How to feed AI (end)
In this post, I would like to consider only typical mistakes when thinking about where to start making AI. The topic does not pretend to be true in the last resort, only tries to point out obvious errors.
By AI, I mean a program that can perform diverse tasks, leading them to the same tasks that they already know how to solve.
A person receives through vision millions of bits per second, touch gives a person hundreds of thousands of bits per second, and through hearing, a person receives tens of thousands of bits per second.
The remaining channels of sensations provide an incomparably smaller amount of information, are very difficult to implement, slow and have no ability to communicate (it is hardly possible to build AI based on smell).
Let me explain: human children raised in a flock of wild animals have a markedly lower level of development, as a rule they cannot speak. People raised in urban settings are more likely to have a high IQ than villagers. From this we can make the assumption that AI for its normal development will be very gluttonous to information full of meaning.
For example, just reading books can hardly compare with communicating with people, because books do not provide feedback. A more subtle example: inscriptions in foreign languages: even after reading them a hundred times, it is impossible to understand their meaning. Although if you try to stroke the snake, and your native guide will scream something, you will quickly understand what exactly he meant.
It is desirable, of course, in real time. Any delay reduces the chance to draw an analogy and relate two events. Especially if the intervals are different. This rule is not necessary; rather, it simply follows from the previous two. Since information can have feedback and be extensive only if it is relevant. Of course, you can communicate with letters the size of a couple of gigabytes once a day, with a conscious interlocutor, but it will be extremely difficult to identify the relationship with honey through specific actions.
In my opinion, these rules are common to any AI, both algorithmic and neural network. For both real and consciousness-in-matrix.
Based on the theory, it can be assumed that AI will need information in the amount of at least 1% -10% of what a person receives (The conclusion is based on the fact that blind people can still think). If there is less information, then the AI will most likely not be intelligent enough (a blind-deaf-dumb child, broken by paralysis and without tactile sensations, is unlikely to become a full-fledged person). Also, AI will have to be able to get an answer to almost all of its actions (As an example, we can cite the fact that a person receives an extensive report on his own movements many times per second using touch). A delay in responses can complicate the identification of relationships and reduce the speed of learning, so it should be within seconds or minutes.
Consider the theoretical developments of the participants of the Habr:
OpenMinded - Create AI based vector graphics. Since there is no vector graphics in the real world, this AI will have serious difficulties to adapt to reality, if at all able to do it. As for the flow of information, for training such an AI, perhaps hundreds of people are required to provide at least megabits per second, each of which will have to quickly draw an answer when receiving a message from the AI. And at that pace, the years until AI progresses a bit.
P.S. It was this topic that made me write this post.
The most common option, of course, is text. They wrote about him:
alizar - Algorithmic chatbot Suzette
krestjaninoff - Algorithmic analysis of morphology
gorcer- Algorithmic chatbot Marishko
pablozer - Program for writing an algorithmic AI
The main problem of textual AI is information poverty. To feed AI chat, you need to communicate with approximately a hundred thousand people at a time. Of course, you can feed him the entire textual heritage of mankind, but this is just a static text, you can teach him the program to make sentences, but not understand their meaning.
Chess AI, about which Latobco wrote :
This AI has the same problems with information saturation, every 2 minutes, 300 thousand opponents are needed all the time. Not to mention that the world in which AI exists is extremely limited in the number of available options for action.
You can give the required amount of information:
Video - if you, of course, can process such a stream of information, provide feedback, and give it some meaning. Wallpaper in flower Your AI is unlikely to inspire,
Tactile sensations - if you can build a tactile sensor,
Sound - there probably also have their difficulties,
Or any other constant, fast, extensive source that changes from answers.
Before you start creating AI, think carefully about whether you can provide AI information in your own way.
Vision - 10 ^ 6 bits / sec.
Touch - 10 ^ 5 bits / sec.
Hearing - 10 ^ 4 bits / sec.
The first 3 years of life (12 hours of sleep) - 5.25 * 10 ^ 13 bits.
Blind child, the first 3 years - 5.2 * 10 ^ 12 bit = 650 Gigabytes of information
About approximately the whole amount of AI needs to be fed in order for it to have some chance of self-awareness.
P.S. If someone has any ideas, additions, or you notice a mistake - you are welcome to discuss
The ideas and complements of the collective mind follow below.
(Group of questions: Man: "his quote ..." my subjective opinion):
igudym:
“Already structured information of a much smaller volume arrives at the level of thinking.”
Yes, much smaller amounts of information will reach the deeper layers of artificial intelligence, but the ability to systematize information and filter out excess is necessary.
OpenMinded:
“The main thing is that the information is logical and consistent, otherwise there will be indigestion.”
Yes, for artificial intelligence, the saturation of information, its complexity is also important.
Unnamed:
"Intelligence is a property of social organisms. “Adyn thing” cannot be created. Only a community. ”
Artificial intelligence needs an environment with adequate, by the level of development, interlocutors.
dersd:
“The most primitive way of learning is solving problems to eliminate discomfort at the current time.”
Yes, artificial intelligence must have a certain set of organs (able to feel = sensors) that can feel harmful effects (discomfort), avoiding which is the simplest action .
twdragon:
“So the AI will have to not only train, but also be conditioned, supplying a mixture of useful information with“ water ”to the input in order to train him to concentrate resources to solve a particular problem.”
Yes, artificial intelligence does not need all of the incoming signals critically, and many of them can be replaced with “water”, noise, but you need to carefully consider how to do this, and leave the AI an opportunity to concentrate on the details that interest him.
OpenMinded:
“The laws of the world are much more important than detailing its objects.”
Biga:
“I believe that the creation of AI is possible without robotics and without simulating the real world, but it will be necessary to communicate with such AI in a non-human language. “
Yes, artificial intelligence can understand the laws of the physical world on the basis of a simplified model of reality such as the 3D world, but this is not enough to learn to speak with humans on an equal footing.
Unnamed:
“Personally, I would generally extend the concept of intelligence to the entire nervous system, to the smallest nerve nodes and endings. And I would also include the chemical subsystems of regulation. ”
Artificial intelligence is not only a square neural network, but all neurons in general either participating in the processing or transmission of information. Systems that can have any sufficient effect on the neural network can also be considered parts of it (for example, humoral regulation).
OpenMinded:
“Complex environment -> complex mental model -> complex behavior.”
Artificial intelligence, to be complex enough, must exist in a complex world and be complex enough in structure to have complex behavior.
grokinn:
“Olga Ivanovna Skorokhodova was a deaf-blind woman, while a scientist and a writer ...”
Despite this, intelligence was formed before her loss of basic feelings (5 years), in addition, she was one of the few disabled people who became famous. Such a rare exception rather proves the rule of the need for a huge amount of information for consciousness.
mechmind:
“Pheromone ants ensure the coordinated work of their many thousands of“ communities. ”
klakhman:
“ Insect communication, for example, occurs in a very large part due to secreted enzymes ... ”
Yes, odor-based artificial intelligence is possible in theory,
but not feasible with our level of technology and due to the lack of possible interlocutors.
I thank everyone who expressed their opinion.
I will wait with interest for response articles from twdragon about metaprints and from ami.
By AI, I mean a program that can perform diverse tasks, leading them to the same tasks that they already know how to solve.
Bit of theory
A person receives through vision millions of bits per second, touch gives a person hundreds of thousands of bits per second, and through hearing, a person receives tens of thousands of bits per second.
The remaining channels of sensations provide an incomparably smaller amount of information, are very difficult to implement, slow and have no ability to communicate (it is hardly possible to build AI based on smell).
Argument One: AI Must Have Access to Content
Let me explain: human children raised in a flock of wild animals have a markedly lower level of development, as a rule they cannot speak. People raised in urban settings are more likely to have a high IQ than villagers. From this we can make the assumption that AI for its normal development will be very gluttonous to information full of meaning.
Argument Two: AI Must Have Feedback
For example, just reading books can hardly compare with communicating with people, because books do not provide feedback. A more subtle example: inscriptions in foreign languages: even after reading them a hundred times, it is impossible to understand their meaning. Although if you try to stroke the snake, and your native guide will scream something, you will quickly understand what exactly he meant.
Argument Three: AI Must Have Access to Timely Information
It is desirable, of course, in real time. Any delay reduces the chance to draw an analogy and relate two events. Especially if the intervals are different. This rule is not necessary; rather, it simply follows from the previous two. Since information can have feedback and be extensive only if it is relevant. Of course, you can communicate with letters the size of a couple of gigabytes once a day, with a conscious interlocutor, but it will be extremely difficult to identify the relationship with honey through specific actions.
In my opinion, these rules are common to any AI, both algorithmic and neural network. For both real and consciousness-in-matrix.
A bit of practice
Based on the theory, it can be assumed that AI will need information in the amount of at least 1% -10% of what a person receives (The conclusion is based on the fact that blind people can still think). If there is less information, then the AI will most likely not be intelligent enough (a blind-deaf-dumb child, broken by paralysis and without tactile sensations, is unlikely to become a full-fledged person). Also, AI will have to be able to get an answer to almost all of its actions (As an example, we can cite the fact that a person receives an extensive report on his own movements many times per second using touch). A delay in responses can complicate the identification of relationships and reduce the speed of learning, so it should be within seconds or minutes.
Consider the theoretical developments of the participants of the Habr:
OpenMinded - Create AI based vector graphics. Since there is no vector graphics in the real world, this AI will have serious difficulties to adapt to reality, if at all able to do it. As for the flow of information, for training such an AI, perhaps hundreds of people are required to provide at least megabits per second, each of which will have to quickly draw an answer when receiving a message from the AI. And at that pace, the years until AI progresses a bit.
P.S. It was this topic that made me write this post.
The most common option, of course, is text. They wrote about him:
alizar - Algorithmic chatbot Suzette
krestjaninoff - Algorithmic analysis of morphology
gorcer- Algorithmic chatbot Marishko
pablozer - Program for writing an algorithmic AI
The main problem of textual AI is information poverty. To feed AI chat, you need to communicate with approximately a hundred thousand people at a time. Of course, you can feed him the entire textual heritage of mankind, but this is just a static text, you can teach him the program to make sentences, but not understand their meaning.
Chess AI, about which Latobco wrote :
This AI has the same problems with information saturation, every 2 minutes, 300 thousand opponents are needed all the time. Not to mention that the world in which AI exists is extremely limited in the number of available options for action.
A little moralizing
You can give the required amount of information:
Video - if you, of course, can process such a stream of information, provide feedback, and give it some meaning. Wallpaper in flower Your AI is unlikely to inspire,
Tactile sensations - if you can build a tactile sensor,
Sound - there probably also have their difficulties,
Or any other constant, fast, extensive source that changes from answers.
conclusions
Before you start creating AI, think carefully about whether you can provide AI information in your own way.
In numbers
Vision - 10 ^ 6 bits / sec.
Touch - 10 ^ 5 bits / sec.
Hearing - 10 ^ 4 bits / sec.
The first 3 years of life (12 hours of sleep) - 5.25 * 10 ^ 13 bits.
Blind child, the first 3 years - 5.2 * 10 ^ 12 bit = 650 Gigabytes of information
About approximately the whole amount of AI needs to be fed in order for it to have some chance of self-awareness.
P.S. If someone has any ideas, additions, or you notice a mistake - you are welcome to discuss
UPD3
The ideas and complements of the collective mind follow below.
(Group of questions: Man: "his quote ..." my subjective opinion):
Input Information Structure:
igudym:
“Already structured information of a much smaller volume arrives at the level of thinking.”
Yes, much smaller amounts of information will reach the deeper layers of artificial intelligence, but the ability to systematize information and filter out excess is necessary.
OpenMinded:
“The main thing is that the information is logical and consistent, otherwise there will be indigestion.”
Yes, for artificial intelligence, the saturation of information, its complexity is also important.
Education and the world:
Unnamed:
"Intelligence is a property of social organisms. “Adyn thing” cannot be created. Only a community. ”
Artificial intelligence needs an environment with adequate, by the level of development, interlocutors.
dersd:
“The most primitive way of learning is solving problems to eliminate discomfort at the current time.”
Yes, artificial intelligence must have a certain set of organs (able to feel = sensors) that can feel harmful effects (discomfort), avoiding which is the simplest action .
twdragon:
“So the AI will have to not only train, but also be conditioned, supplying a mixture of useful information with“ water ”to the input in order to train him to concentrate resources to solve a particular problem.”
Yes, artificial intelligence does not need all of the incoming signals critically, and many of them can be replaced with “water”, noise, but you need to carefully consider how to do this, and leave the AI an opportunity to concentrate on the details that interest him.
OpenMinded:
“The laws of the world are much more important than detailing its objects.”
Biga:
“I believe that the creation of AI is possible without robotics and without simulating the real world, but it will be necessary to communicate with such AI in a non-human language. “
Yes, artificial intelligence can understand the laws of the physical world on the basis of a simplified model of reality such as the 3D world, but this is not enough to learn to speak with humans on an equal footing.
Internal structure:
Unnamed:
“Personally, I would generally extend the concept of intelligence to the entire nervous system, to the smallest nerve nodes and endings. And I would also include the chemical subsystems of regulation. ”
Artificial intelligence is not only a square neural network, but all neurons in general either participating in the processing or transmission of information. Systems that can have any sufficient effect on the neural network can also be considered parts of it (for example, humoral regulation).
OpenMinded:
“Complex environment -> complex mental model -> complex behavior.”
Artificial intelligence, to be complex enough, must exist in a complex world and be complex enough in structure to have complex behavior.
Other:
grokinn:
“Olga Ivanovna Skorokhodova was a deaf-blind woman, while a scientist and a writer ...”
Despite this, intelligence was formed before her loss of basic feelings (5 years), in addition, she was one of the few disabled people who became famous. Such a rare exception rather proves the rule of the need for a huge amount of information for consciousness.
mechmind:
“Pheromone ants ensure the coordinated work of their many thousands of“ communities. ”
klakhman:
“ Insect communication, for example, occurs in a very large part due to secreted enzymes ... ”
Yes, odor-based artificial intelligence is possible in theory,
but not feasible with our level of technology and due to the lack of possible interlocutors.
End
I thank everyone who expressed their opinion.
I will wait with interest for response articles from twdragon about metaprints and from ami.