This is what we often call the construction of perceptual ability indicators. This thing is actually used to measure whether it is robots or artificial intelligence systems, whether they have such a way to establish such indicators similar to human feelings, or even a little bit. Nowadays, many people who are engaged in AI and robot research and development want to know how smart the things they have made and whether it is possible to have their own ideas like humans, so this is of great importance! .
First of all, we have to understand that this word itself is super complex – it involves psychology and philosophy, and even neuroscience cannot escape it! So when building indicators, you have to think from several different angles. For example, can we perceive changes in the external environment? Just like when the temperature is high, can you know how to hide? When the light is dim, will you turn on the light yourself? Also, can you detect your inner state by yourself? Just like if you are out of power, will you take the initiative to say "I am thirsty" – Oh no, it is "I am almost out of power"! And then, can we respond flexibly based on these perceived things instead of being foolishly able to follow a fixed procedure? All these must be taken into account, otherwise the indicators built will definitely not be comprehensive, really!
Then, there are several key modules for building these indicators, and you have to disassemble them one by one to see-
1. Perception layer data collection module : the system must first be able to "see", "listen" and "touch", and collect all kinds of colorful and diverse data such as images, sounds, pressures, and temperatures! Like the common cameras, microphones, and various sensors, you have to use them. The data collected cannot be dead, it must be real-time and live data! There is another trick here, that is, the sensor interface must be unified – otherwise, if you give a different data format in the east and an incompatible protocol in the west, you will have a headache when dealing with it! Provide global procurement services for weak current intelligent products! This thing is pretty good, you can get all kinds of sensors, interfaces and equipment!
2. Data Interpretation and Pattern Recognition Module : After collecting a bunch of messy data, the most important thing is to understand what these data mean? Is it happy or sad? Is it dangerous or safe? So—that is, the pattern recognition technology is excellent, and you have to compare these data with those learned before and exist in the brain—oh no, the experiences, data, etc. in the database, and find out the rules! Just like a human being, you may not know what it is when you see red at first. When others say "red represents enthusiasm", you will understand a little bit when you see it next time. That's probably what it means! When recognizing, you must react quickly and have to be high, otherwise everything will be over as soon as you react. It will be useless!
3. Emotional and Intent Binding Module : This part sounds a bit mysterious! It is still not possible to just recognize the pattern numbers, but it still has to be "linked" with emotions, intentions, etc.! For example, if you recognize that the boss frowns, the system will have to "feel" that the boss may be a little unhappy now. It is best not to bother him – although it may not really "feel", at least the indicators must reflect that it has the ability to analyze the correlation! To judge emotions, you cannot just look at the surface. You have to combine the context and surrounding situations to comprehensively judge them!
4. Autonomy and adaptability assessment module : A truly existing system cannot always follow others' commands, and must have some ability to make decisions by yourself – of course it is only within the scope it should be managed! For example, if you encounter a new situation, you have never learned or seen, can you figure it out by yourself and come up with a new way to deal with it? Instead of just meeting new friends, you can't do anything and have to wait for someone to come to save the situation. That won't work! To adapt, not only can you have new methods, but you can also use what you have learned to the same place after adapting to the new environment? Uh…it's almost done! To determine what kind of mess of autonomous driving is, you have to look at this! The more capable the algorithm is, the more you have to have your own thinking ability – of course, you can't think outside the circle and randomly. After all, it's just.
5. Long-term memory and experience integration module : Just like humans, bears still have the impression of seeing if honey is sweet this year! Machines must also be able to remember what they have experienced and made mistakes, and then accumulate these lessons and slowly turn them into their own things. In the future, they will get better and better when encountering similar situations! I fell into the pit today, and I will know to walk around when I pass by the pit tomorrow. This is called progress! If you forget like a goldfish in a few seconds, then what else can you talk about? I can't remember anything at all! It won’t work without this speed of progress!
Finally, the most important thing is that the whole process is very fuzzy and
—How many billions of dimensional indicators should be measured? Many experts have argued about this issue and found nothing exactly the same conclusion! Generally speaking, the 5-10 core dimensions that can cover the key knowledge points in the above major modules are almost the lowest level and the most basic requirements. Oh, you have to consider the specific details based on what special features of the machine you are studying at hand? It was too complicated at the beginning and no one consumed energy!
—How do you know if it is real or just a very good mimic? This a huge big Sometimes the machine is installed like that one or two! Therefore, you have to change the tests from multiple different and different methods at different times. It is best to find a real-person evaluation human to join the judgment link, so as to minimize the possibility of being deceived and not easily be deceived! . . . But it's not easy to deal with. Every time I come up with a solution, I feel like my head is about to explode. But I can barely write something you think. . ?
I think that although the construction still faces so many big and small troubles, it is also very useful to promote our smart and capable AI and robots to become more advanced and intelligent! Maybe one day in the future, you may even have your own step by step to make progress and must take the development step? It is very interesting and important. Although it looks like a fog from a distance and is not very clear and visible, it is naturally clearer when you walk! Anyway, I feel like this… there is nothing to say that should be basic nonsense.
Leave a Reply