Home
Sometimes I think intelligence is nothing but pattern recognition. Someday in the near future a computer scientist will write code that rapidly compares and stores complex patterns. To populate the computer model with data, the programmer will let his software read the entire Internet, or as much as it can, and look for patterns. After a few days of chewing through content on the Internet, the software will appear to understand everything about humans, from our language to our history. The reality of this apparent "understanding" would be nothing more than pattern recognition.

Consider a simple example: A computer could learn by pattern recognition that humans raised in specific parts of the world usually say "bless you" to a person within earshot who has just sneezed. Then overlay the pattern that people usually whisper in a library and your artificial intelligence knows it should whisper "bless you" when someone in a library sneezes. In a more complicated scenario there might be hundreds of patterns intersecting, but a computer could easily contrast and compare them.

I could be wrong, but I suspect that artificial intelligence will grow out of one page of clever pattern recognition code combined with exposure to every pattern revealed on the Internet. It would take about a week to turn the pattern recognition code into what appears to us a sentient being with super intelligence. Futurists call that day the singularity.

As with most of my posts, some of you will tell me about all of the fiction books that say the same thing but said it first. I haven't read any of those books. Nor do I know anything about actual AI research, so perhaps it's obvious that pattern recognition is the key, and the real problem is that it's a hard nut to crack.

I came to my hypothesis that intelligence is just pattern recognition because people who are not terribly bright have trouble understanding analogies. And analogies are just patterns.

Logic isn't a big part of human intelligence. Put three humans in a room with a problem and each will have a different idea of the logical solution. Humans are rationalizers, not logical beings. Computers don't need logic to act human because humans don't have enough logic that anyone would notice some was missing.

Experience is little more than having more patterns to draw from. New situations are never identical to old ones, but they might follow a general pattern. For example, when I was in my twenties and someone said they would call me back I assumed it was true. Now I look at the entire situation, and all the patterns involved, and half the time I correctly tell myself I'll never hear from that person.

My two questions for the day:
  1. Is intelligence much more than pattern recognition?
  2. If intelligence is mostly pattern recognition, how likely is it that someone will write code that scours the Internet for patterns and uses that as a base to create super intelligence?
 

 
Rank Up Rank Down Votes:  +25
  • Print
  • Share

Comments

Sort By:
0 Rank Up Rank Down
Mar 25, 2014
Helloworldo said:
"It is obvious that our various and ambiguous languages add considerable complexity to AI. As a first step, let's create a single unambiguous language for all mankind. It may take a generation or two before we're all on the same page, but humans are smart like that. This will allow our robot counterparts to get up to speed more quickly. "
One problem with this idea is that language demonstratibly affects thinking, and that different thinking is the source of most "genius." By having everyone speak the same language, you would stifle the formation of new ideas.

@Scott,
One of the problems with pattern recognition is that it requires a different type of computing, granted one that will be more available in the future. Neural networks (multiple node computing) is the only way I'm aware of to hold "Platonic ideal" objects and compare new observations to the platonic ideals and determine whether a three legged stool falls under the idea of "Chair" or "Tricycle." Most pattern recognition software on a home PC creates artificial nodes and simulates the idea of millions of processors instead of one (or 2 or 4 now), which is obviously slow and cumbersome. I was priviledged to work with a supercomputer using 68 servers each with 8 cores-made life a lot faster for me.
 
 
Mar 5, 2014
I don't think the internet is really adequate - true understanding would require observing (in some fashion) real human relations.

Even genius robots would have very differing opinions/analysis. Imagine a robot that has "true" understanding of Jane Austin novels, and prioritizes those patterns. That will be a very different robot than one that has prioritized Stanley Kubrick films. Just as humans (including genius humans) have wide variety, genius robots would necessarily need to have differing world views.
 
 
+4 Rank Up Rank Down
Mar 3, 2014
Regarding mimicking human intelligence, it always makes me think why we place so much emphasis on creating intelligence like ours, which is hugely flawed, emotional, and irrational. It's like purposely building something that sucks.

That said, with emotions taken out of intelligence, we wouldn't live long.
 
 
Mar 3, 2014
Pattern recognition isn't enough. you also need to know what is significant. in the library example, there are hundreds of things that happen to your body and mind in that scenario. Which are significant? Saying the words? looking up? facial expression? Computers would have a near impossible time trying to extract that which is significant from any human interaction. It's easy for us because we "know" two (at least) things:

- what we are consciously aware of
- what we have been conditioned to think of as significant by our community (which can be different in different communities or even between individuals within a community)

From there we can build the patterns we need. if we can inject that learning into a computer then there is no reason why they could not do the same but that, i think is the hard part.
 
 
Mar 3, 2014
The equation for intelligence: http://new.ted.com/talks/alex_wissner_gross_a_new_equation_for_intelligence
 
 
Mar 3, 2014
The idea has merits but also has flaws. Patterns and recognition are the feedstock of genius, there is also the small issue of processing such patterns (projecting out the effects, potential side effects, priority, sequencing, testing methods, etc.). Second to make it worth anything there would need to be a feedback mechanism to determine which have 'merit' and which are rubbish.

A lot of issues are also dependent on the current state of reliable technology. It's one thing to project a genius idea based on a 50 times powerful laser but another to develop such a laser.
 
 
Mar 2, 2014
"I came to my hypothesis that intelligence is just pattern recognition because people who are not terribly bright have trouble understanding analogies."

I came to the conclusion women don't like sex because they have trouble saying yes when I offer it."
 
 
0 Rank Up Rank Down
Mar 2, 2014
>Head out of Ass Theory
I love it!

:-)) Sure, I'm just suggesting >>individual cognition located in a sense of self,
is mostly bias. I wonder if your idea of Point of View can be replicated in terms of an individual's bias and pattern recognition.

>>>if human intelligence was truly reducible to mere pattern recognition, our lives would be arid indeed...

I'm curious how you know they aren't? (Any more that what you give it) The deer out my window, does it's life have meaning? It has pattern recognition, some would equate as intelligence to the level of a 4 year old.

I might go as far to say most people only generate a facsimile of human understanding and you would be hard pressed to know if I am capable of relating any of it to my own needs and feelings. Is a sociopath somehow no longer a Homo Sapien?
 
 
0 Rank Up Rank Down
Mar 2, 2014
Scott,

Evolution has not reached a point where human offsprings are born with objective skills, like for instance, speaking a human language at birth.

Applying your imagination to this fact, the human baby is nothing but an impotent organic mass with a non-zero chance of developing intelligence.

And in contrast, the programmed robot will be guaranteed to deliver intelligence.


Egad, Scott. You are provoking the Pope!!!

.

 
 
Mar 2, 2014
I liked this post.... I have a similar idea of this pattern recognition... It's called the Head out of Ass Theory. - you heard it hear first.

Generally put, "The longer you keep your head out of your ass, the more you'll learn".

Now learning is different from understanding and understanding is different from knowledge. Though intelligence is the method by which we convert information (by keeping our head out of our asses) into useful knowledge, it doesn't necessarily mean an intelligent person has a lot of useful knowledge. An intelligent person has the *ability* to convert information in to useful knowledge. A dumb person has a *less of an ability* to convert information in to useful knowledge.

There is a different level of understanding that has to do with the knowledge that you know. In a sense, it is keeping your head out of your ass even further. It goes on. The more you know, the more you can put together in to more and more useful knowledge.

But the critical part of Scotts argument is pattern recognition which is basically keeping your head out of your ass and paying attention to the world around you and taking that information and applying to statistically guessing at what's going to happen next.

Humans are somewhat unpredictable. A more complicated scenario would lower your odds at guessing what's going to happen as opposed to a more common situation that has more reliable outcomes.

Situation #1: If I'm a normal looking guy with a normal body and nothing is odd about me and I walk into a brothel with a fist full of cash, chances are relatively good I can get laid.

Situation #2: If I'm a normal looking guy with a normal body and nothing is odd about me and I walk into a bar and start talking to a pretty enough girl, chances are I'm not going to get laid.

What's the difference between the two examples?

Conclusions:

#1 Women are money grubbing, unpredictable !$%*!$ who are absolutely impossible to understand.

#2 Keep a fist full of cash on hand. Just in case.

A note to Scott: I haven't purchased your book because it has one major flaw described right in the title.... "... and still be successful". I've been busting my ass and I use a lot of the same thinking that you do. The difference is that you ARE Successful. I am not. I have yet to have luck come across my path and help propel me to a more successful place. I too use "systems". I too look at a lot of pattern recognition. I am persevering, and for right now, that is all I can do. Would you have written the book and made the same arguments if you were still just plugging away at a job with a dumb boss and trying to made ends meet while complete undeserved piles of !$%* make millions? Would your "systems" make sense? Or would they be a fun notion that helps you keep some form of sanity in an insane world?

I loved God's Debris. I like that kind of high level thinking and thought experiments. I even came to your restaurant in Dublin a while ago (during a trip up north) to get my books signed but it was closed.

ps. I love this blog. It's a nice refreshing rattling of thought experiments.
 
 
0 Rank Up Rank Down
Mar 1, 2014
@MTBob

[Give it whatever point of view you want, In fact, program it to take the context and Point Of View depending on the biases of the person interfacing with it.]

I believe you misunderstood what I was getting at. What I meant by 'point of view' was not 'biases' but 'individual cognition located in a sense of self'; it could plausibly be argued that a machine possessing that faculty is actually alive. (How compatible the POV of such an ent_ity would be with ours is a familiar trope in science fiction.)
 
 
Mar 1, 2014
Pattern recognition beyond the abilities of humans is already here. It's called "machine learning" and it is the fundamental underpinnings of many modern applications -- search being the most obvious one. How do you think Google seems to know what you are about to type before you type it?

Pattern recognition is *necessary* but not *sufficient* for intelligence. It can give software the ability to answer virtually any question. What it does NOT give software is the ability to ask those questions, and it is the asking that is the true hallmark of intelligence. Google Search is the most powerful question-answering machine ever built, but until someone asks it a question, it just sits there, doing nothing. It doesn't come up with questions of its own.
 
 
Mar 1, 2014
Mathematics is mostly about pattern recognition. Intelligence is more. That is what I learned from a good book titled The Language of Mathematics. Since you (and many commenters) are heavy in math experience, I think it is easy to see the bias toward pattern recognition equating to intelligence. But much of intelligence includes creativity, which is how to use recognized patterns.
 
 
+5 Rank Up Rank Down
Mar 1, 2014
Scott, do you employ an algorithm--one that could be coded in an existing software language--to create Dilbert?
 
 
-2 Rank Up Rank Down
Mar 1, 2014


Focusing on Cat Videos? My God, it is becoming human.

Saying pattern recognition is difficult, or that we can't do it yet, doesn't make the statement untrue.
I would say, mostly pattern recognition, and I'd bet it's already being done now.
I love spell-check.

Meaning? Give it whatever meaning you want. If your meaning of your life is whatever you want it to be, program it with whatever meaning you want.

Give it whatever point of view you want, In fact, program it to take the context and Point Of View depending on the biases of the person interfacing with it. If it's a southern redneck interfacing with it, have it kick in it's Duck Dynasty sub-routine. If it's a, let someone else pay the bill POV, activate the democrat loop. If the person interfacing wants to dictate my bedroom activities, bring in the GOP. Just look at the person's facebook page and tweets and emails... you will get their bias. Or just pick your own.


>>So how are these computer supposed to tell the difference between internet-stated truths and fiction?
The same way you know to be suspicions of InfoWars if they told you Bush ordered the planes into the WTC, but if NBC, CBS Fox, NYT, were all verifying it too, and add some time for self correction, you might pay more attention. Just be careful of the outliers.


>>But an artificial intelligence can never truly UNDERSTAND the data it is analyzing, because it is incapable of relating any of it to its own needs and feelings (since it is incapable of EXPERIENCING those things).

I think most humans don't understand the data they are analyzing. And you relate understanding data to experiencing needs and feelings? Just add in some bias and suspicion.

Mimic the patterns of a thousand humans, drop the exceptions. How can you know if it's in touch with it's feelings? I don't even know my needs and feelings.

>>Imagine if an AI decided that 1/4 of the world spoke Chinese, so 1/4 of the time it spoke in Chinese.
?? If you interact with it in Chinese, it responds in Chinese....

Isn't teaching a child mostly done through pattern recognition? Watching and listening to you?

>>>For example: if I give you a scale and 7 identical balls plus 1 ball that weighs slightly less. What is the minimal number of times you would need to weigh those 8 balls in order to determine with 100% certainty which ball weighed slightly less?

6, 3, 1, 0, 2.

OK, which of those 5 answers was produced by a human and which by an artificial intelligence?

If humans are wrong some, all, most, part, of the time, should Artificial Intelligence mimic that too? Are we making a better intelligence or non-detectable machine intelligence? An always correct intelligence, or a human-like intelligence?

If we are going for human-like, make sure to program in some error, stupidity and bias?
 
 
Mar 1, 2014
Pattern recognition is part of intelligence, but not all of it. And computers can use other methods too. E.g., A.I. researchers spent about 40 years trying to teach computers to recognize and process patters like human grandmasters do. And Deep Blue used pattern recognition. But the main reason D.B. beat Kasparov was, D.B. could just analyze more moves per minute, and without forgetting any of the variations.
 
 
Mar 1, 2014
It is obvious that our various and ambiguous languages add considerable complexity to AI. As a first step, let's create a single unambiguous language for all mankind. It may take a generation or two before we're all on the same page, but humans are smart like that. This will allow our robot counterparts to get up to speed more quickly.
 
 
Mar 1, 2014
So how are these computer supposed to tell the difference between internet-stated truths and fiction? For example, what happens if the computers decide, from the massive volume of material available, that the !$%* sites portray actual human social interactions in all cultures?
 
 
-1 Rank Up Rank Down
Mar 1, 2014

[Chess actually supports my point. Humans use patterns and today's computers use brute force to calculate all paths. I'm suggesting that computers will someday be built that can recognize patterns as well as a chess master, thus not needing to calculate every path. -- Scott]

I see that point. A computer that could play chess like a person would be part of the way down the road to thinking like a person, and having AI.

I think the problem, which a lot of people are pointing out, is the pattern recognition. The first chess software was written that way 30 years ago, and it didn't work. Any half decent chess player could tell you 10 rules that make a position good or bad, and these could be give to a computer. The problem is weighting them, deciding which overrules the other, and maybe most importantly, deciding when to ignore them.

Humans just *do* that sort of thing. If it is solely pattern recognition and no higher function it's a very complex interplay.

I guess for your theory to be testable, some leap would need to occur in the future about how we teach computers pattern recognition. I would lay a good size wager that for Chess, you wouldn't get there just be downloading and parsing the internet.
 
 
Mar 1, 2014
From Gizmodo / NY Times

Hidden away within Google's X laboratory, where all kinds of secret projects are underway, its engineers have been working on creating an artificial brain. With 16,000 computer processors and freedom to learn whatever it chooses from the internet, though, it turns out that the brain does just what you do online: watch cat videos.

The project has produced one of the largest neural networks ever created. The idea is that such technology can take data sets and notice patterns and trends with them, all by itself. But when presented with a data set of 10 million digital images found in YouTube videos, it decided to, umm, learn how to identify cats.
 
 
 
Get the new Dilbert app!
Old Dilbert Blog