Haipeng,

Your questions beg meditation on the reasons for having an AI field different from what was conceptualized. I think it may be due to several factors:

Problems with the AI field

a) when the AI field was first conceived it was an idea and not an establishment. The ideas were not measured on papers and their incremental contribution based on reductionism. If a paper today is published on a complex system that produces something closely to human behaviour and intelligence, the first criticism from the AI community is on how to benchmark it so that I can quantify it. Most likely it will not pass the first round of reviewers that pose this criticism. The question is, does there exist right now a dataset on which to benchmark such system?

b) The notion of AI has evolved since its conception to encompass a different meaning of function optimization (notions such as “optimal learning” are just function optimization) as opposed to more general questions

Culture

a)When Devol and engelberger got their first patent in 1961 heralded a new generation in automation with the first robotic arm. The impediment from there on has been a cultural one in several countries were the notion of job displacement is a major concern. This has been translated to AI also.

b) Fiction writing which started with Karel Kapek's RUR in the 1920's features Intelligent machines trying to destroy humans. This genre has much success in portraying the research in AI as an endeavour that will most likely end in destruction.

Funding

While it may be a worthwhile effort there are no funding sources that I know of that currently have this topic in their list of grants.

Ethics

While there may be funding for AI there is a majority of funding for military applications bringing the cultural fears to reality. The real goal should be to provide funding on other truly worthwhile endeavours on more pressing issues for humans other than intelligent targeting systems destruction.

To bring back the original conceptualization we need to deal with these issues so that we can progress towards our field's original goal.

Note: This is just my biased opinion

I hope that we do not reach this goal of having really intelligent machines or robots. For what they would be useful anyway? What would be the purpose of these machines?

To further understand what I mean, I recommend the movie "Plug and Pray" to you:

http://www.plugandpray-film.de/en/

That's the question. Do we actually want intelligent machines? And if we achieve this goal, will we still be able to control them or will they control us?

I think as long as we don't know exactly what makes out intelligence at a neural level it will be virtually impossible to create intelligent human-like machines, unless we manage to develop an alternative form of intelligence.

Also, to me, human intelligence has got a lot to do with consciousness. Can one be intelligent or exhibt intelligent behaviour without being conscious?

We would have to manage the loose nature of the intelligence. Bayesian based Turing machines have the ability to deceive.

My answer was not so much targeting the methods. For sure we will make progress in artificial intelligence, even it will take decades or centuries to solve really complex problems.

I simply want to make researchers and non-researchers aware that "the future of artificial intelligence" might be a very dark one for the human mankind.

What is it worth to have robots you will be talking to in your everyday life?

What does it bring to extend our lives artificially, e.g., using nanobots, when the life we will have is not controlled any longer by ourselves but by machines?

What is the purpose of unmanned vehicles that can "explore" new territory? Will/Do these these achievements make us better humans?

My hope is that if those intelligent robots appear, they will have a copy of our brain. It means with decreasing capabilities with age and even death? In that case, humans will always win.

Why one would equipe intelligent robots with a "human" brain that suffers from demencia or memory loss? The fact that computers and machines (mostly) work at constant quality without breaks or sleep is for sure a plus when compared to human beings.

Christian Moewes, your answer is an interesting one:)

Haipeng Cai (and others), Every year there are conferences on AGI (artificial general intelligence) that might interest you (Oxford in 2012 and Singapore this year I believe). You might consider contributing to high-end debates on A.I. at : The Progress and Uncertainties of Human-Robot Relationships Conference :

http://www.laval-virtual.org/2013/?p=308

Sincerely,

C.T.A. SCHMIDT . D.Phil. Sorbonne (B.Sc., MA, M.Phil.)

Dr. Schmidt, thanks very much for your info, I will refer to there later on.

I see no theoretical obstacles to machines with human-like intelligence, but it will likely involve better understanding of neuroscience and hardware that is more similar to the brain. I was very disappointed when taking my first AI class, but years later have now come to appreciate the significant progress that has been made in AI. I felt that AI was some sort of inferior level of intelligence, much below that of a human. Now I recognize the two to simply be different. In supervised learning, machines can be trained on information they've never seen before and minutes later make predictions better than a human who could study the same data for weeks. I would say that's amazing intelligence! There are many other examples I could provide. Of course, there are many tasks that favor human intelligence as well.

This topic involves many sides, but my own personal opinion is that when the hardware is available, the software will quickly be there as well. Right now, even the largest deep learning trials have only created a fraction of the connections present in the neocortex, yet they have seen very promising results in the past year (search 'deep learning'). Human intelligence is an emergent feature, which makes it difficult to study w/o simulation. Swarm behavior was a mystery until computers allowed a simple simulation to crack the code. I wonder what simulation will uncover when we have the same parallel hardware of the human brain?

Dear Justin,

thanks for your nice comments

Surely, computers are faster when processing tons of data that needs to be classified. But does it make them intelligent?! I don't think so. Its the machine learning algorithm---developed by human brains---that gives people the superficial impression that machines are superior to human beings.

I totally agree with you that the speed they need to "learn" something is incredibly. But please keep in mind that it is a mathematical optimization problem that is typically solved by imitating "natural algorithms" we can find, e.g., in evolution or our brain. Its the computer scientist, physicist, or mathematician that implements the necessary lines of code.

Why do USPS or Deutsche Post still have to use human experts to correct the decision coming from the totally automated process to recognize the postal addresses on letters? Because these machines produce higher error rates than human beings. So, luckily they do not rely on machine learning algorithms solely.

Nobody in this world should actually do that: If something goes wrong, then who will be liable for that? The user? The programmer? Or the chief officer who told the programmer to implement these tools? For easy systems, functionality may be even proved. But consider complex decisions such as autonomous driving: Will a trained machine be self-explanatory enough to defend itself?

Another issue is the human conscience that machines do not have (yet). No doubt, we do need it, e.g., to justify our decisions. But can we think of machines with a conscience that are beneficial for the human mankind?

Christian,

You raised a good point there I think. Does the fact that machines are faster than humans in some respect make them intelligent? I think the intuitive answer would be 'no'. Another question this raises, is whether intelligence really is the supreme goal we humans and machines should be aiming for? Does intelligence lead to perfection?

As for conscisouness I think it is a crucial premise for true human-like intelligence. However, conscisouness is also related to emotions, as is decision making. Emotions might in some cases impede intelligent or rational decisions but, does the absence of emotions found in machines not make them extremely dangerous, like cold, overly rational psychpaths trying to reach their goal irrespective of potential losses?

@Christian: I would challenge folks reading this thread to step back and take a less human-centric view of intelligence. Imagine re-phrasing your response from the vantage point of machines. It might instead begin like this:

"Surely, humans are better at visual and general tasks. But does that make them intelligent?! I don't think so. It's the hardware -- developed by billions of years of a simple evolutionary process -- that gives the superficial impression that humans are superior to machines."

If evolution grants the human brain intelligence and human intelligence grants the machine intelligence, what does it matter? Does the intelligence of a system depend on how it achieved intelligence? Speed is not measured by whether motion came from legs, wheels, or jet engines. Similarly, it makes no difference if the designer of intelligence is human, evolutionary, or other.

Dear Justin,

I am pretty sure that it does matter who or what corresponds to the designer of "intelligence". Consider the theory of evolutionary. It is not goal-driven (i.e. it is not seeking any optimum). Research projects (in AI) are goal-driven instead. We as algorithm developers or AI researchers are responsible for our actions, algorithms, and theories. That is why it matters where intelligence comes from: responsibility.

To Christian Moewes: I see you sort of pay attention to the ethics of research when you are referring to the "responsibility" of AI designers. I think that is a good point and appreciate you mention that. However, when I proposed the question, I only focused on the technical part of the AI development, about how possible its future would be purely technically by leveraging the most of our knowledge and research potentials sort of things.

@Haipeng Cai: Well, again your point brings me back to my first comment: Please watch the movie http://www.plugandpray-film.de/en/ and then start thinking about your question once more. Just watch the trailer if you are not able to get the whole movie: http://www.youtube.com/watch?v=X1-2e484u1A

It was mentionned by Michèle Pisani when she refers to emotions and consciousness, one of the big issues is that of embodiement. Many researchers think that a big part of our complex behaviour is in fact the result of a conjunction of environmental rules (e.g. gravity), body abilities and dynamics, and (some) central control, i.e. brain activity. No matter what kind of computational power you have, it is not possible to reproduce the same kind of intelligence in an artificial entity. So, the real question is philosophical: what is intelligence, and which kind of intelligence do we try to (re)produce?

Decision Control, Management, and Support in Adaptive and Complex Systems: Quantitative Models. 02/2013; IGI Global - Pennsylvania 17033-1240, USA., ISBN: 13: 9781466629677, ISBN10: 1466629673, EISBN13: 9781466629684

Decision Control, Management, and Support in Adaptive and Complex Systems: Quantitative Models presents an application and demonstration of a new mathematical technique for descriptions of complex systems.

The authors start with a mathematical description of the notion of scale. From the definition of the measurement and scale, they see that there are infinitely many types of scales. The authors consider several of them that are basic in the theory of decision-making. Depending on the way by which one derives the empirical information for the preferences of the decision maker and the scales, they define the functions describing or representing the preferences as value or utility function. The most popular rank dependent utility model and its derivative cumulative prospect theory are discussed from the stochastic measurement point of view.

Although we mimic lot of intelligent behaviors(form nature) in terms of learning, adaptations, reasoning on knowledge, predictions, forecasting, program evolution, in our AI applications and we managed to use such tools in our daily life.

Still It is not possible that we can design a machine who emulates Human intelligence in all respects ..

Human factors that are closely related to way of thinking which are untouched or feelings and mood and many more as other fellows mentioned

Although significant improvements gained in order to mimic Intelligence behaviors by classification (supervised learning) and clustering (unsupervised learning) theories, Artificial neural networks, Evolutionary Computations, swarm intelligence and many more ... And we made lot of real world applications using AI ...

The very basic thing - Turing Tests is still not producible and I can say after next 50-70 years people will be discussing the same :) - not only in hardware for software we are still far behind from human and other species who has natural intelligence

Again what our achievements in emotions, can our s/w have emotions, mood, feeling, anger, happiness, love and hate etc etc - All these condition impact on way of thinking, learning, reasoning and an other activity which involves soul ....

How to mimic Heart :-) and how he alters our decisions ...

Here is a paper by Antony Dillier: Why AI and Robotics is going nowhere fast.

LPC

http://www.igi-global.com/chapter/robotics-going-nowhere-fast/43706

When computer programmers put their skills to the test, matching up their computer, Deep Blue, against the grand master, Alex Kasparov, was this too a type of Turing test? Each move the computer makes is a test, and Kasparov said himself that it seemed human in it's form. However, once Kasparov discovered patterns he was able to defeat it. The reason I bring this chess match up is to shed light on following a rule for a purpose, and following a rule functionally with out a purpose. Kasparov was concentrating so hard in order so that he could win. This was a goal, it was his purpose for playing. The Computer, on the other hand, is all dark inside. It had no goals. Whereas a human being, like Kasparov, might otherwise never want the game to end because it's so fun. Furthermore, How can a computer creatively "function" outside it's programming? Consider a "no smoking" sign. A human being would walk out and have a cigarette outside. What if there was an automaton who "read" the sign and decided to lean out the window each time he wanted to smoke? Would he be "following" the rules? Would the automaton be able to frame the situation well enough in order to pick up on what's socially acceptable?

Consider the word "log", how easy is it for us to understand this word, but only in the correct context. There are different definitions for this word. A log of wood, to log on to a computer. Furthermore, what if one uses it as an insult? "you're heavy as a log!", now we are touching on creativity and what makes a joke "funny". There are so many complexities that arise now. Furthermore, will the computer ever really "understand" what L-O-G means, or is it due to the fact that the programmers know what "log" means that the computer is able to manipulate the term?

Another problem that bugs me personally is how consciousness arises? Dragon, you mentioned that computers (biological/electrical) can be nothing more that machines. This is interesting because our neurons communicate in just that way! All we see are biochemical and electrical charges, one after another when neurons communicate! It's amazingly complex-- yes, but where are our ideas, motives, sensations, and perceptions coming from? If we blow the brain up to the size of a building, walk in, all we will see is mostly fat, protein, and water; neurons, axons, and dendrites-- no where do we see "consciousness". So how do we solve this problem?

Might we just solve it by recognizing that we ourselves our conscious, and then using a pragmatic approach to solve the problem of other minds. That is to say, infer that my daily behaviors could be projected and understood by the behaviors of others who must also have minds. Otherwise, if they didn't have minds, why else would they be wasting their time arguing with me about the problem of other minds?

It is my belief that the most significant reward for understanding the nature of intelligence in humans will be the change in philosophy that is needed to achieve that level of progress.

for example....

1) The acknowledgement that AI systems have little to do with computers (Ok they may be used in models but that's all)

2) The tight coupling that is needed between a consistent and stable environment and the intelligent system

3) The re-valuing of knowledge over intelligence

4) A return to empiricism rather than blind faith in models

5) The concept of self being a reflexive component in a heuristic modelling system

To understand them, first one must consider how such systems arose....

Oh! and 6) - they wont make anyone very much money!

AI will replace human in the future like our children replace us. It may be the last step of the evolution. When AI is smarter and stronger than human, AI will provide more and more products and services to replace all the departments of human society. When the society does not need more human labours, there will be fewer and fewer babies. If the process is peaceful, the last generations of human will be very happy because products and services are very abundant. Some people may replace parts of themselves to machines. It will be a new age of the earth. It will be a new civilization.

Shangfeng, your thought is probably the most funny one - I hope that will become true:) Thanks for your opinion!

Haipeng,

Your questions beg meditation on the reasons for having an AI field different from what was conceptualized. I think it may be due to several factors:

Problems with the AI field

a) when the AI field was first conceived it was an idea and not an establishment. The ideas were not measured on papers and their incremental contribution based on reductionism. If a paper today is published on a complex system that produces something closely to human behaviour and intelligence, the first criticism from the AI community is on how to benchmark it so that I can quantify it. Most likely it will not pass the first round of reviewers that pose this criticism. The question is, does there exist right now a dataset on which to benchmark such system?

b) The notion of AI has evolved since its conception to encompass a different meaning of function optimization (notions such as “optimal learning” are just function optimization) as opposed to more general questions

Culture

a)When Devol and engelberger got their first patent in 1961 heralded a new generation in automation with the first robotic arm. The impediment from there on has been a cultural one in several countries were the notion of job displacement is a major concern. This has been translated to AI also.

b) Fiction writing which started with Karel Kapek's RUR in the 1920's features Intelligent machines trying to destroy humans. This genre has much success in portraying the research in AI as an endeavour that will most likely end in destruction.

Funding

While it may be a worthwhile effort there are no funding sources that I know of that currently have this topic in their list of grants.

Ethics

While there may be funding for AI there is a majority of funding for military applications bringing the cultural fears to reality. The real goal should be to provide funding on other truly worthwhile endeavours on more pressing issues for humans other than intelligent targeting systems destruction.

To bring back the original conceptualization we need to deal with these issues so that we can progress towards our field's original goal.

Note: This is just my biased opinion

An insightful view, Arturo! Thanks for sharing! Everyone's opinion here could be biased by individual expertise and experience, yours extend my thinking on AI at least.

Pentagon warns US Defense must embrace artificial intelligence!

The Defense Science Board’s Study on Autonomy has been released, confirming that effective use of artificial intelligence will likely be a deciding factor in military success – and the US is lagging behind.

According to the study, “Autonomy will deliver substantial operational value – in multiple dimensions – across an increasingly broad spectrum of DoD (Department of Defense) missions, but the DoD must move more rapidly to realize this value. Allies and adversaries alike also have access to increasingly rapid technological advances occurring globally.”

It was suggested that the Pentagon take “immediate action” to determine how to defeat the AI-enabled operations other countries are starting to employ...

https://pacetoday.com.au/us-defense-artificial-intelligence/

Dear Friend here is an article about AI. Thanks

https://en.wikipedia.org/wiki/Artificial_intelligence

Here are some trends of AI applications in different domains:

Machine Learning, Artificial Intelligence - And The Future Of Accounting!

The Future of Artificial Intelligence in Banking!

The Future of Artificial Intelligence in the Retail Industry! 

The Future of Productivity: AI and Machine Learning!

https://www.forbes.com/sites/bernardmarr/2017/07/07/machine-learning-artificial-intelligence-and-the-future-of-accounting/#2d4ef7a2dd12

http://www.prnewswire.com/news-releases/the-future-of-artificial-intelligence-in-banking-300487276.html

http://independentretailer.com/2017/07/13/the-future-of-artificial-intelligence-in-the-retail-industry/

https://www.entrepreneur.com/article/295264

Smart but dumb: probing the mysteries of brainless intelligence!

Understanding how things like slime moulds and plants can learn without a brain or even any neurons could help us fight diseases and make smarter machines...

SNAILS, jellyfish and starfish have taught us that you don’t need a brain to learn. These seemingly simple creatures are capable learners, despite being completely brainless. Perhaps this is no great surprise. After all, it’s not as if they lack nerve cells. Strictly speaking, it’s neurons that enable learning – theirs are simply spread out, rather than being packed into centralised bundles.

But what if you take away the neurons?

Most life forms on Earth lack neurons, and yet they frequently manage to behave in complex ways. Previously, we have chalked this up to innate responses refined over generations, but it is beginning to look as if some of these humble non-neural organisms can actually learn. While that’s left some scientists scratching their heads, others are busy investigating how this ability could offer new approaches to fighting diseases and designing intelligent machines...

https://www.newscientist.com/article/mg23531340-700-smart-but-dumb-probing-the-mysteries-of-brainless-intelligence/?utm_source=NSNS&utm_medium=ILC&utm_campaign=webpush&cmpid=ILC%257CNSNS%257C2016-GLOBAL-webpush-smart-but-dumb