The Instigator
tvellalott
Pro (for)
Winning
4 Points
The Contender
MasturDbtor
Con (against)
Losing
3 Points

True artificial intelligence is improbable.

Do you like this debate?NoYes+1
Add this debate to Google Add this debate to Delicious Add this debate to FaceBook Add this debate to Digg  
Post Voting Period
The voting period for this debate has ended.
after 2 votes the winner is...
tvellalott
Voting Style: Open Point System: 7 Point
Started: 5/14/2011 Category: Science
Updated: 5 years ago Status: Post Voting Period
Viewed: 2,458 times Debate No: 16506
Debate Rounds (5)
Comments (8)
Votes (2)

 

tvellalott

Pro

Hello DDO!

OPENING
I will of course be arguing that 'true artificial intelligence' is improbable.
To avoid semantics, I will try to describe exactly what I mean by true artificial intelligence:
I agree with the definition of AI as provided by Google...
"The theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages."

However, I am of the belief that computers cannot be programmed to acquire knowledge and skills, but only apply a pre-programmed set of rules and calculations, however vast that code may be. That is not true intelligence, which allows new ideas to evolve from previously known information.

To begin, let me explain the Chinese Room argument; Suppose a man is in a room, secretly being given English instructions on how to manipulate Chinese characters. To anyone outside the room, it would appear the man can understand these characters, but it is only a deception.

AI works in much the same way. When we play a modern video game, it would appear that the computer 'knows' what to do; a computer program might easily outclass you in a game of Chess. However, that same program could not simply be put into a robotic claw and expected to apply. New code would have to be written to operate the claw, translate a camera image into data and apply that chess program.

There is no reason to believe a computer can be programmed to create new and logical code to apply to new operations.
MasturDbtor

Con

=Knowledge And Skills Are Acquired In Humans Because of Programmed Rules(DNA, Brain Chemistry)=

The opponent says computers can not be programmed to acquire knowledge and skills, but only to apply a preprogrammed set of rules and calculations. However, he ignores that even in humans acquiring knowledge and skills does follow a preprogrammed set of rules and calculations.

Everything that exists has a scientific explanation, even if not yet discovered. As human acquisition of knowledge and skills exists it must have a scientific explanation.

Scientists have already gone to work mapping this.
en.wikipedia.org/wiki/NMDA_Receptor
This receptor is used in learning and memory. Humans use this among other systems to acquire knowledge and skills.

How these receptors and the chemicals that bind to them work depends on the physics of these molecules, which were manufactured and put in place by the body according to the genetic code in the DNA, which is analogous to the programming in artificial intelligence.

Since we know that the process of acquiring knowledge and skills in humans is ultimately based on a set of rules and programming there's no reason to think that when programming becomes sophisticated enough that we would be unable to recreate that same process in artificial intelligence.

Furthermore science is already showing promise in progressing towards the creation of "Self-improving machines". See this interesting YouTube video featuring Steve Omohundro at Stanford University.

=The Chinese Room Argument=

The Chinese Room is interestingly parallel to those subconscious things people do that are simply embedded in our DNA. Think about "hunches" we have about people or our reflexes. We never had to learn how to do these things. Our "programming" tells us to when we respond to certain stimuli, just like a robot's programming tells it to do things when we put in stimuli. Animals in turn make more some would argue all their decisions on instinct, on their internal programming. We evolved to where we were programmed to understand how to acquire new knowledge and skills on our own. The same can happen to artificial intelligence as we become smarter in how to design their programming.
Debate Round No. 1
tvellalott

Pro

Rather than nit-pick aspects of my opponent argument, I will try and address his argument as a whole.

COUNTER-ARGUMENT
The ability to learn does not intelligence make.

My opponent has (in my mind) sufficiently proven that machines are capable of learning. While I was perhaps a bit ambiguous about my definition of True AI in the opening I distinctly said: “That is not true intelligence, which allows new ideas to evolve from previously known information.”

My opponent has failed to address this point.

Also, while I am happy to watch YouTube videos, posting one that is over an hour long is a bit excessive. Sources should be used to reinforce points, not make them.

ARGUMENTS
The Dartmouth Principle can be true without ‘True AI’ being achieved.

The Dartmouth Principle is as follows: "Every aspect of learning or any other feature of intelligence can be so precisely described that a machine can be made to simulate it.”

I, of course agree with this. My opponent is correct when he says:
“Since we know that the process of acquiring knowledge and skills in humans is ultimately based on a set of rules and programming there's no reason to think that when programming becomes sophisticated enough that we would be unable to recreate that same process in artificial intelligence.”

HOWEVER, there is a very important distinction to make: Planking.
The latest internet craze: photos of people lying flat in various and precarious places for laughs.
What purpose does it serve? Well it might amuse the masses to some extent, but it is fairly useless in the grand scale of things.
The distinction is some individual or group thought up planking. What pre-programmed set of rules did they use to logically come to the conclusion it was fun?

Computers can ONLY follow a pre-programmed set of rules, no matter how all-encompassing those rules might be, they are still locked into that cycle. That is the very nature of computers. Computers don’t need learning receptors; they need imagination, creativity and originality. How does one even begin to code original thought?

The cost of our true intelligence comes is high. We are bored, narrow-minded, selfish and illogical creatures.


REBUTTALS
“The Chinese Room is interestingly parallel to those subconscious things people do that are simply embedded in our DNA. Think about "hunches" we have about people or our reflexes. We never had to learn how to do these things…"

Not only are you are utterly wrong, but your entire counter argument is a red herring. Your hunches and your reflexes are based on things you learn during your life and have nothing to do with your DNA. Your DNA doesn’t change from your conception to your death. Your hunches and reflexes do.

Let me explain the Chinese Room argument:
The man in the room is a metaphor for a computer.
The instructions he is receiving are a metaphor for the programming a computer ‘understands’.
The man may appear to understand Chinese (as a computer may appear to be intelligent) but in fact has no idea what he is doing.

A computer may know appear to know, understand and be capable of interacting with a myriad of things, but it is merely an illusion created by clever programmers.


CONCLUSION
We must come back to the resolution: True AI (as I defined it in R1) is improbable.

I have explained my argument; In order to be considered truly intelligent (like a human) a computer must be able to program itself. This opposes the very nature of a computer! Computer programming has become more complex and sophisticated in the last decades, but that is only because of the evolution of the way we interact with binary. This indicates that while computers may grow faster and faster into infinity, the nature of them is unlikely to change.

Resolution affirmed.
MasturDbtor

Con

Intelligence

It is easy to conceive of how it is possible to make a program "which allows new ideas to evolve from previously known information".

It would need a memory, and a complex set of ways to compare and contrast information.

That's not all that different from how humans build intelligence. We take in sensory data from our surroundings, some consciously, some subconsciously, over time again somewhat consciously and somewhat subconsciously we put things together and draw conclusions that we then apply to everyday life to pursue new things and consequently gather new information that is run through the same process.

The mechanisms of how this works is complex, but that's intelligence in a nutshell.


Planking

As for Planking and what purpose it might serve to a potential computer program consider this...


Utilitarianism



Logically everything has a reason for it and reasons upon the reasons, except for happiness. You want to be happy to be happy, no other reason. Everything else has another reason. Utilitarianism is a theory of logic, morality, and ethics that most logical, moral, and ethical thing to do is what ever increases everybody's happiness. One's own happiness would figure into it. For one thing to be good for anyone else's happiness you have to have some yourself.


Now imagine a robot programmed to be utilitarian or alternatively you even programmed a hedonistic robot. Hedonism refers to considering only your own happiness. Or a robot could even be programmed to be logical but still have some experience of feelings and then rationalize that the most logical thing to do is to pursue happiness and pleasure.

Now imagine that this robot was programmed to respond positively and negatively to certain sets of sensations, sounds, etc. at first, but with a framework that sets in motion certain complex sets of rules that determine how these responses and preferences change in response to the robot's observations in the environment. This would make the robot's development of likes and dislikes almost indistinguishable from a human's. All our preferences must also come from somewhere. These things don't happen for no reason. Psychology has already shown how past events shape us in the future, and how temperament at birth even factors in.

The only difference here is the robot would be conscious of it.

Easy fix. Program the robot's central processing unit, the part of its hardwire responsible for most of its movements, its decisions, etc. to not be aware of the rest of the robot's brain, but still programmed so the rest of it is stored information that is fed into the central processing unit. Make the robot somewhat capable of looking into the rest of it, but not without conscious choice and effort and not an unlimited ability to do that. Make another part still completely inaccessible.

Then you have a robot with a conscious mind, a subconscious mind, and an unconscious mind.

If the robot was programmed with a framework to potentially develop a wide range of interests, pleasures in interaction with a wide variety of environmental stimuli the robot could develop a sense of humor through thought processes hidden in its subconscious leading it to get a kick out of that YouTube video or even try to post one itself.

Imagination

Imagination, creativity, and originality, our ability to do those things still all comes from somewhere, the ability for it to develop is encoded in our DNA, and its development is dependent on our environment. We already have science on some of the neurology of it. We know for example that these things are more active in the right hemisphere than the left. Replicate a similar set of processes as the human brain and you will get True A.I.
Debate Round No. 2
tvellalott

Pro

REBUTTALS
INTELLIGENCE & IMAGINATION
“It would need a memory, and a complex set of ways to compare and contrast information.”

You’re simply rehashing an argument that has already been rebutted by the “Chinese Room Argument”.

The throbbing question is: How do you make the step from “Doing a task” to “Understanding a task”?

“Replicate a similar set of processes as the human brain and you will get True A.I.”

Perhaps this is the answer? Replicate the mind?
However, the brain is not a computer.
There is no distinction between software and hardware in the brain.
There are all sorts of processes going on that we don’t have control of.
They are fundamentally different!
I have not once denied that a computer can simulate some of the base functions of a human brain, but that doesn’t bring us any closer to an actual thinking machine. Again, how do you make the leap to original thoughts? How do you code a machine to write code for new tasks? None of these questions have been answered.

UTILITARIANISM
“Logically everything has a reason for it and reasons upon the reasons, except for happiness. You want to be happy to be happy, no other reason.”

Ummmm, what? Everything has a logical reason for it, except happiness? O_O
This is blatantly wrong. Everything has a scientific reason, even if we aren't certain of what it is. The argument I've made is that humans are masters of finding illogical ways of reaching desired conclusions.

My opponent goes on to make a lot of assertions about how a computer could be programmed to act like a human, even going so far as to describe one potential problem as “an easy fix”.
Unfortunately, there is no currently known way of even beginning to tackle this task. I can imagine a robot that can think; it’s been done many, many times before (Terminator, I, Robot, 2001: A Space Odyssey). Imagination is something Humans are capable of, but difficult to translate into 1’s and 0’s.

CONCLUSION
My arguments are based on two assertions which I have provided arguments for:
Computer programming hasn’t gotten any more advanced. We have much smoother and more powerful ways of interacting with it, but it’s still the same binary code at the core.
Artificial Intelligence is ‘just pretend’. It gives the illusion of knowledge, but will fail when an unknown variable is introduced. The Chinese Room Argument (which my opponent hasn’t refuted) deals with this directly.

Resolution affirmed.
MasturDbtor

Con

1. Not the Same as the Chinese Argument

In the Chinese Room the man "speaking" Chinese only takes information from the outside and spits it back out without involving any process within the room other than taking information from outside and sending it outside.

Having a "complex set of ways to compare and contrast information" would be more equivalent to if the person in the room already had mounds of information in there from past exchanges and while in the room was very busy squaring the different pieces of information together to discover new information(analyzing), and even possibly to come up with novel solutions. That's vastly different from the guy who is just taking information in and sending it out.

2. "The throbbing question is: How do you make the step from "Doing a task" to "Understanding a task"?"

How do humans make that step?

We don't fully understand this yet, but we do know the answer is scientific and involves physical mechanisms within the human brain.

There's no reason why replicating that same system of interaction in a machine would not produce the same result.

3. Rebuttal-No distinction between software and hardware in the brain.

"Software" and "hardware" are terms we use for computers, but there are analogies with the human mind. The hardware is always in the computer, the software is put into the computer for temporary use. For example, a computer game CD would be "software", the computer's built in mechanism to read CDs is part of its hardware. The computer does not contain within it the information to execute that program, but once it is able to "read" the software it applies the instructions within the framework in its hardware and can do it.

Likewise, a human might need to get to a town and not know the way to go. The information about how to get there is not within the human's brain, it is not within his hardware. He pulls out a map and reads it, then he knows how to do it. The brain itself is analogous to the hardware, containing the framework needed to interpret the map, which is the software. We do not shove the map into our brains, but we become aware of the information on the map with our eyes just as the computer becomes aware of the information in the disc with its scanning equipment.

4. Processes We're Unaware Of

How does that make us more "thinking"? That we're unaware of these processes means that for these processes we're really just following our previous programming. It's good we have these, though since otherwise we would have to think in order to do such simple things as digest. This just makes the point that even if AI still had some processes that weren't under the control of its "intelligence" that wouldn't mean it isn't True A.I. so long as it does use intelligence for some things.

5. Original Thoughts/Imagination

Pro asks "How do you code a machine to write code for new tasks?" Pro defined "true intelligence" as "allows new ideas to evolve from previously known information". Any new idea we have is based off of our previous experiences, sensations, old information stored in our brains, and then put together again in new ways based on present needs. This all follows a physical process in our brains. No magic, no matter how wonderous our ability to do this might seem it is all science. If a robot can have old information and past experiences and can go over them again and again, and can receive sensory input about its environment and can apply this to achieving certain outcomes and goals including joy and entertainment we have original thought and imagination in a machine.

6. Utilitarianism

Clarification: Pro seems to be confused by this argument. I was not arguing that happiness has no logical reason for it in the sense he implies. Happiness, like everything has a cause. What I was referring to is that "happiness" does not have to be justified by an "end", does not require a logical reason to explain why someone would have that as a goal. When someone does something to make themselves happy it would be silly to ask why they wanted to be happy and what they planned on using their happiness for.

The point was while we think of robots as logical and hence that they would not consider "emotion" this is because our society has placed a false dichotomy between considering logic and considering emotion. Since happiness is the only thing good in and of itself it is logical to put logic towards the goal of achieving happiness. It's not hard to conceive of an adequately programmed robot, programmed with feelings to apply logic towards "happiness".

7. Binary Code Rebuttal

There's nothing about artificial intelligence that requires us to keep it in binary.

Human beings have our own programming too, in our DNA. We have adenine, guanine, thymine, and cytosine for each "spot" on the genome. This is analogous to the 1s and 0s in artificial intelligence. The difference is instead of binary we are machines run on a quaternary code. If it has to be done to enhance artificial intelligence we will switch over to using a system of 0s, 1s, 2s, and 3s.
Debate Round No. 3
tvellalott

Pro

tvellalott forfeited this round.
MasturDbtor

Con

All arguments extended.
Debate Round No. 4
tvellalott

Pro

I’d like to apologise to my opponent for forfeiting the last round. It really couldn’t be avoided. I’d also like to extend my warmest thanks to him for a fun, challenging and educational debate.

FINAL SUMMARY
My opponent’s entire argument rests on the assumption we will one day be able to replicate the thought processes of the human brain with computer code. You cannot simply keep saying “A computer will eventually be able to replicate the processes of a brain, we just can’t do it yet”. This is an argumentum ad ignorantiam and just doesn’t cut it. I have already explained numerous reasons why this is very unlikely to happen:

    • I have argued that this is an impossible task as the brain and the hardware/software combination work completely differently. If anything, a computer would have to work completely differently to a brain to achieve the same goal.

    • I have argued that the processes of the brain are drastically more complex than binary code. My opponent suggests that we upgrade from binary to a quadrinary-type code, which is parallel to our DNA. I really have to face palm at this counter argument. Our DNA is part of our development, yes. However, our DNA isn’t what makes us think or learn. The appropriate sections of our brains do. Trying to convert DNA into computer code is a complete waste of time. Any parallels between it and computer code are irrelevant.

    • I have presented the Chinese Room argument, which destroys the notion that computers in their current form will ever be able to think. I’ll reiterate: Even if everything that every person on Earth has ever known was uploaded into a computer and complex algorithms were written to cross reference data across every field and new ‘ideas’ were formulated using incoming space data for true randomness, it is STILL a façade of intelligence. If any new data was introduced, for example pointing the computer’s camera at someone planking and telling it to analyse the person to try and understand what they were doing, it couldn’t. Computer’s only function because of carefully structured logic; Human thought is illogical.


My opponent answers my own question (“How do we make the step from “Doing a task” to “Understanding a task”) with his own question (“How did humans make that step?”)
With great difficulty; we evolved from simple life forms into complex ones over billions of years. There were millions upon millions of tiny steps.

I will rebutt the rest of my opponents counter-arguments in my conclusion, since they all stem from the same point…

CONCLUSION
My opponent seems to want to suggest that by creating a mechanical brain, we could achieve artificial intelligence. I would have probably had to try an entirely different line of arguments to that, but he hasn’t. He has only argued that one day we might be able to replicate the processes of the human brain with code. I have already explained the fallacy of this.

In summary, the human brain is a far different beast to the lowly computer. There is no doubt that computers of the future will be able to solve all our problems, efficiently and quickly. There is no doubt that the artificial intelligence of the future will be mind boggling complex, to the point of exceeding humans in many areas. However, computers ultimately can only do what we have programmed them to do. This is their only purpose.

The resolution is “True Artificial Intelligence is Improbable”. I have fulfilled my burden of proof. The resolution is affirmed. Vote PRO.
MasturDbtor

Con

MasturDbtor forfeited this round.
Debate Round No. 5
8 comments have been posted on this debate. Showing 1 through 8 records.
Posted by tvellalott 6 years ago
tvellalott
Sorry about the for forfeit. Moved house last week and didn't have the internet on. :(
Posted by gizmo1650 6 years ago
gizmo1650
Fun topic. We had this debate in my computer science class. The answer we ended up with is that humans do not even meat our criteria for intelligence.
Posted by tvellalott 6 years ago
tvellalott
Time is running out. I better finish this noaw.
Posted by Cliff.Stamp 6 years ago
Cliff.Stamp
Shaping up to be an interesting exchange.
Posted by tvellalott 6 years ago
tvellalott
Exactly. I love the entire subject.
Posted by Cliff.Stamp 6 years ago
Cliff.Stamp
It is interesting, the foundational question is one of paradigms, what about if we are actually in the room now.
Posted by tvellalott 6 years ago
tvellalott
I'm playing devils advocate here after I read an excellent article on The Chinese Room argument which made me doubt the possibility. I too am hoping this isn't a waste of my time.
Posted by Cliff.Stamp 6 years ago
Cliff.Stamp
Nice topic TV, hope that Mr. Bator puts forth solid argument.
2 votes have been placed for this debate. Showing 1 through 2 records.
Vote Placed by quarterexchange 5 years ago
quarterexchange
tvellalottMasturDbtorTied
Agreed with before the debate:--Vote Checkmark0 points
Agreed with after the debate:--Vote Checkmark0 points
Who had better conduct:Vote Checkmark--1 point
Had better spelling and grammar:--Vote Checkmark1 point
Made more convincing arguments:Vote Checkmark--3 points
Used the most reliable sources:--Vote Checkmark2 points
Total points awarded:40 
Reasons for voting decision: both forfeited but Pros arguments about how computers may advance but not change to be able to program themselves stood. Con also forfeited repeatedly
Vote Placed by RoyLatham 5 years ago
RoyLatham
tvellalottMasturDbtorTied
Agreed with before the debate:--Vote Checkmark0 points
Agreed with after the debate:--Vote Checkmark0 points
Who had better conduct:--Vote Checkmark1 point
Had better spelling and grammar:--Vote Checkmark1 point
Made more convincing arguments:-Vote Checkmark-3 points
Used the most reliable sources:--Vote Checkmark2 points
Total points awarded:03 
Reasons for voting decision: Good topic, heavy with semantics. Con's argument is that since human bras are purely physical it's likely that what the do can be replicated physically. I think that's compelling. "Learning" via self-modifying code has been done, so the question is whether than meets the definition.