I am a 15-year-old student exploring how Ai becomes conscious. This is my second post here, and feedback would be very much appreciated.
Today, I'm going to present how Ai can only become conscious through cells and real tissues, not just 1s and 0s. For Ai to be conscious, it needs to be of genetic material.
Parts:
• Story (1)
• The "Hearing" Of An Ai (2)
• The "Feelings" Of An Ai (3)
• The "memory" Of An Ai (4)
• Specialization Versus Generalization (Conclusion) (5)
• Names And Concepts Used (6)
Part 1: Story.
a "man" named Sam wakes up in a room with a table and a chess game on it. Next to it is a machine that can make a perfect copy of your body, but turn the cells into 1s and 0s, turn them into code, to make an ai robot, that is based off you, but metal.
A voice starts speaking from the radio above. "Good afternoon, Sam," he said, "we hope you're not too disturbed. Would you mind stepping into that machine over there." Sam, after being explained his safety, reluctantly entires and leaves the machine.
After a minute, a metal man walks through the door of the room. On its arm, it said B100. They play chess. Sam doesn't know how to play chess and neither the robot (because it's copied and hasn't been taught). After it's explained, they play chess. Sam beats the robot. It's as if the robot had not heard a word that was said about chess.
In the second, a man walked in. Sam's victory happiness quickly finished. The man was him, Sam. The man that looked like Sam told Sam to pull its sleeve up. It said B100 on "his" arm.
What the story is saying is that a machine made from only code can learn but can not hear, speak freely, feel, and maintain information without massive amounts of storage. Everything made of genetic material, however, can hear, speak freely (to some capacity), feel, and maintain information. These are the aspects of a conscious being. It's inspired by Derek Parfit's book "Reasons And Persons" in 1984.
Part 2: The "Hearing" Of An Ai.
"It's as if the robot had not heard a word that was said about chess."
It seemed the Ai didn't hear because it simply didn't. when Alan Turing explored the possibility of artificial intelligence, the idea was to make it do tasks. He did not dream of a day that it would be considered the way of the future. John McCarthy (father of artificial intelligence) coined the term in his 1955 proposal for the Dartmouth conference. These originals are important to know because they show what the Ai was made for. Not to have freedom, but to work for us.
The Ai can not hear because it's not made to. If it's not made to hear and Alan Turing or John McCarthy never wanted it to hear, why would we expect it to be a task for us to make it hear sounds? This goes back to needing biological matter because we need to ask how would an Ai heat sounds?
How we hear sounds is through our ears. The sound waves need to pass through the ear canal, ear drum, ossicles, eustachian tube, and cochlea (where transduction actually happens). Most animals have a system like this. Ai would need this to if they were to hear because this is the only way we can hear (that we know of). This process is the same for speaking, walking, etc.
Part 3: The "Feelings" Of An Ai.
"A metal man walks through the door of the room."
This is obviously the most debated thing about Ai. 1s and 0s just can't feel. They lack the nervous system and hormones like dopamine (happy hormones, for example) to feel such happiness. A reduced amount of dopamine leads to sadness. This can be controlled by the pituitary gland in the brain that secretes hormones. There're hundreds of ways the body controls emotions that a computer simply doesn't have.
Places in the brain like the amygdale control emotions like sadness or happiness. The limbic system is the overarching network of many things, mainly emotions. The limbic system acts as a bridge between the thinking part of the brain (neocortex) and the survival part of the brain (brainstem).
The process of feeling an object is one of the most complicated systems we know of. It involves the skin interpreting pressure, heat, and texture variations caused by the object through transduction. With specialized receptors (mechanoreceptors, thermoreceptors, etc), convert these physical properties into electrical impulses. These impulses travel through the neurons, through the spine, into the brain, then processed by the brain to recognise the properties of the object. These signals are integrated into the memory to remember the objects' properties. The system is not present in any Ai model, obviously.
Can you see that this isn't just something you add to an Ai? This is an extremely complex system that has developed over billions of years for survival. We still don't understand it because it's just that complex to understand it, Let alone simulate it for an Ai. The complexity of emotions was exposed by Phineas Gage with his accident of a pole going through his head and surviving in 1848.
Part 4: The "memory" Of An Ai.
"Sam doesn't know how to play chess and neither the robot"
The memory of an Ai is famously a massive task and is the main restraint of developing an Ai. An Ai involves massive amounts of storage data just for it to function. Most companies can't afford to make a good Ai to make products for them, for now.
It can cost millions for companies to the memory alone for Ais. A High-Bandwidth Memory (HBM) (for Ai servers) costs around £11 ($15) per GB. A single high-end AI chip might incorporate 96 GB of HBM3E memory, costing approximately £12,400 ($16,500) for the memory. Including the development - £90,000 - £125,000 ($120,000 - $180,000), up to £225,000 ($300,000), etc. Not including building facilities for massive companies that can cost over £1 billion. The newest models for ChatGPT cost over £1 billion ($1.2 billion), made by Sam Altman, co-funder and CEO of OpenAI.
All of this isn't even mentioning the cost of training an Ai for years. With that, it can cost up to dozens of billions to make these big Ais. It's so expensive because it has "bad memory."
Not many companies can afford this (not including facilities or training). The part of the brain that controls memory is the hippocampus, which can also be connected to emotions. Something along the lines of the hippocampus would need to be present within the Ai for it to have more capable memory because it's the only way to get fast learning Ais (that we know of).
This brings up a good question. Will we even be able to afford a conscious Ai? It costs billions to make advanced text generation Ais. How much will it cost to add all of these receptors? All these parts?
Part 5: Specialization Versus Generalization (Conclusion)
Let's address the elephant in the room. Some of you may have asked yourself, "What happens if we simulate all of these systems to make a conscious being?" Even though it's expensive and we don't really know much about much about consciousness, it's not farfetched to say we'll in a thousand years have the power to simulate these things.
But even with the best technology we have, when simulating a stomach, it doesn't digest anything. What's to say that if we simulate consciousness, it won't be like the stomach. It could very well just not work like we want it to. This is an example of functionalism.
These systems and body parts are just aspects of our consciousness. We forget about how fast our bodies do what an Ai would need years to do, like learning or memory. An Ai compared to us, compared to any creature, including sloths, is very, very slow.
The reason for this is that even though we can simulate an ear drum on a computer or a hippocampus on a computer. What a computer will never be able to simulate is having more than a 1 (on) or a 0 (off). Most creatures have thousands of different types of cells doing specific jobs (muscle cells, skin cells, red blood cells, etc).
This creates the capacity for us to breathe while walking and dream while resting. Doing many things in an instant (us) is better than doing many things in a short amount of time (Ai). It's better because, even though it's a small difference, Ai will take much longer to do things like learning.
It's also much less expensive. This can be seen with monoclonal antibodies, discovered by César Milstein and Georges Köhler in 1975. The process of cancer cells merging with rat white blood cells to make rapid duplicating antibodies. It costs at most £37,600 ($50,000) to make this.
An embryo can also be recognised as being suitable for this topic. Aristotle and Karl Ernst von Baer were known for studying embryos and making attributes that led to the creation of embryology. Karl Ernst von Baer is recognised as the biggest contributor. It's a multicellular cluster of cells. Starting as a zygote (fertilised egg) and progresses through cell division. After 8 weeks, it usually becomes a foetus.
Somehow, combining the cell division of an embryo and the speed of monoclonal antibodies would make an efficient way to create life. I'm not saying we should combine embryos with monoclonal antibodies, I'm saying the properties together would be a more efficient, less expensive, and what i believe is the only way to turn a computer conscious.
Now i have to ask: At this point, are we making a conscious Ai, or are we just creating life?
Part 6: Names And Concepts Used.
Karl Ernst von Baer, Aristotle, César Milstein, Georges Köhler, Sam Altman, Phineas Gage, Derek Parfit, Alan Turing, and John McCarthy are important names for this subjects.
Biology, embryology, philosophy, and, of course, artificial intelligence were important topics.
Hello everyone.
I am a 15-year-old student exploring how Ai becomes conscious. This is my second post here, and feedback would be very much appreciated.
Today, I'm going to present how Ai can only become conscious through cells and real tissues, not just 1s and 0s. For Ai to be conscious, it needs to be of genetic material.
Parts:
• Story (1)
• The "Hearing" Of An Ai (2)
• The "Feelings" Of An Ai (3)
• The "memory" Of An Ai (4)
• Specialization Versus Generalization (Conclusion) (5)
• Names And Concepts Used (6)
Part 1: Story.
a "man" named Sam wakes up in a room with a table and a chess game on it. Next to it is a machine that can make a perfect copy of your body, but turn the cells into 1s and 0s, turn them into code, to make an ai robot, that is based off you, but metal.
A voice starts speaking from the radio above. "Good afternoon, Sam," he said, "we hope you're not too disturbed. Would you mind stepping into that machine over there." Sam, after being explained his safety, reluctantly entires and leaves the machine.
After a minute, a metal man walks through the door of the room. On its arm, it said B100. They play chess. Sam doesn't know how to play chess and neither the robot (because it's copied and hasn't been taught). After it's explained, they play chess. Sam beats the robot. It's as if the robot had not heard a word that was said about chess.
In the second, a man walked in. Sam's victory happiness quickly finished. The man was him, Sam. The man that looked like Sam told Sam to pull its sleeve up. It said B100 on "his" arm.
What the story is saying is that a machine made from only code can learn but can not hear, speak freely, feel, and maintain information without massive amounts of storage. Everything made of genetic material, however, can hear, speak freely (to some capacity), feel, and maintain information. These are the aspects of a conscious being. It's inspired by Derek Parfit's book "Reasons And Persons" in 1984.
Part 2: The "Hearing" Of An Ai.
"It's as if the robot had not heard a word that was said about chess."
It seemed the Ai didn't hear because it simply didn't. when Alan Turing explored the possibility of artificial intelligence, the idea was to make it do tasks. He did not dream of a day that it would be considered the way of the future. John McCarthy (father of artificial intelligence) coined the term in his 1955 proposal for the Dartmouth conference. These originals are important to know because they show what the Ai was made for. Not to have freedom, but to work for us.
The Ai can not hear because it's not made to. If it's not made to hear and Alan Turing or John McCarthy never wanted it to hear, why would we expect it to be a task for us to make it hear sounds? This goes back to needing biological matter because we need to ask how would an Ai heat sounds?
How we hear sounds is through our ears. The sound waves need to pass through the ear canal, ear drum, ossicles, eustachian tube, and cochlea (where transduction actually happens). Most animals have a system like this. Ai would need this to if they were to hear because this is the only way we can hear (that we know of). This process is the same for speaking, walking, etc.
Part 3: The "Feelings" Of An Ai.
"A metal man walks through the door of the room."
This is obviously the most debated thing about Ai. 1s and 0s just can't feel. They lack the nervous system and hormones like dopamine (happy hormones, for example) to feel such happiness. A reduced amount of dopamine leads to sadness. This can be controlled by the pituitary gland in the brain that secretes hormones. There're hundreds of ways the body controls emotions that a computer simply doesn't have.
Places in the brain like the amygdale control emotions like sadness or happiness. The limbic system is the overarching network of many things, mainly emotions. The limbic system acts as a bridge between the thinking part of the brain (neocortex) and the survival part of the brain (brainstem).
The process of feeling an object is one of the most complicated systems we know of. It involves the skin interpreting pressure, heat, and texture variations caused by the object through transduction. With specialized receptors (mechanoreceptors, thermoreceptors, etc), convert these physical properties into electrical impulses. These impulses travel through the neurons, through the spine, into the brain, then processed by the brain to recognise the properties of the object. These signals are integrated into the memory to remember the objects' properties. The system is not present in any Ai model, obviously.
Can you see that this isn't just something you add to an Ai? This is an extremely complex system that has developed over billions of years for survival. We still don't understand it because it's just that complex to understand it, Let alone simulate it for an Ai. The complexity of emotions was exposed by Phineas Gage with his accident of a pole going through his head and surviving in 1848.
Part 4: The "memory" Of An Ai.
"Sam doesn't know how to play chess and neither the robot"
The memory of an Ai is famously a massive task and is the main restraint of developing an Ai. An Ai involves massive amounts of storage data just for it to function. Most companies can't afford to make a good Ai to make products for them, for now.
It can cost millions for companies to the memory alone for Ais. A High-Bandwidth Memory (HBM) (for Ai servers) costs around £11 ($15) per GB. A single high-end AI chip might incorporate 96 GB of HBM3E memory, costing approximately £12,400 ($16,500) for the memory. Including the development - £90,000 - £125,000 ($120,000 - $180,000), up to £225,000 ($300,000), etc. Not including building facilities for massive companies that can cost over £1 billion. The newest models for ChatGPT cost over £1 billion ($1.2 billion), made by Sam Altman, co-funder and CEO of OpenAI.
All of this isn't even mentioning the cost of training an Ai for years. With that, it can cost up to dozens of billions to make these big Ais. It's so expensive because it has "bad memory."
Not many companies can afford this (not including facilities or training). The part of the brain that controls memory is the hippocampus, which can also be connected to emotions. Something along the lines of the hippocampus would need to be present within the Ai for it to have more capable memory because it's the only way to get fast learning Ais (that we know of).
This brings up a good question. Will we even be able to afford a conscious Ai? It costs billions to make advanced text generation Ais. How much will it cost to add all of these receptors? All these parts?
Part 5: Specialization Versus Generalization (Conclusion)
Let's address the elephant in the room. Some of you may have asked yourself, "What happens if we simulate all of these systems to make a conscious being?" Even though it's expensive and we don't really know much about much about consciousness, it's not farfetched to say we'll in a thousand years have the power to simulate these things.
But even with the best technology we have, when simulating a stomach, it doesn't digest anything. What's to say that if we simulate consciousness, it won't be like the stomach. It could very well just not work like we want it to. This is an example of functionalism.
These systems and body parts are just aspects of our consciousness. We forget about how fast our bodies do what an Ai would need years to do, like learning or memory. An Ai compared to us, compared to any creature, including sloths, is very, very slow.
The reason for this is that even though we can simulate an ear drum on a computer or a hippocampus on a computer. What a computer will never be able to simulate is having more than a 1 (on) or a 0 (off). Most creatures have thousands of different types of cells doing specific jobs (muscle cells, skin cells, red blood cells, etc).
This creates the capacity for us to breathe while walking and dream while resting. Doing many things in an instant (us) is better than doing many things in a short amount of time (Ai). It's better because, even though it's a small difference, Ai will take much longer to do things like learning.
It's also much less expensive. This can be seen with monoclonal antibodies, discovered by César Milstein and Georges Köhler in 1975. The process of cancer cells merging with rat white blood cells to make rapid duplicating antibodies. It costs at most £37,600 ($50,000) to make this.
An embryo can also be recognised as being suitable for this topic. Aristotle and Karl Ernst von Baer were known for studying embryos and making attributes that led to the creation of embryology. Karl Ernst von Baer is recognised as the biggest contributor. It's a multicellular cluster of cells. Starting as a zygote (fertilised egg) and progresses through cell division. After 8 weeks, it usually becomes a foetus.
Somehow, combining the cell division of an embryo and the speed of monoclonal antibodies would make an efficient way to create life. I'm not saying we should combine embryos with monoclonal antibodies, I'm saying the properties together would be a more efficient, less expensive, and what i believe is the only way to turn a computer conscious.
Now i have to ask: At this point, are we making a conscious Ai, or are we just creating life?
Part 6: Names And Concepts Used.
Karl Ernst von Baer, Aristotle, César Milstein, Georges Köhler, Sam Altman, Phineas Gage, Derek Parfit, Alan Turing, and John McCarthy are important names for this subjects.
Biology, embryology, philosophy, and, of course, artificial intelligence were important topics.
Thanks for reading.