Gábor Túri: My first date with artificial intelligence

The development is now at the point where you can "chat" with our AI partner as a visual partner. In the course of this, a young girl who could be said to be pretty was introduced in this German broadcast, who later turned out to be artificial, just a digital "being", who was not actually sitting there among the conversational partners.

A few days ago I saw a short discussion on German TV about artificial intelligence, abbreviated as MI, or using the English term - artificial intelligence - about AI.

The topic is already fundamentally interesting, since AI - as a branch of computer technology - aims to make IT machines intelligent. In this context, intelligence should be understood as the quality that enables a living being (including humans) to act appropriately and foreseeably in its environment. This includes the ability and response to sensory impressions, receiving, processing and storing acquired information as knowledge, and understanding different languages ​​to achieve problem solving. In other words, very tempting options.

The development is now at the point where you can "chat" with our AI partner as a visual partner. In the course of this, a young girl who can be said to be pretty was introduced in this broadcast, who later turned out to be artificial, just a digital "being", who was not actually sitting there among the conversational partners. All of this provided a surprising experience, because she had a pleasant voice and charmingly admiring eyes, and her shapely lips moved perfectly in accordance with the words she uttered, and we could even see her tongue for a moment.

It was incredibly lifelike!

The "girl" gave correct answers to all the questions asked in the program, and she also looked back at her interviewer with encouraging understanding if he did not ask her one of her questions clearly enough.

Of course, I will note right away - if only for the sake of my female readers - that of course there can be male MI answering partners, they must be pretty and well-cut. What's more, as a self-deprecating quip, I can't wait to see when the gender frenzy will break in here too, and there will be a respondent of the third gender. Or - excuse me - even fourth, fifth and many genders, so that they can satisfy the representatives of the 72 (plus VAT?) gender identities "known" so far.

Encouraged by what I saw, I searched the world wide web for a relevant link. Well, not because of the girl who would answer me there, rolling her eyes charmingly, but because of an AI program that answers my written questions in writing. Because of course - as the senator Caius Titus once put it - the word disappears, the writing remains: "Verba volant, scripta manent" .

Summa summarum: I was interested in how much I can trust MI's answers.

I started from the assumption that the response information packets had to be fed into his memory by someone or others. Because let's not forget that artificial intelligence doesn't draw its answers from its own well, since it has no intelligence - in Hungarian it means "no mind of its own" - only stored information.

And all of this is fed into the system by a multitude of programmers, even a network of them, with years of long work, who - and let's not forget this! – representatives of real intelligence. In other words, members of the human species homo sapiens have been collecting and passing on the collected information for nearly 200,000 years - from one generation to the next.

At the same time, the human being has a major flaw, namely that it is not infallible. As the saying goes: "To err is human". That's why I got curious and gave it a try. I wanted to know if the information fed by the species called humans was correct in our creation.

I asked ChatGPT my first question in German; I quote the conversation verbatim:

Me (in German): "Does it matter to you if I ask the questions in Hungarian or German...? Is there a difference in the use of information sources available to you because of this?"

ChatGPT (in German): "As an artificial intelligence, I can communicate in several languages, including Hungarian and German, and I can understand and answer your questions in any language. My answers are based on the information available to me during the training process, as well as data available on the Internet. Whether I can give you an accurate and complete answer depends on the availability and quality of the information I have access to. However, if you have a specific question, I always try to give the best answer based on the available information.”

Then I continued in Hungarian.

Me: "Of course, there are articles published only in Hungarian that have not yet been published in the German online world. My question is: if I ask this question in German, can you also go to the Hungarian online information?"

ChatGPT (again in German): As an artificial intelligence, I can access information from the Internet, regardless of the language in which it was published. So if there are articles in Hungarian that are not available in German, I might be able to find them and provide information about them if they are available on the Internet and I can access them.

I thanked him for his answer, to which he replied, this time in Hungarian:

"Willingly! If you have any further questions, please do not hesitate to contact me.”

I have to say, I started to like it, so the next day I asked him if he knew this Latin quote: "Quosque tandem abutere Catilina" , which we learned in the 3rd grade of high school, so I knew its origin. I wanted to get tested, well...

ChatGPT's answer: “The quosque tandem abutere is actually a Latin phrase from Cicero's In Catilinam , written in B.C. In 63 he said in the Roman Senate. The full sentence reads: Quo usque tandem abutere, Catilina, patientia nostra?, which translates to: How long will you abuse our patience, O Catilina? Cicero addressed a member of the Roman Senate, Lucius Sergius Catilina, who was involved in a plot to overthrow the Roman government. Cicero warned Catiline that her actions were intolerable and that the Senate would not allow her to continue abusing their patience and tolerance. Today, quosque tandem abutere is often used to express frustration with someone who behaves in an unacceptable or abusive manner and who has gone too far in their actions.”

Since the answer was surprisingly correct, I wrote to myself that MI is a good thing and works well. But my joy proved premature.

I didn't "meet" the AI ​​for a while after that. Perhaps I would have already forgotten "about him" if one day I had not looked for the meaning and origin of the coats of arms visible in the painting for one of my writings in relation to Dürer's picture of King Sigismund of Hungary and the German-Roman Emperor. in the section of Sigismund's era (1387-1437) in Szalay-Baróti's book A magyar nemzets története , published at the end of the 19th century. . Then, suddenly, my curiosity awoke again, whether "my" artificial intelligence could answer this too, and whether I would have received my answer sooner with the help of AI? I asked like this:

Me: "What coats of arms do we see in Albrecht Dürer's painting for King Sigismund around 1509?"

ChatGPT: "The portrait of King Sigismund by Albrecht Dürer (circa 1509) shows several coats of arms. One of the coats of arms is on the king's right and represents the Duchy of Bohemia-Moravia-Silesia, of which he was the governor during the reign of King Sigismund. The coat of arms depicts an eagle's nest containing three young, with a snake in the eagle's beak. The other coat of arms is on the left side of the king and represents the Hungarian kingdom. This coat of arms depicts three convex mountains, between which an oak tree stands, and below the mountains the eagle of the Árpád House can be seen. In addition, a coat of arms can be seen on the king's chest in the painting, but its symbols are difficult to determine precisely."

Since this answer was completely wrong - I would say stupid - which I can confidently say after half a day of "investigative" work, I replied:

Me: "Unfortunately, this is a mistake! In fact, this painting, which is located in the Nuremberg Town Hall and depicts Sigismund in the German imperial coronation costume, has the GERMAN imperial coat of arms on the top left of the picture, above the emperor's head, the CZECH coat of arms in the middle, and the HUNGARIAN coat of arms (with the double cross) on the right. ; then on the left, below the German imperial coat of arms, the HUNGARIAN coat of arms with red and silver bands (so-called bandages), and on the opposite side, on the right, the BRANDENBURG coat of arms. Please correct this, because the information you provided was wrong!"

ChatGPT: “I apologize for my previous reply, I was wrong.
Thanks for the correction and clarification. The coats of arms shown in Albrecht Dürer's depiction of King Sigismund are indeed the GERMAN imperial coat of arms, the CZECH coat of arms, the HUNGARIAN coat of arms and the BRANDENBURG coat of arms. The German imperial coat of arms can be seen in the upper left corner of the picture, with the emperor's crown floating above it. The Czech coat of arms can be seen in the middle, the Hungarian coat of arms to the right, and the Brandenburg coat of arms in the lower right corner of the picture. Thanks for the clarification.”

But then the AI ​​was wrong again, so I fixed it again.

Me: "The emperor's crown is NOT FLOATING above the German imperial coat of arms, this is again a mistake. The imperial crown can be seen ON Zsigmond's HEAD."

ChatGPT: “I apologize again for the inaccuracies in my previous responses. I was indeed wrong when I claimed that the emperor's crown floats above the German imperial coat of arms in Albrecht Dürer's depiction of King Sigismund. Furthermore, it is correct that the imperial crown can be seen on Sigismund's head in the picture. Thanks for the clarification and correction.”

My answer to this: "You're welcome!"

ChatGPT: "Okay, is there anything else I can help you with?"

I do not. Thank you very much."

ChatGPT: “You're welcome! If you have any further questions at any time, do not hesitate to get back to me!”

That was the end of our conversation, and I smiled at his last sentence. Although it sounded nice and good, it was indeed a polite farewell, but it did not encourage me - I will be very hesitant in the future! Based on my current experience, I am not at all sure that I will always receive an accurate answer from the system.

Of course, in the past there were also differences between certain sources of information, which in most cases had political reasons rather than knowledge. So, for example, we knew that there is a difference between the Great Lexicon of Réva and the New Hungarian Lexicon. However, since we were aware of this, we filmed the volumes in the spirit of this. We knew that when we were looking for a politician or historical person in the lexicon, from which and when we should accept what.

And this is exactly what we don't know about artificial intelligence! Where does the information you are looking for come from, i.e. with what motivation and from which sources was it fed?

What I found most interesting, however, was that based on my answers and corrections, ChatGPT from me as well - and that would be the essence of MI. That he can learn from the questions asked and from the corrections.

However, if this is the case, then anyone can "teach" them, and even supply them with false information accidentally or even intentionally. And let's face it: it's very dangerous.

So, no matter how charming ladies give us their seemingly certain answers, we always have to be on the lookout, because not all information is "true".

There is no doubt that sooner or later the system will be perfected - in light of how much knowledge it stores - but until then (and I think after that too) I'd rather reach for the bookshelf or attend live lectures where I can ask my questions to really intelligent people , to which they will answer, and maybe they will also be interested in my opinion.

Because it is this interaction that sharpens human intelligence and develops our knowledge.

So I'll stick with analog, old-fashioned respect…

(TTG)

Featured Image: Pixabay