In respect to neural networks and how they are trained, it is not so dissimilar to the manner in which you might train a child to learn. When you see an AI agent playing Super Mario Brothers, you might be tempted to believe that it is thinking but in fact it is simply accomplishing things using rewards-based optimizations. This goes back to the “self aware” question that was posed earlier.
No, artificial intelligence cannot think. A necessary precondition for thinking is libertarian freedom--the ability to indeterministically choose between alternatives--and computers (fortunately) lack this ability.
Human thinking is not only to logically combine existing knowledge and arrive at a solution to a problem or make a decision but to reason and imagine outside of the established galaxy of knowledge to create a new kind of knowledge. The thinking of machines on the other hand is a capacity of reciting input procedures and logical operators to solve problems using an existing knowledge.
No, artificial intelligence cannot think. A necessary precondition for thinking is libertarian freedom--the ability to indeterministically choose between alternatives--and computers (fortunately) lack this ability.
If by thinking we basically understand "reason" (that is, conceive and generate actions in a logical, sequential and useful way, definitely yes: AI does.
If by Artificial Intelligence you include the cybernetic system including the operator then of course all Artficial Intelligence programmes can think. We have shown this in our research on Stock Exchanges in India which are electronically networked with switching circuits and Boolean Field logic see for e.g. the proof of the P vs. NP problem on our researchgate page www.researchgate.net/profiles/Soumitra_Mallick.
Soumitra K. Mallick
for Soumitra K. Mallick, Nick Hamburger, Sandipan Mallick
We go on the way first to define what is 'think'. 'Think' is a set of actions to perceive things around, abstract rule underlying, raise questions spontaneously and be able to understand others' questions, do analysis with knowledge to answer questions, interpret the feeling. From this perspective, the AI now indeed can do some thinking in its specific area. However, mostly, the question or problem the AI to resolve still needs very specific and sophisticated interpretation in machine language e.g. programming of a mathematics model, as well the way for AI to give the answer and solution. From this point, it just does very limited thinking actions by itself. It's still very far for AI to do really free perception, abstraction, and interpretation.
My opinion, there is limited think for artificial intelligence. Its think depends on the technical training of the human. But, until now, no artificial machine able to reach to the level of human thinking.
In a philosophical sense this question boils down to the following questions:
Is it possible– and why and how – to implement intelligence in and by non-biological architectures which is comparable with the total and the whole of human thinking and consciousness.
Can (with what objective?) or may (under which circumstances) this intelligence/thinking exist without executory, supervisory or strategic control of responsible human beings?
What is the teleology of an uncontrolled growth of artificial intelligence (all cognitive capabilities conjointly) in a long range?
You might want to take a look at this question:
https://www.researchgate.net/post/Is_it_IN_PRINCIPLE_possible_to_implement_human_desires_and_intentionality_in_a_Human_Level_Machine_Intelligence_How
I am agree with Jesus Retto.
The key point of AI is "reasoning system"
Dear Adriana Santos-Caballero
Why not build-in the very training needed (that which may be necessarily required, but no more)? (You would be very hard-pressed to show HOW that would be, or would have to be, different from "external" (e.g. parental) guidance -- kids "don't listen" that much anyway; and, thus more than you might imagine, behavioral change likely relies on their "internal", "self" guidance (correctly termed: major innate guidance factors) in play throughout child development, i.e. ontogeny, some only having their effects (or big effects) ("emerging") late in child development, i.e. adolescence or young adulthood.
Dear Imre Horvath
Again, we have "philosophical concerns" posing (or posed as) questions SUPPOSEDLY TO BE ANSWERED [seemingly] simply by using the cognitive facilities OF ONE'S OWN MIND (or should I say "minde") , and apparently requiring NOTHING ELSE (or at least no guidance on how the answer should be obtained, specifically, what is required to be OBSERVED).
The real problem with these "questions" situations is that the meaning of the question could not even be agreed upon and, likely, could not remain the SAME question in any reliably expressible way, EVEN WITHIN ONE'S OWN MIND. The reason for this assertion and for related assertions, below, are BASED ON FACT (or at least VERY reliable, oft-proven valid, strong data from THE SCIENCE OF MEMORY -- among the very best and strongest findings in all psychology).
The real problem, though, is that even the "philosophical" questions, THEMSELVES, are too much to entertain or contemplate without the bolstering and guidance of external (direct, empirical, real, real-time) observations. This primary fact (fact which has primacy) is shown by all the best and good research on the nature of our Memories -- the limitations of which it would be irrational to deny.
Dear Others:
For all those several above who basically delimit why AI cannot be like a human:
It is clearly possible that you cannot conceive of the needed sort of open [though inherently guided** ] system (in the hardware/software) to progress and change on its own forever. BUT: You certainly do not prove otherwise, and you simply cannot prove the impossibility of my last sentence.
** FOOTNOTE The "inherently guided aspects" have yet to be discovered and well defined, to say the least; but, there is no reason to believe that is impossible.
I would agree with Prof. MacGregor's response. Thinking also involves postulation which computers can not do. Postulation requires collapsing a huge body of knowledge & observation indeterminately into a simple pattern. For instance, computer can never come up with statements like "There exists a set", "All humans are born equal" etc. Computer will never be able to understand what equality among humans means!
In my opinion, 'think' might not be the right keyword even humanoid robots like sophia and erica looked they think before answering questions posted.. computers (AI) can analyze data and assist in decision making but they still work according to the algorithms..
Although it seems like some here think that there is some magic happening in biological systems that artificial systems "just cannot have", the point is:
However you define "think", as long as a human is able to perform it, a (theoretical) computer can do it.
The human brain consists of neurons that can be emulated (1:1). The synapses in a human brain can be emulated (1:1). Any input can be emulated(1:1), so the entired brain can be emulated(1:1) (theoretically, we might need a really big computer). A human brain can think, so this computer program would be able to think.
This would be an extremely inefficient ansatz, but it would work.
emulating stuff 1:1 is most propably not the right way to do it, but it proves that it a thinking computer is possible.
There is no magic happening here. The problem with Computer Scientists who 'think' that "Theoretical Computer" can do whatever humans can do is first they reduce humans to the level of Computer and then establish equivalence. This reductionist approach is at the heart of problem solving approaches humans adapt.
If you read about CNN and its application in AI, you can see that: if we want to solve a complex problem, we can divide into some layers (to handle every task). By combining these results we can solve that complex problem. That is the "think".
“Computers (AI) can analyse data and assist in decision making but they still work according to the algorithms…”
Nowadays that statement is losing strength. The new important support for AI is Deep Learning, by which the programming of algorithms is no longer the axis of an intelligent device. Some machines are now able to learn by themselves, processing the complex data provided by their sensors. Something else: that learning may not even be supervised.
Dear @Veerendra K Rai,
actually there is no need to reduce a human to the level of a computer.
We are able to emulate the particle behaviour of atoms and molecules. As the human body consists of those, a human body can be simulated => Postulation.
q.e.d.
I am pretty sure that this is not the way self conscious programs will be written, it is just a proof that it is possible. Claiming "a computer cannot do it" is not a proof for anything. Computers are actually able to generate postulations.
Creating postulations is actually pretty simple (assuming the computer is able to perform logical operations), and obiously way easier than proving them.
Dear @Daniel Knuttel,
Actually reduction is happening. We can emulate behavior of atoms and molecules and since humans are made of these molecules & atoms hence humans can also be emulated. This is reductionist approach- reducing a human to an aggregate of atoms and molecules.
Dear Veerendra K Rai,
Reduction implicates that something has been omitted. Please explain what I have omitted. Please also add a proof that the things I have omitted do exist.
As I said, creating postulations is easier than proving them.
I agree with Veerendra K Rai and Kirk MacGregor.
Dear Daniel, yes indeed the statement "we are able to emulate the particle behavior of atoms and molecules" is not true. Moreover the "atom" and "molecule" itself is just a term, a definition of something we do not fully understand. We are very far from understanding the nature of things that is why we constantly apply reductionism, then slowly we expand it..... but I am afraid that the expansion must be infinite to arrive at an answer.
In my own opinion, AI can only person jobs assigned to it. The thinking is from human not from the robot.
if Alan Turin were alive today, it would be interesting to know if his question, can machines think, has been answered to his satisfaction. The evidence is it has not and perhaps never will be.
Yes, basic logical thing already available. If thinking you mean choosing of right answer so artificial intelligent in some cases better than real.
About Big tata, nowadays we rely on computer or mobile phones mostly.we use a lot kind of App to help our life,reading news,navigating driving,banking, shopping,paying and gathering...
All App have collected our personal information.They know us better than ourselves.they know our saving, favorite shopping,political tendenvy,favorite news...
we are following AI thought to live.
AI can only the configured to act in particular way or perform certain tasks. It does not have its on mind.
Valuable view points by respected colleagues on this discussion
Logical thinking is equal to logical calculating or logical reckoning.A App.
has collected hundreds or thousands most clever intelligent thinking ways.It's database has gathered all information passing the website,and the website can share information with others.
So AI is becoming more and more intelligent.
Artificial intelligence methods can
1. Learn from her self-experience (Learning Machine)
2. Approximate uncertain data (universal neural network approximation)
3. Decide for any NP Problem (Turing Machine)
4. Approximate NP-Hard Problem (Evolutionary Algorithm)
...
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training