Skip to content
 
If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Fark)   Is it true that "artificial intelligence" is just a fancy phrase for "computer program"?   (fark.com) divider line
    More: Survey, Artificial intelligence, Cascading Style Sheets, Takuma Sato, The Muppets, Politics, Computer, Main, Holder of multiple patents  
•       •       •

298 clicks; posted to Discussion » and STEM » on 19 Jan 2021 at 10:14 AM (5 weeks ago)   |   Favorite    |   share:  Share on Twitter share via Email Share on Facebook



63 Comments     (+0 »)
 
View Voting Results: Smartest and Funniest


Oldest | « | 1 | 2 | » | Newest | Show all

 
2021-01-19 9:48:40 AM  
No... It is the description of the last four years of the White House.
 
2021-01-19 9:52:42 AM  
Artificial Intelligence just means that it is a technique that doesn't really work yet. Once a technique becomes practical it stops being "AI" and just become just another "computer program".

/Holder of multiple patents on "Artificial Intelligence Methods"
//Quotes!
 
2021-01-19 10:01:07 AM  
A.I. is really overblown.  Deep learning is even a bit of a misnomer.  It's really just very good pattern recognition.

/IT Guy
//Son is student learning robotics @ AI
/// Slashies!
 
2021-01-19 10:02:29 AM  
Think of AI like a flying car.  We know it's coming, and it's only about 10 or 20 years away.
 
2021-01-19 10:14:14 AM  
"AI" is a symptom of the fact that as our society becomes more and more dependent upon computers, the more confusing that shiat seems to get. It also gives us a nice pre-excuse for mass layoff in the future.

Magnets, for example. That's AI. Because I have no other explanation.
 
2021-01-19 10:22:54 AM  
And regular intelligence is just a fancy phrase for meat program.
 
2021-01-19 10:23:58 AM  
No.  You can debug computer programs.
 
2021-01-19 10:25:14 AM  
The best definition I've ever found is that AI is computer code that modifies itself.

Just run through the implications of that in your head.
 
2021-01-19 10:26:20 AM  

leeto2: A.I. is really overblown.  Deep learning is even a bit of a misnomer.  It's really just very good pattern recognition.

/IT Guy
//Son is student learning robotics @ AI
/// Slashies!


We always referred to them as "expert systems" instead of artificial intelligence. They aren't really intelligent, and can't really make inferential guesses outside of a small area. We ever develop anything that can make truly make correct inference based guesses, over a wide area of data, we've probably developed artificial sentience not A.I.
 
2021-01-19 10:27:12 AM  
Human behavior is largely a form of biological "programming." It's just that our base code is hard wired into our brains instead of being burned onto a ROM chip.

The human brain as a computer is very good at "programming" itself. The problem arises when it programs itself with faulty information and has to be "deprogrammed."
 
2021-01-19 10:33:14 AM  
"AI" as we think about it now is the ability for a computer to recognize patterns in input data that it wasn't explicitly programmed to look for and produce some useful output based on finding them in other data sets.

This is why you hear about an AI being "Trained", someone writes a program that is able to recognize, classify, and discriminate between patterns in pictures.  No developer writes code that says "look for a bird, this is what a bird looks like".  Instead someone hands the software thousands of pictures of birds, then hands it millions of random pictures, and asks the software to find things in the millions of random pictures that "look like" the thousands of pictures of birds.
 
2021-01-19 10:34:00 AM  
Fark user imageView Full Size

Fark user imageView Full Size

Fark user imageView Full Size
 
2021-01-19 10:36:31 AM  

Xcott: No.  You can debug computer programs.


You can enbug AI.
 
2021-01-19 10:36:47 AM  
Despite the joke memes above, I find the AI hate and "No True Scotsman" semantic quibbling kind of tedious. Independent of the name, what people call "AI" does some pretty cool stuff that would be hard to do with alternate programming models, regardless of whether you consider anything less than Commander Data to be "not really artificial intelligence".
 
2021-01-19 10:37:22 AM  

coffeetime: No... It is the description of the last four years of the White House.


That would be RS (Real Stupidity), not AI.
 
2021-01-19 10:46:55 AM  
If Intelligence is defined as "the ability to acquire and apply knowledge and skills", then artificial intelligence is a whole lot different than a program, which usually follows very specific and logical inputs/outputs.

But build enough dynamic programs that are able to handle a wide range of situations while working in conjunction with one another, and you might get the emergence of intelligence (and mayyyyyybe sentience).

Which begs the question: What makes intelligence artificial if it's just as capable as the original? Artificial is defined as being created by humans, but any system sufficient enough to create true intelligence is certainly looking like it might currently be beyond human understanding (reference: some neural networks whose inner workings are not understood).

With that thought in mind, Artificial Intelligence to me is a system that emulates intelligence to a close degree (example: AI's following along with natural dialog progression), but is still lacking in capabilities to acquire new knowledge and skills. Almost like a precursor to true intelligence.

Any system capable of offering full parity (or more) of human intelligence is simply another form of intelligence....to me at least.

Your philosophical mileage may vary. 😝
 
2021-01-19 10:53:17 AM  

coffeetime: No... It is the description of the last four years of the White House.



That wasn't artificial intelligence, that was natural stupidity.
 
2021-01-19 10:56:07 AM  

dittybopper: The best definition I've ever found is that AI is computer code that modifies itself.

Just run through the implications of that in your head.


Is that really the best definition? I used to do that when I was learning ASM, and I was definitely not programming AI. It's just that the traditional programming languages for AI (such as LISP and Prolog) have this feature built in. As far as I know, modern day machine learning does not depend on self-modifying code.
 
2021-01-19 10:56:24 AM  

coffeetime: No... It is the description of the last four years of the White House.


No, that's "organic stupidity" and "naturally occurring toadyism."
 
2021-01-19 10:57:46 AM  

Ambitwistor: Despite the joke memes above, I find the AI hate and "No True Scotsman" semantic quibbling kind of tedious. Independent of the name, what people call "AI" does some pretty cool stuff that would be hard to do with alternate programming models, regardless of whether you consider anything less than Commander Data to be "not really artificial intelligence".


The problem isn't whether or not it can be useful.

The problem is whether or not we can actually control it once it becomes powerful enough.

Consider the recent actions by Google, Facebook, Apple, and Twitter to suppress things they don't like (Meaning Trump, and his supporters).   Those were, mostly, humans making those decisions.   This is something we inherently understand.   Politics is politics, monopolies are monopolies, and so forth.

Now imagine a company that has a very large monopoly on something important, and it's run by an AI.

It decides, based on something we may not have any comprehension of, that it needs to take Action X, and Action X is detrimental to human beings, or some subset thereof.   How can we predict that?  How can we stop it?   Pull the plug?  What if it doesn't want us to pull the plug?   What if, because it's such a monopoly, we can't pull the plug?   What would happen, for example, if Google went permanently offline?   *EVENTUALLY* something would take its place, of course, but it would cause significant disruptions, and the use of alternatives would swamp the (much smaller) search engines because of excess traffic.

Now, if I write some software and there happens to be a bug in it, I can go in and change what I've written, and if I do it correctly, the software will work correctly into the future, until something happens to change that.

AI code is self-modifying, though.   It has to be or it's not actually "intelligent", as a system with static code can't learn.   So if I'm a programmer for the Very Big Important AI Corporation, I can't just go in and change "if x >= 1" to "if x > 1" and expect that to solve the problem.   And because the code is self-modifying, there is no way to really guarantee that after I've "fixed" the problem that it won't "unfix" itself later on.

That's if I could even identify the problem at all.   AI systems are becoming so complex that the sort of "down in the weeds" programming that many of us do doesn't apply anymore.   You're basically teaching a machine.   And how you're teaching it, and what you're teaching it, can have some very unexpected results.

The Real Reason to be Afraid of Artificial Intelligence | Peter Haas | TEDxDirigo
Youtube TRzBk_KuIaM
 
2021-01-19 10:59:55 AM  
It's all a computer program when you get right down to it. It's just that it's either self-contained (so-called "artificial intelligence" that can mimic a simulation of basic intelligence, like the old Eliza program) or self-modifying ("deep learning" neural networks and the like).  Even then it's still a computer program.  It's just one that can expand its "understanding" of things, insofar as it is able to understand anything in the way we think of understanding.
 
2021-01-19 11:00:02 AM  
Humans are just fancy monkeys.
 
2021-01-19 11:00:15 AM  

kbronsito: coffeetime: No... It is the description of the last four years of the White House.


That wasn't artificial intelligence, that was natural stupidity.


Psychopusher: coffeetime: No... It is the description of the last four years of the White House.

No, that's "organic stupidity" and "naturally occurring toadyism."


Alternate Intelligence.
 
2021-01-19 11:02:49 AM  

dittybopper: Ambitwistor: Despite the joke memes above, I find the AI hate and "No True Scotsman" semantic quibbling kind of tedious. Independent of the name, what people call "AI" does some pretty cool stuff that would be hard to do with alternate programming models, regardless of whether you consider anything less than Commander Data to be "not really artificial intelligence".

The problem isn't whether or not it can be useful.

The problem is whether or not we can actually control it once it becomes powerful enough.

Consider the recent actions by Google, Facebook, Apple, and Twitter to suppress things they don't like (Meaning Trump, and his supporters).   Those were, mostly, humans making those decisions.   This is something we inherently understand.   Politics is politics, monopolies are monopolies, and so forth.

Now imagine a company that has a very large monopoly on something important, and it's run by an AI.

It decides, based on something we may not have any comprehension of, that it needs to take Action X, and Action X is detrimental to human beings, or some subset thereof.   How can we predict that?  How can we stop it?   Pull the plug?  What if it doesn't want us to pull the plug?   What if, because it's such a monopoly, we can't pull the plug?   What would happen, for example, if Google went permanently offline?   *EVENTUALLY* something would take its place, of course, but it would cause significant disruptions, and the use of alternatives would swamp the (much smaller) search engines because of excess traffic.

Now, if I write some software and there happens to be a bug in it, I can go in and change what I've written, and if I do it correctly, the software will work correctly into the future, until something happens to change that.

AI code is self-modifying, though.   It has to be or it's not actually "intelligent", as a system with static code can't learn.   So if I'm a programmer for the Very Big Important AI Corporation, I can't just go in and change "if x > ...


Is your head really shoved so far up your ass that you consider silencing someone who openly incites riots, treason, racism, insurrection, violence, bigotry, conspiracy theories, and just about everything bad about humanity to be "things tech doesnt like"??
 
2021-01-19 11:11:29 AM  
Artificial Intelligence, as it's currently used in industry, is usually just used to define software that has a level of self-governed prediction built into it with the ability to make decisions on that prediction.  The idea that it is a Program is just a definition of the medium used to create the AI, similar to how our Intelligence uses neural connections, pathways, and electrical impulses as its medium.
 
2021-01-19 11:16:28 AM  
"Artificial intelligence" or earlier in my career referred to "fuzzy logic" both meaning "using statistics to produce a result"
 
2021-01-19 11:18:51 AM  
You want to see where AI is, consider your Echo device.  The pattern matching is great for all the crap it collects on you and you can see this in your Echo View (if you have one).  But ask it to play music and if you don't get it *exactly* correct, it farks up.

Me: "Alexa, play fifty one fifty by Van Halen."
Alexa: "I'm sorry, I can't find fifty one fifty by Van Halen but here's other music by Van Halen..."
Me: "Alexa, cancel.  Alexa, play five thousand, one hundred and fifty by Van Halen."
Alexa: "Here's five thousand, one hundred and fifty by Van Halen on Amazon Music."

...another...

Me: "Alexa, play Hey Now by London Grammar, Arty remix."
Alexa: "I couldn't find Hey Now by London Grammar, Arty remix but here's other songs from London Grammar..."
Me: "Alexa cancel.  Alexa, play Hey Now by London Grammar."
Alexa: "Here's Hey Now by London Grammar on Amazon Music."
(plays wrong version)
Me: "Alexa, cancel.  Alexa, play Hey Now by Arty."
Alexa: "Here's Hey Now, by London Grammar, Arty Remix on Amazon Music."


A farking condition statement would fix that (though targeted) but even a modest AI could handle it...fark, their pattern matching (very modest "AI") could fix this but they only seem to apply that to things they can sell to you and not music requests.
 
2021-01-19 11:19:22 AM  

batlock666: dittybopper: The best definition I've ever found is that AI is computer code that modifies itself.

Just run through the implications of that in your head.

Is that really the best definition? I used to do that when I was learning ASM, and I was definitely not programming AI. It's just that the traditional programming languages for AI (such as LISP and Prolog) have this feature built in. As far as I know, modern day machine learning does not depend on self-modifying code.


I think so.  Not all self-modifying code is AI, but AI *REQUIRES* self-modifying code.

Otherwise, it couldn't "learn".
 
2021-01-19 11:23:18 AM  
Define "Intelligence."

/for me, intelligence is the capacity to observe, analyze, and effectively act upon patterns in our environment
 
2021-01-19 11:23:51 AM  

dittybopper: How can we stop it? Pull the plug?


Versioning, or restore from early backups. :-)

If it really is AI, then there needs to be a punishment/reward system. We use that all the time and it doesn't have to be harsh. We hate being wrong and we like being right, for example.

I'm way more optimistic about AI superior to humans, and unlike Elon, I want to take the chance.
 
2021-01-19 11:28:57 AM  

dittybopper: I think so.  Not all self-modifying code is AI, but AI *REQUIRES* self-modifying code.


I disagree.  A lot of basic AI isn't self-modifying at all.  Consider, for example, a standard alpha-beta search engine for playing chess, or a Haar cascade face recognizer.  Once you have them set up, they do their thing without any modification of their own behavior.

Now, you can augment these algorithms so that they update their heuristics or their dataset, but you don't have to, and they're "AI" either way.

In addition, a lot of "self modifying" AI is really only "modifying" itself during a training phase, only to be completely fixed in deployment.  Again, you can design it to continue training itself during deployment, but you don't have to.

I think a better definition of "AI" is a system that either learns on its own or reasons on its own---"on its own" meaning that although it is running a prescribed algorithm, its intelligent behavior is emergent rather than specified in its programming.
 
2021-01-19 11:29:10 AM  
Well when you realize "intelligence" is just a biological program, then yes, it is just a computer program.
 
2021-01-19 11:29:26 AM  

SurfaceTension: Define "Intelligence."

/for me, intelligence is the capacity to observe, analyze, and effectively act upon patterns in our environment


We can't discount ineffective or irrational response. Look at the whole shiatshow of conspiracy theories and politics. Plenty of creatures of intelligence arrive at horrifically "wrong" conclusions and take the "wrong" path all the time. That's real intelligence in action.

Right vs wrong debates immediately drive us to "morals" and where do they come from? No matter what we come up with for AI, it won't have morals or a "soul," whatever that is.
 
2021-01-19 11:29:39 AM  
Back in college when I was getting my CS degree I took a couple of classes of AI and at least back then... (mid nineties)  yes. They were nothing more than computer programs.

LISP was an awful monstrosity of a programming language and a pain in the ass to work with. I still have nightmares trying to match closing parenthesisisisisis. XD

We had things like ELIZA that was little more than an elaborate switch statement... and could basically only answer questions that had been prepped for...

I'm sure we've come a long ways, but I'm pretty sure if we've not come to the point where we can call AI anything other than "elaborate computer programs"

/my 2 cents
//yay slashies!
 
2021-01-19 11:33:48 AM  
Most of what the media and by extension the general public call AI isn't really close to what a computer scientist would mean by the term.

It's become a meaningless phrase that gets attached to anything that uses an algorithm to make a decision now.

We'll probably need to come up with something new to call actual research into things that can think,
 
2021-01-19 11:37:48 AM  
expert systems. (necessary sufficient conditions)
models.              (con suff necss) (history character impact)
(themostats = robots)

from agents to AI: 2bfuzzy or not to b

from the 80s when I was still thinking

*************

Pragmatics (empirical).
     object, relations, mediation

This a pragmatic empirical epistemology via accounting.
ie: Mystical, Rationalizing Pragmatic Authoritarian :

EMPIRICAL - failure to reject NULL hypothesis.
INTELLigence recognition malnipulation sysnthesis.


1: necessary sufficient conditions ( cause  effect  co relations )
3: structure function evolution
4: history character impact

QUADS:  1:DATA       -  full, complete. accurate
     2:reporting  - elegant synthetic paradigm
     3:systems    - efficient effective rational
     4:evaluation - open truth justice
 
2021-01-19 11:49:07 AM  

Orallo: LISP was an awful monstrosity of a programming language and a pain in the ass to work with. I still have nightmares trying to match closing parenthesisisisisis. XD


Lisp is truly the language from which God wrought the universe.

We had things like ELIZA that was little more than an elaborate switch statement... and could basically only answer questions that had been prepped for...

I'm sure we've come a long ways, but I'm pretty sure if we've not come to the point where we can call AI anything other than "elaborate computer programs"


A state-of-the-art AI attempting to be Eliza: here and here.
 
2021-01-19 11:54:53 AM  
AI is basically glorified statistics.

With machine learning, you take a known set of input parameters and results, and randomly split them in half.

In step 1, you train the system by feeding in the input parameters and the result, and it goes through a bunch of math to come up with internal settings.

In step 2, you test the system but running the other half of your data through and seeing how often it is correct and incorrect and if that score is good enough or not.  If it is, then you let it loose in the world, and if it isn't, then you can shake the box and try again, and if that doesn't work, you have to make sure everything on the input list is relevant (causation vs correlation) to the output and that there aren't outside influences that aren't accounted for, get larger sample sizes, try different internal engines, or bang your head.

The Machine Learning class at Coursera taught by Andrew Ng is really interesting (and free), though you need a background in linear algebra, stats, and programming if you want to fully understand it.

The biggest difference between AI and a regular program is that it uses what it has seen in the past to figure out the answer vs following a pre-programmed algorithm.

For instance, with OCR, there isn't someone pre-programming what an A looks like. Instead, say a million letters are sent through with different fonts and centerings (the output from a letter detection algorithm, which is a different AI problem), where it can see the 25,000 As vs the 975,000 non-As.
 
2021-01-19 12:00:16 PM  

urger: Artificial Intelligence just means that it is a technique that doesn't really work yet. Once a technique becomes practical it stops being "AI" and just become just another "computer program".

/Holder of multiple patents on "Artificial Intelligence Methods"
//Quotes!


There's a lot of truth to this.   Consider Siri or Alexa.   "Understands" spoken natural language and often does the right response.   We have Teslas that can monitor the road and, to a limited degree, drive themselves.   No one blinks an eye.   In the 80's and 90's, that would have been groundbreaking AI.      Now, it's "Dammit, Siri, that's not the song I asked for."

Sure, you can see that there's not a huge amount of depth to the responses -- Siri and Alexa don't think too hard about what you're asking.   But in 20 years?
 
2021-01-19 12:03:25 PM  

Orallo: Back in college when I was getting my CS degree I took a couple of classes of AI and at least back then... (mid nineties)  yes. They were nothing more than computer programs.


Same.  I think in one class we had to design basic decision making algorithms.  The most interesting thing was how a different path could be taken with each iteration even though nothing had really changed.
 
2021-01-19 12:04:13 PM  
No. A computer program is a set of instructions a human wrote, in order to tell the computer what to do.
AI is a (very complicated) set of rules an algorithm has created (or "learned"), by observing data and finding patterns.

There are multiple approaches to learning, depending on whether we
- tell the computer the outcome we're looking for (supervised learning)
- let it sift through data and find similarities, with no predetermined outcome (unsupervised learning)
- put it in a simulator and let it find its way to a solution by trial and error (reinforcement learning)
etc.

Reinforcement Learning is probably the closest to the way humans learn how to do a task. And it's come a long way in the past decade (MuZero is particularily mind-blowing and has challenged some previously held beliefs about what we can achieve with it) but we are still far from Artificial GENERAL intelligence.

Which is not to say we can't build useful stuff with what we have now.
 
2021-01-19 12:04:47 PM  

UberDave: You want to see where AI is, consider your Echo device.  The pattern matching is great for all the crap it collects on you and you can see this in your Echo View (if you have one).  But ask it to play music and if you don't get it *exactly* correct, it farks up.

Me: "Alexa, play fifty one fifty by Van Halen."
Alexa: "I'm sorry, I can't find fifty one fifty by Van Halen but here's other music by Van Halen..."
Me: "Alexa, cancel.  Alexa, play five thousand, one hundred and fifty by Van Halen."
Alexa: "Here's five thousand, one hundred and fifty by Van Halen on Amazon Music."

...another...

Me: "Alexa, play Hey Now by London Grammar, Arty remix."
Alexa: "I couldn't find Hey Now by London Grammar, Arty remix but here's other songs from London Grammar..."
Me: "Alexa cancel.  Alexa, play Hey Now by London Grammar."
Alexa: "Here's Hey Now by London Grammar on Amazon Music."
(plays wrong version)
Me: "Alexa, cancel.  Alexa, play Hey Now by Arty."
Alexa: "Here's Hey Now, by London Grammar, Arty Remix on Amazon Music."


A farking condition statement would fix that (though targeted) but even a modest AI could handle it...fark, their pattern matching (very modest "AI") could fix this but they only seem to apply that to things they can sell to you and not music requests.


thats just a better decision tree.
an 'AI' is when it goes "You have shiat taste.  Play some Hendrix"
 
2021-01-19 12:05:49 PM  

FrabjousDay: urger: Artificial Intelligence just means that it is a technique that doesn't really work yet. Once a technique becomes practical it stops being "AI" and just become just another "computer program".

/Holder of multiple patents on "Artificial Intelligence Methods"
//Quotes!

There's a lot of truth to this.   Consider Siri or Alexa.   "Understands" spoken natural language and often does the right response.   We have Teslas that can monitor the road and, to a limited degree, drive themselves.   No one blinks an eye.   In the 80's and 90's, that would have been groundbreaking AI.      Now, it's "Dammit, Siri, that's not the song I asked for."

Sure, you can see that there's not a huge amount of depth to the responses -- Siri and Alexa don't think too hard about what you're asking.   But in 20 years?


That's what I think.  And I still marvel at being able to control my house with voice commands.
 
2021-01-19 12:06:53 PM  
Lots of different types of AI.  Machine Learning and Expert Systems are just two specific examples of things within the AI world.

Artificial Intelligence is any computer program that can emulate intelligence (our very poor definition of intelligence notwithstanding).

For example, a doctor has training and needs to use intelligence to identify and diagnose symptoms.  People try to write programs that can do this, but the spectrum is too broad to build a proper diagnostic machine.  So there are specific applications like reading x-rays.  Humans are trained to read x-rays to identify things like doppler effects that look like breaks.  Computers can read these and do calculations to identify every x-ray pixel that is the result of a doppler effect and only take a couple hours to read an x-ray which humans can do in 5 seconds.  AI in this case is to use a different type of mathematical function (based on 3-D right angle turns) that reduces the computations required by 3-5 orders of magnitude so now the computer can also read an x-ray in about 5 seconds.  It is not "intelligence" like humans use; it is artificial, but it has the same result.  Artificial Intelligence.

People often think "Intelligence" is the operative word in AI.  It isn't.  "Artificial" is the word you should be paying more attention to.
 
2021-01-19 12:08:53 PM  

UberDave: FrabjousDay: urger: Artificial Intelligence just means that it is a technique that doesn't really work yet. Once a technique becomes practical it stops being "AI" and just become just another "computer program".

/Holder of multiple patents on "Artificial Intelligence Methods"
//Quotes!

There's a lot of truth to this.   Consider Siri or Alexa.   "Understands" spoken natural language and often does the right response.   We have Teslas that can monitor the road and, to a limited degree, drive themselves.   No one blinks an eye.   In the 80's and 90's, that would have been groundbreaking AI.      Now, it's "Dammit, Siri, that's not the song I asked for."

Sure, you can see that there's not a huge amount of depth to the responses -- Siri and Alexa don't think too hard about what you're asking.   But in 20 years?

That's what I think.  And I still marvel at being able to control my house with voice commands.


people in the industrial revolution marvelled at being able to push a button and have a machine do something.

we've revolutionized the button.
 
2021-01-19 12:17:54 PM  

FleshMonkey: Lots of different types of AI.  Machine Learning and Expert Systems are just two specific examples of things within the AI world.

Artificial Intelligence is any computer program that can emulate intelligence (our very poor definition of intelligence notwithstanding).

For example, a doctor has training and needs to use intelligence to identify and diagnose symptoms.  People try to write programs that can do this, but the spectrum is too broad to build a proper diagnostic machine.  So there are specific applications like reading x-rays.  Humans are trained to read x-rays to identify things like doppler effects that look like breaks.  Computers can read these and do calculations to identify every x-ray pixel that is the result of a doppler effect and only take a couple hours to read an x-ray which humans can do in 5 seconds.  AI in this case is to use a different type of mathematical function (based on 3-D right angle turns) that reduces the computations required by 3-5 orders of magnitude so now the computer can also read an x-ray in about 5 seconds.  It is not "intelligence" like humans use; it is artificial, but it has the same result.  Artificial Intelligence.

People often think "Intelligence" is the operative word in AI.  It isn't.  "Artificial" is the word you should be paying more attention to.


in a lot of ways i think the term "artificial intelligence" was a mistake.  "automated intelligence" is a far better description of what we have at this point.

we have not created an artificial replication of an intelligence.
we have automated the repetition of the known outputs of an existing human intelligence.
what we have isn't a reproduction of the thing or the process.  its just a reproduction of the results through a different much more base and simplified process.

not to say that isn't incredibly useful.  any tool is useful, and thats an amazing tool.  but its nothing like the I, Robot situation people seem to want it to be.
 
2021-01-19 12:25:24 PM  
While yes, AI is A computer program it is not "just a fancy word" for it. That idiotic simplification makes it sound like there's no difference between Minesweeper, Photoshop, Excel or Cyberpunk 2077. While they are all "computer programs" what they do, how they do it and whether you may get a seizure or not greatly varies. (Looking at you Excel!"
 
2021-01-19 12:36:41 PM  

dittybopper: The best definition I've ever found is that AI is computer code that modifies itself.

Just run through the implications of that in your head.


Nah.  Doing this isn't really that difficult.  I wrote a program in high school that did this on a farking Commodore 64-in Basic.

Pretty easy, actually-the location of where a Basic program is stored is stable and known on a C-64.  Just Poke said location with new data late in the program, then Goto the beginning of the program, and it will run the modified version the second time through.

It would look something like this (30 year old memories ahead):

10 Print "Hello"
20 Rem 100
30 Poke A, B [A being the memory location of the o in Hello, B being the ASCII value for an !]
40 Poke C, D [C being the memory location for the Rem command; D being the code to change the command token to a Goto command token instead]
50 Goto 10
100 Print "Goodbye!"

Running this the first time would give you:

Hello
Hell!
Goodbye!

Running it the second time would give you:

Hell!
Goodbye!

As the program had already modified itself the first time you ran it.
 
2021-01-19 12:42:25 PM  

edmo: SurfaceTension: Define "Intelligence."

/for me, intelligence is the capacity to observe, analyze, and effectively act upon patterns in our environment

We can't discount ineffective or irrational response. Look at the whole shiatshow of conspiracy theories and politics. Plenty of creatures of intelligence arrive at horrifically "wrong" conclusions and take the "wrong" path all the time. That's real intelligence in action.

Right vs wrong debates immediately drive us to "morals" and where do they come from? No matter what we come up with for AI, it won't have morals or a "soul," whatever that is.


One reason I use the modifier "effectively" is because of just what you say  -- people can perceive patterns, but take actions that are ineffective, like believing conspiracy theories. I'm just saying on a continuum, people of greater intelligence tend to be more effective in their actions than those of lesser intelligence.

This recognizes, of course, that the least intelligent human is still pretty dang intelligent compared to all other animals.
 
2021-01-19 12:43:51 PM  

dittybopper: AI code is self-modifying, though.   It has to be or it's not actually "intelligent", as a system with static code can't learn.   So if I'm a programmer for the Very Big Important AI Corporation, I can't just go in and change "if x >= 1" to "if x > 1" and expect that to solve the problem.   And because the code is self-modifying, there is no way to really guarantee that after I've "fixed" the problem that it won't "unfix" itself later on.


The AI code which is in a production system doesn't have to involve modification. While being trained, there is modification of something (code or data). But the resulting tool might be saved, and a copy which does not do any modification becomes the production tool. A pattern-matching tool which finds faces and eyes, then overlays glasses over the image, might be driven by the stored AI-trained data, but be unable to change its behavior.
 
Displayed 50 of 63 comments


Oldest | « | 1 | 2 | » | Newest | Show all


View Voting Results: Smartest and Funniest

This thread is closed to new comments.

Continue Farking





On Twitter



  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.