Computer systems must be aid to humans, but as we get more creative, we delegate as we can. When the decision support system, big decisions might be taken. When the system is actually a solution, we have seen many "stupid" decision made simply because the system to assign their decision algorithms instead of instincts and emotions. Artificial intelligence is often understood as a computerized simulation of human thinking. But as it stands it is based on forms of logic. We find that the employer is not always connected to the most logical solution is as illogical as it sounds! However, this way, so we must find ways to overcome this deficiency with our systems of artificial intelligence (AI).
Maybe we haven't quite broken in the most complicated forms AI. We believe that we can achieve much more. Technology does not grow in a linear form, as we believed in the past. We now have the evidence, based on historical growth, that technology is growing exponentially. This means we can expect to see very human of similar systems in the very near future. AI systems can take the form, we look forward to. For example, while consumers await the neural network with lots of lights and microscopic transistors, researchers have developed a system with a functioning brain, animals in the operation of the installed processor. Much progress has been achieved using this technique artificial. With this study and other discoveries, all of a sudden, the attributes that we thought could never be processed AI suddenly are.
We will not be enough time to discuss every aspect of AI. At least we can do is look at the way they are and to say that some of the ways we can overcome this.
Too much complexity: think about AI systems is currently divided into three parts. The first part will enter (where it is fed information used for decision-making). The second part is that we call the "hidden node" (where the algorithm lives that will handle the information and decide what results will be). The third part of the output (where he organizes and displays its decisions for the end user or system).
The hidden nodes, you can have more than one node. In fact, more hidden nodes you have more complex AI System. However, if follows that if you have too many hidden nodes, complexity can become too high, and there are too many templates for NN school. Who would think your computer might have too many templates for study? It's not so much a problem of processor performance is more of an introduction of inconsistencies in your algorithms, causing several crashes and congestion, especially in such a way that the system does not recognize. So use less hidden nodes for more simple or basic problems.
Remembers: you want your system actually "learn" rather than simply "remember". When he just remembers, she cannot properly analyze new problems or emissions. There is no need for establishing an overly complex systems to compensate for the deficiencies memorization. This problem can be mitigated through post with new information, to see if he learned anything else. Specifying your analysts and programmers to develop algorithms to find ways of classifying a completely new information and store it, but in addition, receive feedback from the output and store (effective learning from mistakes like people)
Fees: if you "teach" system for too long, it can "learn more". This can cause it to "paranoid" and see models out there really doesn't exist. As a result of an error, you can start, an increase of the year. It would be inappropriate for your system, or leave it vulnerable to hacker manipulative tactics. Can be reduced by stopping training and testing to see what time a stop error is reduced. That is why you should spend enough time to develop ways to test your system.
, Local solutions: you can get a solution that is just one local. Therefore, although the solution works, it might not be the optimal solution. This problem can be mitigated by testing solutions received to determine the best.
Limitations: the hard truth is that your system can solve the problems of AI, it's being done to tackle. As a result, some problems cannot be effectively dealt with in the system. The good news is that it can solve the problems that need!
Takeaway message is that the AI system should not be used to replace human. If so, this should be basic or simple in nature. The system should be used to monitor how the AI system examines emissions. Furthermore human review and intervention is absolutely necessary to ensure the effectiveness of the system.
0 comments:
Post a Comment