‘Artificial intelligence isn’t yet intelligence. Humans are still smarter’
Maxim Akimov (Innostage) about the hype of neural networks, the benefit and danger of artificial intelligence for humanity
Will robots replace us and will we become jobless? What is artificial intelligence in general and who is smarter, intelligence or humans? Where will thousands of artists and photographers whose art is now successfully replaced by the neural network? Will the script of films about Terminator become a reality? In an op-ed column for Realnoe Vremya, head of the Centre for Counter Cyber Threats at Innostage CyberART Maxim Akimov reflects on what world we are rapidly moving to now, when artificial intelligence entered a new era.
“ChatGPT was lucky to become the first to reach new horizons”
Debating on this theme, first of all, it is necessary to find out what artificial intelligence is, how we understand this concept. Is it intelligence in general, how real intelligence is it? I offer to have a look at it with the eyes of a programmer. From this perspective, it looks like a myriad of functions and a myriad of conditions a programmer sets by picking. In other words, in fact, a programmer sets the methodology of work of artificial intelligence. This means any programme of this kind in the end is a product created according to a specific person’s image and intelligence.
But it develops further. The creation of the data model depends on the amount of information it is inserted. The more information the neural network ate up, the better answer it provides. One should understand that these shouldn’t be sporadic events. The neural network needs to be fed with millions, billions of events before it starts to give a quality result. Results are accumulated, added, verified and at some point the neural network will start enriching itself.
Neural networks have existed in our world for a long time, but precisely ChatGPT has recently made it. I think the wave of discussions around it is caused specifically because it is qualitatively different from previous versions of neural networks. It is a well-formulated model of data processing that gives quite good results, good answers to specific requests.
Such a qualitative leap was inevitable because technologies develop, data is accumulated. The thing is that ChatGPT was lucky to become the first to reach new horizons. Like we once went to space in 1961: everybody had been pursuing it, but we were the first to go. So the same happens to neural networks and artificial intelligence: everybody was beating around the bush, and in the end somebody annually made quite a good product.
Code outsourced by neural networks and end of image copyright?
For instance, a real programme can be developed with its assistance — and the IT community has already started using it. I have colleagues who jokingly decided to try to ask ChatGPT to help programme some specific function. They received quite an answer of quite a good quality — they just need to fix some parameters a bit and you have ready-to-use code in your hands. Instead of wasting your waste on wiring thousands of lines of code, now you can give this task to the robot. And you just give it some logic. The most important thing is to formulate the request, and you will get the correct, quite satisfactory answer.
We see how our colleagues are “playing” with ChatGPT at other companies and even in other countries: videos with people checking how correctly artificial intelligence does tasks are spreading on YouTube. And if there are still questions for the third version of the programme, the fourth is already quite good, both from the perspective of the correctness of the code and its security.
Our stand at PHDays had cards with our products designed like graffiti. So a neural network generated these images. There is another very practical and specific sphere where ChatGPT and similar programmes can be used even today — the generation of unique, free images to illustrate your content. Many think why they should pay for photos and images from photo banks if a neural network can be asked to draw them? For instance, we have also started doing it. Everything is absolutely legal, copyright isn’t violated.
The neural network can become a serious assistant to a specialist in humanities: for instance, to get content for an article or presentation, collect some information. A human just needs to check it out and “thank” it. Labour costs can reduce with the help of artificial intelligence and redirect your time to another, more constructive area.
Are we going to become jobless?
If we think about the list of specialists whose work can become easier right now, with the appearance of the superscientist ChatGPT — it is designers, programmers, journalists, copywriters, secretaries, translators...
Of course, there is a question here: what can humans do if robots start doing all routine work? I have already explained how time can be saved when developing a programme code. You will ask where should neophyte programmers, juniors go? Because all routine simple work is usually on their shoulders.
So are there going to be fewer developers as artificial intelligence develops? I don’t think so. If you notice, more electronic services are appearing, the world of information technologies is developing. And the more services, the more people should be involved in this. I understand the fears because it is logical to wait for people to be laid off as their tasks are automated. But somebody has to maintain this automation. However, people will likely have to repurpose quickly. For instance, programmers will not learn how to write thousands of code lines themselves but think of the logic of this code, set this logic to the neural network and feed it with the as accurate as possible request for a programme. Programmers will not learn to monotonously insert a code but pay more attention to analytics, logic and the creation of the correct task for the neural network.
“Nobody knows what this lead to”
But any coin has two faces. The scariest thing in terms of the development of neural networks is that nobody knows what all this extensive development will lead to. ChatGPT is developing too fast. Many have already started to think about hit: our and foreign regulators, ministers, Elon Musk — everybody in the world. They are asking to ban the development of the neural network for some time to understand what to do with it in the future. It is necessary to make up rules of the game to make the process manageable. People in corridors of power and negotiation rooms of the biggest corporations say: “Give us a pause for a year to understand how to live with it.” But it is impossible to concoct how to harness the process and move it to the direction we need. Has everybody perhaps seen Terminator? So today this script doesn’t seem so fantastic and impossible.
Again, we all remember Isaac Asimov’s cycle of stories on robots. He invented an ethic model in the middle of the past century: “A robot should harm humans in no conditions.” But we don’t know what ChatGPT and other neural networks have under the hood. Does it have the logic “Do no harm to humans?” For instance, today the neural network won’t answer the direct question about how to do something harmful and even dangerous, it will stay: “This is a criminal act.” Btu this request can be formulated in a way that it will anyway give an answer because as I already said, artificial intelligence isn’t yet intelligence. Humans are still smarter than it is.
I will put an example that’s the closest to us of how destructive it would be to use a neural network: to task it with developing a virus or a hacking plan. Moreover, its built-in defence can be bypassed too — ask to write a virus without saying it is going to be a virus. Now ChatGPT filters most of such requests, but with certain training it can be made write a specific procedure. I have already heard stories of how hackers are “playing” with it. They are always on the edge because a swindler’s task is to create something new that has never existed. And I won’t be surprised if neural networks star to be used en masse for generating attacks.
Is it going to become smarter than humans ever? Perhaps. Simply because it operates more information than humans. We anyway have limited possibilities of learning information. It increases day after day, and a human brain cannot process it fully and use it like a computer. Though we all know Cyberpunk game, who knows, perhaps if a human memory can be inserted additional chips, modules and similar devices...
Art of the future
I cannot help but remember a recent scandal with a photo that won an international prestigious award, and its author later recognised it was generated by a neural network. Art intelligentsia, painters, artists were in awe because the visual part of what a neural network does is almost undistinguished from what a human could make up. Modern artists, photographers, designers fear they are going to be replaced by artificial intelligence and are demanding to ban this all immediately.
But I think this issue cannot be resolved with bans, while progress cannot be stopped. By banning image generation by a neural network, we can obtain a reverse effect and, conversely, focus on what people have never known and heard. Loopholes will certainly be found.
I think with time the art of classic artists and photographers is going to become something what vinyl discs have become today — pleasure for connoisseurs, rare devotees. I understand the pain of artists who put their soul in it and live and breathe their compositions. But I think they already should think about how to fit in the changing reality. New art is needed where artificial intelligence can be integrated. Perhaps, it will be paintings created using neural networks, but bigger, elaborate, deep ones, a painting with “layers” you sink in immersing in the world created by an artist and AI.
What is ahead, the City of the Sun or enslavement of humanity by machines?
I am not inclined to be dramatic and I think the development of neural networks shouldn’t cause a collapse, total unemployment and mass despair. Humans will simply have more time for creation. When a specialist won’t have to spend time on routine, they can increase the speed of product development and constantly improve its quality. Their efforts will be focused on another direction. Remember the industrial revolution that once happened in Europe: when a machine tool replaced humans, people started to quickly deal with other things after recovering from the first shock.
In this respect I like Jacque Fresco’s ideas who has that routine should be shifted to robots as much as possible and humans should be left creation and art. Humans shouldn’t sweep the streets in the morning, a robot can do it. A human, as a supreme creature, should think about how to make this work better.
I sometimes imagine what our world will be like in 50 years. My images haven’t changed since kindergarten in some way: for instance, I still believe that cars will start flying. Many things will be automated in our world. Human life will be very simplified with the neural network and “artificial intelligence” (in the way I understand it). And we will deal with the creation and use our living, moving intelligence to something useful. So that people in the whole world feel better. I want to believe in it!