Date: 2009-08-22

AI - GENERAL

Without any doubt, the biggest challenge in the development of robotics remains the improvement of AI.

Confirming this intuitive thought, take the most popular type of robots, the humanoid / android robots. Today’s androids are becoming capable of making more and more human-like movements with each passing month, when a humanoid robot gets upgraded for instance or a new model becomes available. Just about 10 years ago, no robot was able to walk in bi-pedal motion, which is basically walking on two feet, but today they can climb stairs, run in circles, avoid moving obstacles while walking towards a specific target etc… Last year, in Holland, a research scientist introduced a new walking mechanism with a robot called flame, which is able to mimic the human walking better than any other robot on the planet and it looks very real. Recently Toyota’s android made the news by running at a speed of 7km/hr. The big dog robot, from Boston dynamics, is able to imitate four legged movement, like no other robot and causes everyone’s jaw to drop. These examples will only get more and better as we progress. There is no doubt that this improving trend is going to continue, and robots will be capable of moving faster, imitate the biological beings better and after a while perform even in a superior way in terms of specified physical movements or tasks…

However when it comes to coordinating these movements with a deliberate autonomous controlling mechanism, things change. The action of thinking like human and then performing actions, is far more complicated than imitating individual physical movements. Based on this, the challenging step is to replicate the human level AI. In order to understand what we are faced with, first define what is being discussed here as “AI”.

The philosophical side and definition

There is of course the philosophical side to this subject and some experts argue that, no matter how complicated artificial entity you create, technically it is never alive and never aware of itself, therefore the real AI is not possible. That is true, but we are talking about a functionality here. As long as a thinking mechanism can practically think, reason, solve problems, recognize patterns, organize concepts in a logical manner like a human, or even better, it means we achieved AI. Another view that has found some acceptance in the past was that if we were intelligent enough to discover how our brain worked, then it would mean our brain would have to be a more complicated organ than it is now, but in that case we would not be able to discover this more complicated organ and so on… I don’t think this paradoxical point of view has any scientific value nor it is based on any evidence or scientific fact. So once again, we are talking about functionality. Think about the famous Turing test, which states that, to an unknowing person, if a machine’s thinking process and therefore conversation is indistinguishable from a human, then it means the artificial entity has achieved the human level intelligence, from the practical point of view.

The progress of AI in the first 40 years of computers and AI winter

Starting with the use of computers in 1950s, the creation of AI has become a major subject of interest. After witnessing the improvement rate of computers, in 1960s, one famous view was that the computers would be able to achieve human level intelligence in 20 years or so... When we look at the history of AI, we see many failures in comparison to the early goals of strong AI. Through 1960s to 1990s, we witnessed many unsuccessful early ambitious projects, such as the failure of machine translation in 1966 (by the way this concept has improved a great deal in our time, and even on the internet you can find many sites that provide “close to human” translation), DARPA’s disappointment with speech understanding Research program in 1970s, disappearance of Japan’s ambitious fifth generation computer project in 1990s. All of these failures, cutting of research funds, public and academicals frustration with the early development and progress of AI, is sometimes called the “AI winter”. Whether this period has ended or still continues is sometimes a matter of debate among experts.

The mistake that the early AI predictors made was, that they believed, ever improving performance of computers will naturally enable human level intelligence… In our time, it has become clear that, the developing of computer hardware alone is not the sole criteria of creating artificial intelligence, without the necessary algorithms that can create a thinking mechanism like humans. There is of course a machine, that enables human level intelligence. And that machine is called the brain. Therefore the biggest hope we have if we want to achieve the real human level intelligence artificially is to discover how brain works first.

Reverse engineering of brain

In recent years, the improvements in brain scanning, and the ever increasing number of institutions and research labs working on brain research, have naturally increased our knowledge base about this most complicated organ of our bodies. In 2008, there were more than 60,000 people in the world working on just brain research. Medicine is now entering the same course of IT, and is becoming an IT field, meaning, it is now progressing at almost the same accelerating rate as the IT. Within a decade or so, our knowledge of the brain will be multiplied by at least several factors of today’s knowledge. Again, when it comes to progress, the biggest mistake to do is to think of the progress linearly, though in fact it is exponential. In order to see an example and understand this trend, see this article that describes accelerating progress in a great detail. It is now being argued by some experts that, because of the geometrical progress rate, ai will reach human level intelligence in 20 years… Another prediction that came recently, argues that we will be able to replicate human brain artificially in about 10 years.

Of course, it is not only brain research that will enable to complete reverse engineering of the brain. First of all, when we “engineer” human body, just like engineered auto parts, we can make things better and better artificially, without being limited to the boundaries of biology. Apart from brain research, a lot of progress is being achieved in medicine, at cellular research, genetics, neurobiology and so on… Even nanotechnology is gradually being a contributor to the engineering of human body. All these progress in other fields will also contribute to the brain research and engineering.

Someone might say, it is not enough just to imitate something, in order to make it work. By reverse engineering of the brain, I do not mean just making an exact physical copy of course. Ever increasing knowledge of brain thinking mechanisms will also enable us to replicate not only how brain is made, but also how it works. As an example, see this ambitious artificial intelligence project that IBM udertook recently.

The present situation of AI

In later articles, I will try to summarize the most well known concepts that present artificial intelligence technology is widely used, and go into greater detail for some.... So keep on reading…

Article By: A. Tuter

(c) Robokingdom LLC. Copying or republishing is prohibited without permission from Robokingdom LLC. See our Terms and Conditions, Use of Content section.

Back to main page