Limits to Artificial Intelligence

The text below is a literal transcription of the spoken video text.

I want to take up my intention again to attempt to let something be experienced about the difference between human thinking and artificial intelligence. Of course, we have a great deal of research available, which is research on the brain, which is scientific research, in which an attempt is made to map the brain in such a way that one believes that the moment will come when science will be able to fully comprehend how the human brain functions. . Watch video on YouTube

I want to take up my intention again to attempt to let something be experienced about the difference between human thinking and artificial intelligence. Of course, we have a great deal of research available, which is research on the brain, which is scientific research, in which an attempt is made to map the brain in such a way that one believes that the moment will come when science will be able to fully comprehend how the human brain functions. That, of course, can then be used fruitfully in imitating it in artificial intelligence. That is how I have always read it, that these are the two poles: on the one hand, artificial intelligence, which, on the other hand, is imitated from how the human brain functions.

It is expected that, at a certain point, this will, as it were, break away from each other, that artificial intelligence will undergo a development on its own in which the foundation of the human brain is abandoned. That is the principle of singularity, that there comes a moment when artificial intelligence is so far developed that it no longer needs human intellect, human intelligence. You can actually only contradict this if you could prove that human intelligence and artificial intelligence are two totally different forms of intelligence.

With brain research, of course, you cannot prove that, because that brain research is precisely the basis for the development of artificial intelligence, so you cannot expect that much attention will be paid to a difference there. And yet, you can gradually see something appearing in the public domain, a kind of impotence that arises in the further development of artificial intelligence, and that is extremely interesting.

So, even if you do not have an experience of the difference between human and artificial intelligence, you could still already read something from that difference. There has been an increasing kind of concern about OpenAI, ChatGPT, because people expect that at some point, the usable data present on the internet will, as it were, be exhausted.

You can imagine it: when you have a question and you ask it to this chat, you almost see that it searches the internet at lightning speed to then compile the answer for you. We used to do that on Google or so, where we would search ourselves for what could be found on the internet about a certain topic. Now, you can also have this done by ChatGPT, which does it for you, and it does that in 1, 2 seconds, or maybe even less. In such a short time, it has searched the internet for usable information and, based on that, provides an answer.

People are convinced that this form of artificial intelligence can develop further from a certain point. So, what has been entered into the internet as data—by us, by people, by everyone who uses the internet—all that data can, as it were, be continued by the system itself. And some doubt is beginning to arise there, and people are starting to see that what humans have entered into the internet as data is more or less running out.

There is still quite a large area left, but this OpenAI does not yet have access to it because those are areas that, for example, belong to the private domain. But a very large part of the open domain has, by now, more or less been screened by this artificial intelligence, you could say. That data is available, and if at some point that is exhausted, then, as it were, the development is also exhausted.

That, in itself, is an extremely interesting fact, that it is also acknowledged that it actually revolves around data entered by humans and that this OpenAI makes use of that data entered by humans. And then it is assumed that this intelligent machine will be able to further develop what data is available. That is precisely the point where voices are beginning to emerge saying that this system cannot do that. That further development is indeed possible, but that you then run the risk of bizarre combinations that have nothing to do with reality at all, that are indeed intelligent but cannot actually occur in reality. In that case, you would again need a team of people to keep such results in check; otherwise, it would become an entire field of information that makes no sense.

You already experience this sometimes, where you ask something and receive a bizarre answer. It does not happen very often yet, and the more complex your question is, the greater the chance that the system will not know how to handle it.

Well, I have used many words, but what it ultimately comes down to is that we engage our experience in this, that you experience what it is that this artificial intelligence system does indeed collect and combine data into a piece of text at an enormous speed—which we will never be able to do—but that what the human mind does, what human intelligence does, namely simultaneously verifying whether what is given as an answer is actually possible, that this does not really take place.

Of course, there are artificial intelligence optimists who believe that this will all still be developed, but you can now gradually see some doubt arising about this, and if you are familiar with human intelligence, then you also know that this doubt is justified and that it will never succeed in surpassing human intelligence in quality with artificial intelligence.

In quantity, certainly, in speed as well, but the quality of human intelligence is truly something entirely different from that of artificial intelligence. That is something I would very much like to convey to you, but that can only be done through these kinds of examples.

A possible example is when you look at a growing child. Then, from the perspective of Waldorf education, there has been—and perhaps still is—the aim to teach children in such a way that they gather intelligent content within themselves, but that it is not placed in a rigid framework, that it is not laced into a dogmatic form, that it is not fixed, so that it can continue to grow and remains open to development.

So that it is not the case that when a person, let’s say, reaches the age of 35, they have only a limited number of concepts at their disposal that are more or less fixed, with no room for change. Human intelligence is capable of much more. It is not a machine that collects data and then analyzes and combines that data until death.

Rather, human intelligence is a living process, in which the concepts you acquire can develop. They cannot develop if they are drilled into you with a “this is how it is and this is how it will always be,” while in the second half of the sentence, it then says: but we must be open to everything, and we must be tolerant, and we must embrace diversity, and we must—well, whatever else.

But the way this is presented to children is such that there is hardly any strength left for developmental concepts. That is the hallmark of human intelligence: that it is a living process, not a reservoir of data that you can analyze and combine at a certain speed—slow means you are dumb, fast means you are smart—but rather, it is a process in which the knowledge you acquire only becomes intelligent when it is integrated into life itself.

Thus, the creativity that an artist has—when they create their sculptures, paint their paintings, or compose a piece of music—that creativity remains within intelligence as well.

Great scientists, discoverers, inventors have not made use of existing data—of course, they have—but they invent something that completely transcends that data and brings something new into existence.

That is life. Of course, you could say: Artificial intelligence will be able to do that too. You can say anything. But when you experience what human intelligence is, then you can no longer say that, because then you know: artificial intelligence depends on existing data, and that data may possibly be expanded by the system, but to what extent that still has anything to do with reality—that remains to be seen.

And we will see that absurd results will emerge.

That is what I wanted to share today about human and artificial intelligence, and I will continue to look for other examples to show that human intelligence is truly something entirely different from what is being developed as artificial intelligence.

Limits to Artificial Intelligence by Mieke Mosmuller