Artificial General Intelligence - around the corner or far distant future?

What side do you take in the global polemics around AGI (Artificial General Intelligence) - the super humanoid that is smarter than all human beings in the history altogether? Is it possible or just another story from Silicon Valley?

This time we covered this topic in more depth trying to answer this question by ourselves. And here is what we’ve got so far.

AGI = AI x Human.
Thus, in order for us to get AGI we need to emerge AI and human beings. Which in turn means that we should kinda put one into another. And so far the majority of the world’ efforts had been aimed at putting Humans into AI. The whole idea of AGI is inspired by the notion that we can replicate people’s ability to reason in various formats not restrained by the context.

Now, let’s consider the concept of reasoning. Does AI really have the ability to reasoning?
Any type of machine learning is built around the concept of, well, learning on the data. The quality of machine learning is measured by the accuracy of predictions it can make after having fewer and fewer samples. And who are the judges? Obviously, people. Now, one may argue that AI has ability to reason. At the end of the day, what else does AI do when it sees that your purchase decisions forming some trend. It starts analyzing the data and gives predictions to you, right?

Now, here we have to be very careful with definitions. Because in fact, what AI does, is just about Pattern Recognition. The more accurate and the more amount of data it has, the higher chances to recognize some pattern is out there. Now, once some sort of pattern is recognized, it gives predictions of the future development of the pattern. If the future decisions made by a person aligned with the pattern, then AI learns even better, gets even more data, and precision of future predictions grow.

The main question is: is it really reasoning?

To get the answer we should understand the difference between the following concepts: knowledge and understanding.
What is knowledge? Well, it is information that is interpreted from some data. Now, what is understanding? It is the next level of intellectual activity that is based on knowledge which enables an intelligent creature to develop new pieces of information from the one that exists.

Is there a difference between phrases: “I know how it works” and “I understand how it works”?
Seemingly no, while to be really precise, the difference is quite huge.

One student knows a mathematical formula, another one - understands it. Now, it is much clearer to us that understanding gives us ability to possess knowledge, while knowledge itself is not enough for understanding. Now, the guy who understands the formula has potential to make new formulas, while the guy who just knows the formula is limited only by this very formula.

The problem is that when an AI starts learning from previous data and generates the future predictions, it doesn’t understand anything. What does it do then? It just knows the previous data basing on which, it starts generating the future predictions.

For example, here is a simple example of a row: 1, 2, 3, 5, 6, 7, 9… could you continue it? Definitely. Do you have any idea what this row is about? No idea. And that’s why you don’t understand this row, because it is absolutely abstract and it just requires you mathematical logic to continue the line. The same is happening inside AI. It has no idea about pictures, people it sees, but it can accurately predict the future numbers based on mathematical logic.

Can we say now that a robot has reasoning?

Now, it’s not all yet.

What separates machines from humans is not only the lack of general reasoning. It is also consciousness.

You may disagree whether consciousness even exists, but let’s just agree upon the definition - it is your ability to be aware of what’s happening in your world and make sense out of it (at least to your own level of intelligence). Now, consciousness is directly related to such a thing as experience.

Now, is there a difference between understanding something and being conscious about something?
Seemingly no, right?

While in fact, it is fundamental.

Let’s ask ourselves, how many people on the planet know what war is? 100%, right? Everybody knows what war is. Now, how many people actually understand what war is? Well, that’s harder to answer, right? Let’s assume 20%-30% of people. If more, then fine, but never 100%, right? Now, the final question is - how many people are conscious about war? And that number will be limited only by those individuals who have ever been to war, period.

Why is that? Because consciousness is about our ability to experience and perceive the world, that couldn’t be emulated without the whole biology in our bodies. In other words, if we don’t have emotions, if we are not afraid of pain, then we don’t have consciousness because only those things enable our ability to experience the world.

Now, are you getting it? Just imagine how many millions years it took the Universe to create a human being? And our smart people from Silicon Valley want to make it in the next decades. That’s just impossible. Because not only do you need to understand the most complex organism known in the Universe so far – human’s brain – but also you have to recreate the entire complex biology inside the human body.

The idea we came up with was simple - do what Elon Musk does with Neuralink. Put Robot into Human, not vice versa. That will simplify your life tremendously!

But that’s great news for people, right? Now, it is us who will become the AGI, not AI.

At the end, we would like to definitely put it clearly – it’s pointless for AI to compete with people. What truly matters is to introduce completely different metrics for AI and enable them to do what they are predestined to do well.

We identify two steps in the process of creating humans:
i) Creating the biology.
ii) Putting it into a different body and grow the organism.

And then we can expand our biological capabilities to get into the AGI level. And we should always think in advance how we are going to design it.