine it will adjust, that last part is where the human fear of it comes from.

[LAwLz]

9 minutes ago, jde3 said:

Look at what they do. Lets say you buy a product on Amazon. Amazon will determine you like that product and send you ad's for similar products.. however the problem with that is you already have it. You no longer need it again. If you just bought a new TV, you don't need another TV. Take a look at facebook for example. If I have a friend that posts a lot of bullshit I don't care about but occasionally respond to and read and comment on their posts Facebook will determine that I like that person and I want to see their posts more than the people that don't post very often that ARE my really close friends that I DO want to see messages from hiding them lower in my feed. Busted algos, there are many more examples.

I think you are drastically underestimating the algorithms you're referring to there (does Amazon even use machine learning for their suggestions?), mixing up different systems with different knowledge, and you're using those examples to condemn all machine learning without understanding it. 

 

Again, please explain why you don't think for example a neural net is really learning. What is your definition of learning?

[jde3]

2 minutes ago, Jito463 said:

The internet.

And some of the most important stuff.. like Philosophy.. oh wait..

[Sauron]

6 minutes ago, Cole5 said:

What? No bait only my honest thoughts. an AI wont get into a dick waving contest with a midget 

...ever heard of Tay?

http://knowyourmeme.com/memes/sites/tay-ai

3 minutes ago, jde3 said:

Look at what they do. Lets say you buy a product on Amazon. Amazon will determine you like that product and send you ad's for similar products.. however the problem with that is you already have it. You no longer need it again.

Actually amazon will recommend you similar products or products that fit well with what you already bought. That's not AI in the sense of a neural network though, it's just a normal algorythm (at least for now).

6 minutes ago, jde3 said:

Take a look at facebook for example. If I have a friend that posts a lot of bullshit I don't care about but occasionally respond to and read and comment on their posts Facebook will determine that I like that person and I want to see their posts more than the people that don't post very often that ARE my really close friends that I DO want to see messages from hiding them lower in my feed. Busted algos, there are many more examples.

To make any sort of guess on what you'd like to see you need a LOT of data. I'm not sure facebook uses AI specifically for your feed, but if they did it would base itself on data it gathered from ALL users and try to guess based on patterns it recognizes in most of the population. Imagine having a human choose what is in your feed based solely on what you click on and where you live - their prediction would probably not be very accurate either, would it?

[jde3]

1 minute ago, leadeater said:

That's not really a neural network though, that is using a form of machine learning but it doesn't come under the research field of machine learning or AI. Those also don't truly self learn either and always have human intervention and correction.

 

AI is the stuff that learns how to play chess or Starcraft 2 better than the best humans on earth by playing, over, and over, and over, and over learning from each experience and outcome then self adjusting. A strategy it finds today may not work tomorrow or in 2 years but that's fine it will adjust, that last part is where the human fear of it comes from.

No i know they are two different topics. It's fine, you know.. you can think its cool but I just think its a fad and a waste of money. Each their own man.

[Sauron]

13 minutes ago, MadyTehWolfie said:

Everything we have started out as millitary technologies. Computers, phones, landlines, all of it.

Hardly - wartime use and military interest only boosted many of these technologies' widespread adoption, but most of them were born in a research lab with no military aims. Regardless, times have changed since ww2 and with globalization it's corporate interests that guide technological development, not war.

[jde3]

13 minutes ago, Sauron said:

...ever heard of Tay?

http://knowyourmeme.com/memes/sites/tay-ai

Actually amazon will recommend you similar products or products that fit well with what you already bought. That's not AI in the sense of a neural network though, it's just a normal algorythm (at least for now).

To make any sort of guess on what you'd like to see you need a LOT of data. I'm not sure facebook uses AI specifically for your feed, but if they did it would base itself on data it gathered from ALL users and try to guess based on patterns it recognizes in most of the population. Imagine having a human choose what is in your feed based solely on what you click on and where you live - their prediction would probably not be very accurate either, would it?

How about I choose whats in my feed? Or just give me a feed? No.. too hard.

 

And go look at your Amazon.. is it actually based on what you want or just related to what you just bought from them. Or better yet look at Youtube are the channels they give you actually shit you want to watch or just a hodge podge of shit you already watched?

 

Good example here.. If I watch a LTT.. great I like Linus because of his personality and production not so much because of the topics or content.. I've never water cooled anything in my life and I couldn't give one flying F about Jay or those other people out there.. but guess what.. Youtube disagrees with me because it can't really tell what I like about a video. All it can do is add tags to a video and show you videos with similar tags.. hardly ground breaking algos.

[leadeater]

3 minutes ago, jde3 said:

No i know they are two different topics. It's fine, you know.. you can think its cool but I just think its a fad and a waste of money. Each their own man.

It's not a matter of what's cool or if I think it is, it's just that what you said wasn't AI or a neural network so if you're saying they don't work or are waste of time it's kind of good to actually talk about it and not something different, that is crap and annoying but all marketing and advertising is.

[Cole5]

6 minutes ago, Jito463 said:

I assumed you were making a Skynet reference about the end of humanity.

Nope, Im more partial to the Machine or Samaritan 

[leadeater]

4 minutes ago, Sauron said:

Hardly - wartime use and military interest only boosted many of these technologies' widespread adoption, but most of them were born in a research lab with no military aims. Regardless, times have changed since ww2 and with globalization it's corporate interests that guide technological development, not war.

It's always been this

 

And if you go back far enough in time just replace money with resources and land.

[jde3]

2 minutes ago, leadeater said:

It's always been this

 

And if you go back far enough in time just replace money with resources and land.

Ya the military did a great job giving us truly important world changing research like philosophy. Oh wait.. no.. hmm.. maybe everything isn't about killing people and taking their shit..

[Sauron]

3 minutes ago, jde3 said:

How about I choose whats in my feed? Or just give me a feed? No.. too hard.

How would you go about that? If you just want them to show every single thing your contacts post, it would get overwhelming very quickly. On top of that you'd never be exposed to other things you might like and be confined to only seeing what your friends post - maybe you'd prefer that, but it's not good for facebook's business...

8 minutes ago, jde3 said:

And go look at your Amazon.. is it actually based on what you want or just related to what you just bought from them. Or better yet look at Youtube are the channels they give you actually shit you want to watch or just a hodge podge of shit you already watched?

Again, you're confusing AI with run-of-the-mill recommendation algorythms... and you're expecting more from them than you would from a human.

4 minutes ago, leadeater said:

It's always been this

 

And if you go back far enough in time just replace money with resources and land.

Yeah, sure - although I suppose in a world war the military effort takes precedence...

[laminutederire]

48 minutes ago, jde3 said:

Sure man.. Machines don't learn at least not like we do. They basically follow instructions and the "algos" we all think are so great are pretty busted to be honest. It's a trend and a marketing term, nothing more. Skip logic.

Well the good thing is that it needs at least 10 000 images of cats to recognize those, so... it'll need 10 000 examples of war won in the modern era to do the job as well. So they're going to have troubles

There however is something very dangerous about demonizing AIs of today. As you said, those aren't as good as most people (especially people who design them) think they are, which will result in AI making costly mistakes for humanity. On top of that you add that we basically have no guarantees on their outputs, so you'll end up fast with something which will shoot on the civilians and so on pretty fast.

[jde3]

5 minutes ago, Sauron said:

How would you go about that? If you just want them to show every single thing your contacts post, it would get overwhelming very quickly. On top of that you'd never be exposed to other things you might like and be confined to only seeing what your friends post

Oh the horror.. How would I ever get by.

 

But seriously, isn't that the point of Facebook?

 

I think that the reason these companies have gotten so big is because they have sold a bag of goods to the advertisers and some day they are going to get wise to the crappy algos and all this is going to change and they will go back to older methods like product placement, native ads and the like. In the 90's hits on a website use to sell banner ads.. this is similar but harder to detect as bullshit.

[leadeater]

3 minutes ago, laminutederire said:

Well the good thing is that it needs at least 10 000 images of cats to recognize those, so... it'll need 10 000 examples of war won in the modern era to do the job as well. So they're going to have troubles

There however is something very dangerous about demonizing AIs of today. As you said, those aren't as good as most people (especially people who design them) think they are, which will result in AI making costly mistakes for humanity. On top of that you add that we basically have no guarantees on their outputs, so you'll end up fast with something which will shoot on the civilians and so on pretty fast.

Humans = global warming

Global warming = death of the planet and life on it

Recommendation = kill the humans

 

[Sauron]

1 minute ago, jde3 said:

Oh the horror.. How would I ever get by.

 

But seriously, isn't that the point of Facebook?

 

I think that the reason these companies have gotten so big is because they have sold a bag of goods to the advertisers and some day they are going to get wise to the crappy algos and all this is going to change and they will go back to older methods like product placement, native ads and the like.

You're confusing "is AI a useful technology that works?" with "is Facebook a good service?".