Journal
The simulator is not just the game it's the psychology of life scaled down to a 16 boxes maze. In the simulator we have basically converted some of the laws of psychology into mathematical equations.
Last updated
The simulator is not just the game it's the psychology of life scaled down to a 16 boxes maze. In the simulator we have basically converted some of the laws of psychology into mathematical equations.
Last updated
My name is Chaudhry Talha here is a brief about the intentions, motives, or views behind Intuitive Artificial Intelligence or Intelligent Artificial Intelligence (IAI). This journal also works as a rough ideas board that I am doing and might change later as the research progresses.
Below is a theory that I see when I look at the data being collected. It's not the final work but I'm sharing it here to deliver the idea.
What if I tell you that every grid in this simulator is a point of life (a life scenario) or life itself and it all depends on the perspective you put it through. Have a look at the image below and read along:
In the above image, there are three games data translated to grids and a path is drawn to each of the grids that show the user journey.
Imagine each of these grids as a completely different scenario of the life of one person or they define a particular scenario in a person's life, for example, a life scenario could be going for a job interview, the user starts at 2(14) and went to 8(10) it was windy (meaning going bad) he took a small risk b opening 6(9) it was still windy this grows the user's fear that they are going to fail and hence they took a big jump to 5(3) and died. Maybe, an AI can learn that when things are going windy don't take big risks from this one. From the second grid, AI can learn when it starts to get wind go up. Remember the AI will never be able to know what the full grid was, it'll just know the path it has explored. Similarly in the third grid, an AI might learn that if you are on an unknown territory or a scenario where you don't know what to do go for a smaller or optional goal rather than running after something that you are not going to get or has greater risk.
It's all about from what perspective you are defining each element in this simulator. Any program can code an intuitive AI into their systems and if needed the AI can use intuitive AI when needed. There will be a control on how much the AI can ask its Intuitive Brain. For example, in an image recognition example if an AI is unable to identify that the animal is a dog three times it is allowed to use its intuitive brain. You can mold this example into anything to it this algorithm here is one o the many other scenarios this example can fit into a problem. Now the one intuitive AI perspective to define this image recognition example would be:
Goal: Need an output string of the animal in the photo
Optional Goal: Just to reverify use a different image recognition algorithm before presenting the result.
Risk: Image recognition result is wrong
The AI will look into the millions of intuitive data points that we have collected using the simulator. It'll pick a starting point, let's say it's index 12. It'll sort millions to only select the grids that have an opened box at index 12 (I am defining it like that, you can use our own perspective to take all the failed ones as well and let AI decides if it wants to follow this path or not based on the hints it's getting). Then from this result, the next sort will be to get all the scenarios where the user has won or found gold. Now on index 12, it has an answer let's say it has detected the dog as some other animals say a cat which is wrong and we know the AI is going to fail and from the sorted out result it randomly picks a grid in which the user has found gold. Consider the path below:
AI has the answer say cat, it's going to move forward and found an optional goal which means it'll run a different algorithm just to reverify and that algorithm also detects wrong say it detected that it's a seal. The reason why it decided to move to 8 and not to any other index is because of the random selection of this grid. Here the task is complete but the risk is higher now as both answers don't match. Here the AI only knows that they don't match and also has no idea if any of them is correct but knows for sure that both are not correct.
It's a cat(True) and not a seal(False)
It's not a cat(False) but it's a seal(True)
It's not a cat(False) and it's also not a seal(False)
Here the AI has to decide either to move forward or go back and for that, we need a zero (No, don't move forward) or a one (yes, move forward). If we take the above to NAND we end up with 1,1,1 and then pass these through an XNOR gate and you'll get 0 which means not to move forward.
You can completely change this output to make sure it continues the path. It's all based on the risk that you want the AI to take. As moving forward you're going to get wind and if you set its score to say -1 which means bad. If the hint is wind, glitter the score can be 0, and if it's smell and glitter the score can be 1. If it's all three it can be 0, and if it's X nothing it can be 2. Based on the next score that you get you can decide either to alter or not. Let's say we have this logic implemented and we ask for the next index to move to and we get 13 and the score as its wind is -1 so AI knows it's in a dangerous zone. Our logic says that if the risk is zero or greater then move forward otherwise choose another grid.
Let's say the AI selects the second one as a winner.
In this new scenario, the AI will start again with a totally different thought process it'll go to 15 get a score of 1. Which is good. The next index the user gets on is 14 and gets -1 so doing +1-1 we get zero and based on our logic zero is greater than -1 so we'll move forward. But what about the previous data i.e. -1? We'll also add it to this result so 0-1 is -1 and we can't move forward. So, we'll need another winner board that has 12 as the opened index. Also, note that the opened box also has a score. In the case of the second grid, the score at opened index is 1 as it has glitter and wind. So, let's say this time we have programmed to select the highest score grid with index 12 as the opened index. So we need the third grid. Its starting score is bad i.e. -1 but in the data, we only have this one left so we're forced to use this grid.
In this scenario, the AI will start and take the first move. Evaluate its score i.e. 1 which is good to move forward. Now next we go to index 6 which is -1 so here I have to re-evaluate. So, -1+0-1+1 = -1 and we know we have to change the grid. let's say we don't have any grid left and are forced to move forward. We'll go to index 3 and here we'll get 2, and then again we get 2, then we get 1, then we reevaluate the image lets say this time we get Dog. The next index would be 11 and that is to present the findings so we'll say that intuitively the AI thinks it's a dog which is correct. Here the output of the reevaluation can be wrong, there is a full chance of that and in this scenario, we can say that we have found a goal so the output will say it's a cat or a seal or some other animal. it'll decide based on the score of each grid i.e. -1, 0, 5 which means that the AI is less confident that it's a Cat and more confident that it's a seal and it's very confident that it's some other animal. This is still not correct but the computer thought about it. Maybe in a world where AI is dominant, dogs are called seals or some other name because of this intuitive decision. We can change this scenario to sort the grids that start at 12 and have all Gold and Monster found only. Depending on your scenarios you can code the risk. So, it has countless possibilities and you can reprogram it to work in the ways you want it to take the decisions, meaning you can control the intuitive amount of the AI. Too much intuition is bad and too little is also bad. Our goal should always be to take the best scenario.
So, this is just one example of how Intuitive AI will take a decision. This situation can be molded to a different logic but the point to take from this example is that there is no one type of programming scenario. You have to program it in a way that you take the best scenarios possible. That is why we used NAND & XNOR gate as they take the best scenarios. The goal of Intuitive AI is not to improve an algorithm but to add a brain that takes decisions where there are no other options left.
I read it somewhere that if we can define anything with a mathematical equation we can get the computer to do that. Because the computer is built on logic. That is why our brain is divided into two parts, anyone who sees a brain diagram can tell it's in two parts and it's only because our decision-making process consists of the logical and the hunch.
Time is very strange. It's a part of our daily life but yet if you do a little research of what time actually is you'll be blown away with what it is and to save you time, no one can tell what time actually is. The problem with the majority of us is that we are either living in the past or living in the future we just don't live in the present or we just don't live in the NOW.
A good example that I give to myself of why this concept of IntiutiveAI that what I'm presenting is worth exploring is if you go back to 1995 when Java was introduced. It was the first Object Oriented language and it was inspired by real-world concepts like objects, Inheritance, Abstraction, etc. Objects are all around us and a similar concept is used to define Java Objects though they are a bit different in nature they are not like the objects around us they act like that. Similar is going to be the situation here, the intuitive thinking of a computer will not be similar to what we have but it'll be something that only machines can do or in other words, it'll be machine intuition that is made for the machine.
“You are awareness, disguised as a person.” ― Eckhart Tolle
Computers are different from humans so self-awareness in computers will be a bit different but can a computer ever achieve it? Intuition revolves around the psychology of humans and the simulator basically defines some of those psychological principles into a mathematical term. When I first designed the simulator I went to the one and only meeting I had with Dr. Vazeer and I was super hiked about this simulator that I got super excited telling him about it that sir it fulfils the aspects and I think we have something here that can revolutionise something if an algorithm is built around it and of course because of this excitement I got carried away and he got lost in the conversation. So, I opened the simulator and gave him the rules and that's it. He played about 5 games and he realised everything that I was trying to say and he said I know even if I play this game a million times I'll still not be able to improve and will always have to make decisions based on my intuition. The simulator was portraying the NOW factor. You have to focus on the now situation, you know you're about to open a random box which can have anything in it, and all the boxes look the same so the box you'll choose, you'll explain your decisions to someone like "I just felt like opening this box so I opened it", that's intuition scaled on a different scale as I wanted to take the real world intuitive decision making into the computer. We did couple of experiments where we gave the simulator to bunch of people to try from there we derived the above response.
Let's see how the simulator works but this time we'll see how every step is mimicking the real-world psychological principle.
Any problem you have, we need to scale it down to a 4x4 grid. The rules are the same for everyone, i.e. you were born, you are going to get aged and one day you'll die. These three simple rules are for everyone, the world is different for everyone and out simulator present that by randomising grid every-time. Someone is religious, someone is atheist, someone loves to eat meat, someone is a vegan, and so on. Our decisions throughout our life shape how our life is going to be. Also, the simulator is designed by keeping an individual in mind, because my belief is that you are alone in the world. Your family, your parents, your kids are just other people you are an individual and you'll be answerable to your decisions only and no one can take your place no can you take someone else's place. That is why when you are playing the game you're all alone, all you have is a world that is changing with every decision you are making, either you'll succeed or just die.
The scoring in the game is also built on psychological principles. You can only get -1, 0, 1, or 2 in the game. Most people will strive for 1, this portrays the 80% of the world population that is middle class. Many people if they see a good calculated chance that they'll be able to get 2 only then they'll take it where some people will go out of the way to take the risk of getting the 2. These are people who are successful. The people which get 0 are the ones who were going good in life, but died early otherwise they would've done more great things but they left behind some sort of legacy, which saved the people after them something meaning they were saved from a total loss which is -1 in the game i.e. the smallest number. Now the people of -1 are the ones who get the lowest possible in this small 4x4 world. They can be people who try to work hard but failed even after taking a calculated risk or following their intuition. They can also be people who don't care and they are just making decisions randomly and ends up failing. In life, we have to be a bit careful and use both sides of our brain to make decisions. The risk of failure is still there but that's the life you do your best and let fate decide.
Now, if an AI is able to play through this world and improves its accuracy to an acceptable amount I can confidently say that it has achieved the intuition decision making and I'll not stop until I am able to solve this.
The 4x4 grid can make more than 16! = 20,922,789,888,000 (twenty trillion, nine hundred twenty-two billion, seven hundred eighty-nine million, eight hundred eighty-eight thousand) unique games. By more, I mean if we consider the one box that opens up randomly then the number of the unique pattern increases. The total the Population Reference Bureau estimates that about 107+ billion people have ever lived on this planet, according to https://www.bbc.com/news/magazine-16870579 so, if you give one unique grid to each of these humans they'll all have unique lives and we'll still have a lot more left for the humans who'll be coming in the future.
If you want to see the 400+ intuitions we were able to capture of people here is the datasheet https://docs.google.com/spreadsheets/d/1-9SaOmx2HAwQPbpjHVPw8OxMih1UH-NUVNxn4R2CyKI/edit?usp=sharing
In conclusion, this is how I was able to place some of the psychological principles of our life into a computer simulation and now using the data which is people's intuitive decision we are in the process of developing an algorithm.
