The Hype of AI
AI as a technology is way over its promise. The AI community needs to tamper down these extreme predictions because they are pure fiction. Having studied and followed this area of computer programming, I have to remind people the history of AI and the many failures in the past. The few successes such as Watson from IBM is not an intelligent system as we know but just a super fast computer making computations of all possible moves...
- Mar. 2017
What is AI?
Let me first define what is artificial intelligence. AI is the ability of a computer system to interact with humans seamlessly. The original turing test as describe by Alan Turing, specify that if a system can fool the user in thinking he or she is interacting with a human like machine, then the machine can be said is "intelligent."
So far, no system has been able to pass that test with satisfaction.
Why is it so hard? After 50 years of research in this area, we cannot duplicate a child of 10 years old. The answer may surprise you. Humans are not predictable 100% of the time. Think about it, if we were, we can easily program a machine to replicate a human reaction based on prior knowledge. Since, we are unpredictable, the problem becomes huristic, meaning it is something that is learned by experience. So far, we have not been able to develop a system that can learn on its own. We have intelligent systems, which can search for a pattern when givien a large set of data, but that is not intelligence. What is intelligent is the ability to learn and to apply that knowledge to new and unrelated fields. So far, only human brain can do that.
It is the trait of humans to invent new devices, to ask questions, to look for better solutions and to think in abstract and create art.
Here is an analogy. Suppose you create a box thst can translate English to Chinese and back. A person speaking English only can communicate with a person speaking Chinese using this device. The question is this. Is that box "intelligent" or just a tool or utility?
A human translator can do exactly what the box does. Yet, it can do much more. It can recognize if the original speaker is not making any sense and question the speaker, are you sure you want to say this? A box that translate will not be able to do that. It will feed data in and out regardless of the content.
Another attribute of humans is that we are each unique down to our finger prints. Our brains are different and even though we may all receive the same training, we are very different as individuals. Even identical twins have some slight differences.
A computer program, no matter how sophisticated, cannot be unique. It is designed and manufactured and programmed in one way only. Suppose you could build a robot and give it a unique serial number, but that in it self does not make it any more unique. It is identical in every way to another robot built by the same blueprint. In order to make it unique, you will need to create something similar to our genes. The combining of genetic codes from two sources to create a new unique entity. This new entity will have various traits, inherited from both parents. How can an AI system do this? A big question mark.
In conclusion, we have a long way to go in creating a human replacement. It is a much harder problem than people thought or realize. You would need to be a "God" of some sort in order to create a human like entity with all our attributes and ability.
Some Related Info
- Google's DeepMind AI learns like a human to overcome "catastrophic forgetting"
Forgetfulness is a major flaw in artificial intelligence, but researchers have just had a breakthrough in getting 'thinking' computer systems to remember.