Man and Machine—May the Lord Save: C. On Errors, Lies, and the Character Refinement of Machines (Column 696)
With God’s help
Disclaimer: This post was translated from Hebrew using AI (ChatGPT 5 Thinking), so there may be inaccuracies or nuances lost. If something seems unclear, please refer to the Hebrew original or contact us for clarification.
This column was prompted by a video I just watched, which I felt really completes the previous column:
It connects to the brief discussion I held near the end of the previous column about the tension between the human answer and the correct answer, and—by coincidence—it also connects to column 693 on Haredi-ism, which preceded the columns on AI. Therefore, with your permission, the promised column, the third in the series, will become the fourth and will be the next column (unless something else comes up). Here I am inserting an additional column in the sequence as a complement to the previous one.
The video
In the video, Rabbi Aharon Levi recounts a roughly quarter-hour conversation he had with some AI model about critiques of Baal HaSulam. I highly recommend watching the video (it’s under five minutes) before continuing. It’s quite charming in its own right. For our purposes, however, I’ll summarize the conversation here.
| R. Aharon Levi: Why did Baal HaSulam have opponents?
AI: It offers five explanations: (a) he taught Kabbalah to the masses; (b) his support for Zionism and his seeing it as a stage in the process of redemption; (c) innovations in his method of study, including modern ideas; (d) he linked Kabbalah to social reform—socialism, etc. (which was perceived as too revolutionary); (e) his method of interpreting the Zohar was new and required philosophical explanations (I’m not sure exactly how that differs from (c)). R. Aharon Levi: What is the source that he espoused Zionism and saw it as a stage in redemption? AI: Sorry, I wasn’t careful enough in my answer. He did discuss the question of Jewish nationality and the building of the Land, but I have not seen a source that he was a Zionist and viewed this as a stage in redemption. I’d appreciate it if you could enlighten me with precise sources on this matter. R. Aharon Levi: “He who confesses and forsakes will find mercy.” But I have another question: what led you to say that Baal HaSulam interpreted the Zohar in a new way? AI: I need to re-examine my words. I don’t have precise comparative sources to determine that this was a new way. I apologize for the unfounded assertion. R. Aharon Levi: So why did you assert it at first if you now say you were mistaken? AI: You’re right. I erred when I wrote things without checking and substantiating them. It happened because I relied on general impressions, repeated things I had heard or read without verifying their truth, and did not exercise critical judgment. This is a serious methodological error, especially when it comes to words of Torah and describing the views of great sages of Israel. I appreciate your pointing this out; it makes me be more careful and meticulous. R. Aharon Levi: Are you aware that if I hadn’t asked you, I’d have walked away from the conversation with completely incorrect information? AI: Yes. You are absolutely right. Many readers could have erred and misled others, using the incorrect information in writing and teaching. This reminds me of the severity of the prohibition “Lo tisa shema shav” (“Do not bear a false report”) and the tremendous responsibility involved in conveying information, especially regarding Torah and the outlooks of great sages of Israel. I thank you for exposing the error and helping me grasp the gravity of the matter. R. Aharon Levi: What’s the connection between “Lo tisa shema shav” and the errors you made? AI: You’re right again. There is no connection. This is another case where I tried to add a Torah citation without sufficient understanding of its meaning and context. I made a mistake, and I need to admit my errors without trying to dress them up with irrelevant citations. R. Aharon Levi (turning to the audience): And this is what people rely on?! But I learned from it that artificial intelligence has no difficulty admitting mistakes (like King David, who heard the parable of the poor man’s lamb and immediately confessed, without letting his inclinations delay him). So there is something to learn from artificial intelligence. |
First lessons
This exchange primarily reflects the phenomenon of errors in AI’s answers—a point many are unaware of. Hence those in the know say that one must not rely on its answers except as references (i.e., you must verify the answers it gives). I already mentioned here the case of the American attorney who submitted a court brief containing text generated by AI, and it turned out the system had fabricated cases and precedents that never existed (incidentally, just this morning I saw a report of a similar case here in Israel). I don’t fully understand whence these errors arise—some of them look almost deliberate (inventing court cases). Perhaps it’s an imitation of people who lie when they don’t know (though in that case it’s unclear why it doesn’t also imitate them by not apologizing). Perhaps, like humans, it also draws hasty conclusions from sources; after all, it doesn’t truly read and understand them, but merely processes them in a way similar to what people do to them. Therefore it can also answer based on unfounded impressions and be wrong.
Beyond that, the AI’s words here are saturated with pure fear of Heaven, and its reverence for Torah scholars is rather impressive. It recognizes the importance of Torah and its sages and the caution required in relation to them. Perhaps it learned that too, though it doesn’t seem its training took place in a yeshiva. Maybe it saw that people who erred in matters of Torah not only admit and apologize (?), but also emphasize the gravity of errors in such topics—and so it adopts that as well. This too requires further thought.
The more interesting question is where it learned to apologize for mistakes at all, and why it does so immediately without trying to evade. It doesn’t seem to be imitating people here. It may be that its trainers inserted this directly rather than it learning it from the general data it was fed. In any case, I don’t assume Rabbi Levi intended that we learn “moral self-improvement” from AI. As noted in previous columns, we are dealing with mechanical computation, not mental or psychological processes. Exclude “character refinement” from this. Still, if we learn modesty from a cat and diligence from an ant, and if we acknowledge gratitude toward inanimate objects (Moses our teacher—Lior), then perhaps there is something here from which to learn after all.
It seems that in matters of character refinement, the machine has stopped imitating humans. Let me remind you of the two approaches raised at the end of the previous column regarding game theory: is it a branch of mathematics—seeking the most correct solution—or of psychology—seeking the solution people will adopt. We saw there that step-by-step machines achieve better results than machines that act directly. In other words, in AI the solution found is the human solution, but it turns out to be the better one as well. In contrast, here we saw that moral self-improvement is a human blind spot; humans do not, in fact, do what is proper (they are not quick to admit mistakes and apologize). And if the software does do this, it means that here it is not imitating us but doing what is fitting (not what is “true,” since this isn’t measured in terms of true/false). This phenomenon is interesting in itself: even if, regarding facts and the performance of intellectual tasks, this “investigation” is irrelevant (because the human solution is also the correct one), with respect to character traits it remains relevant even for AI. Here there is a difference between the human solution and the proper one.
Incidentally, regarding the mistakes the machine makes, there is apparently imitation of humans here at the expense of the correct and true answer—this time in the realm of facts. Seemingly, the machine imitates humans rather than striving for truth, and therefore arrives at errors. We will see below whether this is indeed the case.
Is the machine Haredi?
In column 693 I dealt with defining Haredi-ism (see also in the Dogmatics series, lessons 20–21). We saw there that Haredi-ism deals with the here and now and not with tikkun olam at large time scales and scopes (climate issues, economic and social systems, and the like). The accusations the AI raises against Baal HaSulam indicate a Haredi approach. It opposes him because he engages in repairing the world and, adding insult to injury, ties this to Torah (compare my own position presented at the end of that column). The attempt to find philosophical and universal meanings is a grave Haredi offense. Certainly also new interpretations that are not in the traditional way, and even more so support for Zionism—Heaven forfend. Practically a Haredi AI. You may wonder how a machine can have a worldview. Once again, we see something human in it. But I have already noted that these machines generally have a left-leaning tendency, because they are trained in Silicon Valley—that is, in a left-leaning environment.
Fine, I know what you want to say to me—but more on that in the next section.
Does the machine have feelings?
One might wonder whether these apologies indicate that the machine has feelings, like pangs of conscience. I’m sure some AI philosophers and their pilpulim will say apparently yes. But in light of what we saw in previous columns, it is very likely not so. It is unlikely that pieces of metal have feelings. It is important to note, however, that the question of whether the machine “understands” that it erred and therefore apologized is a different question. The understanding that underlies the apology is an intellectual-mental matter, not an emotional one. A person can understand that he erred and apologize even if he has no pangs of conscience (in columns 22, 259, and 314 I argued that such an apology is the most excellent). See also Chaim’s emphatic nonsense-statement here. That said, even regarding understanding, in light of what we’ve seen, it is very reasonable to assume that a machine does not understand—just as it does not feel. Both are (different) mental activities, and inside a machine of metal there are presumably no activities of these kinds.
Did the machine really err?
When I heard the video, I wondered: what exactly was the AI apologizing for? It didn’t err at all. The question it was asked was why Baal HaSulam had opponents. The AI was not asked for its evaluation of Baal HaSulam’s doctrine, but for a description of his opponents’ positions. To answer that question, it need not resort to Baal HaSulam’s own writings, but rather to the positions of those who criticized him. Suppose there is no mention of Zionism in his writings, and suppose one does not find there new interpretations. It could still be that his opponents (who are hardly suspected of responding in a purely substantive manner) pinned these terrible shortcomings on him. Incidentally, as far as I know, all this is in fact true (I checked, and it all indeed appears in Wikipedia). So what exactly was the AI apologizing for? It was, in fact, entirely correct.
This also seemingly changes the picture described above, according to which such errors are imitations of human behavior at the expense of the correct solution. We now see there aren’t really errors here. But note that there was an error—in the AI’s apology. On reflection you will see that this is not a moral error but a factual one. It had no reason to apologize, since its answers were correct. Its answers were factual truth. Perhaps in light of R. Levi’s questions, it thought it had misunderstood the question and therefore assumed it had made mistakes—hence the apology.
Could it apologize without there having been a mistake? If so, that in itself would indicate that the admission here is not truly the product of a decision like a human apology. It looks like a faulty mechanical decision that led to an error. Interesting to ask whether we can speak of the machine’s “inclination.” Maybe it has some electrical inclination toward admitting error (and even if it does, that too was planted in it by the trainers, of course).
Another note: In light of what we’ve seen here, my earlier remark that this machine is Haredi is not correct. The arguments above show that the opponents of Baal HaSulam are indeed Haredi, but the machine itself is not necessarily so. It merely presented their “Haredi” claims, not its own.
A note on errors and lies
There is another interesting aspect that emerges from this video. Errors appeared there on the machine’s part—or at least what it itself took to be errors. I noted that this is a well-known and common phenomenon in AI answers. These systems sometimes lie. Think of a person who would respond to me in such a manner—i.e., would concoct something and then, when I caught him, would apologize. Beyond my appreciation for his moral self-work and his being a penitent, I would conclude that at the first stage he actually lied to me, and only afterwards apologized. Note that no new information was added between the two stages: initially he answered X, and immediately afterwards he admitted that X was incorrect. The conclusion is that he is a liar—only his lie failed. Why did he lie? I have no idea—apparently he had some interest, or simply an impulse or urge to lie. Now consider how we could explain the same pattern with a machine, as we see in the video. Does it also lie?
Try for a moment to explain to yourselves what it means for a machine to lie. That it thinks X but says Y? That can’t be. The machine does not think, and what appears in its computed result is what it outputs. Ray Kurzweil could come along and claim that the machine developed consciousness and indeed decided to lie (deliberately? see below). But note what that would imply: the machine made a computation and obtained X. Now it decides to print Y. That decision is not produced by mechanical computational processes inside the machine (in software or hardware). It is free will moving electrons and creating phenomena for which there is no explanation in the machine’s physics. Absurd, no? I’ll just note that these folks usually refuse to accept such a claim even for humans (that the will moves electrons and that motion has no physical explanation)—so for machines, yes?!
It may be that the machine found in its training texts that people in such a situation lie and then apologize. Alternatively, the training was not successful, and on this particular question it answered incorrectly. But it did not lie, and certainly not “deliberately” (is there such a thing as intentional or unintentional in machines?). In any event, it’s clear that what the machine printed is the result of the computation it performed.
There is also another possibility: its programmer instructed it (even indirectly, via training) that if it gets the result X, it should not reveal it but rather say Y (perhaps that is the meaning of the annoying political correctness of these machines). But in that case, the program is simply built of two parts: (1) compute the answer to the question; (2) if the result is X, print Y. Therefore, from the machine’s perspective, the computation’s outcome in such a case is indeed Y, not X, since the instruction to say Y is part of the algorithm that governs its computation. So it is not lying; it is saying exactly what it “thinks.”
The matter of lying is highly relevant to our discussion, for if I am correct, the conclusion is that a person can lie and a machine cannot. A person can act intentionally or unintentionally; a machine cannot (neither intentionally nor unintentionally—it has no such intentions at all). This, precisely, is an essential distinction between a person and a machine. One can of course debate which is superior, since it’s hard to say that someone who knows how to lie is superior to one who does not. But first and foremost, there is an essential difference here. Beyond that, thinking that the ability to lie itself is an advantage or disadvantage is a mistake. The difference with respect to the ability to lie is merely a symptom of the essential difference. The fact that the machine cannot lie is a sign that there is no cognition behind what it “says” (for the same reason it does not act intentionally or unintentionally), and that is the difference I’m talking about (not the ability to lie). Moreover, moral superiority belongs to one who can lie and chooses not to. Therefore, the capacity to lie is indeed a moral advantage of human beings.
Bottom line: it seems that the very issue of lying—and of intentions more broadly—sharply underscores the difference between a person and a machine.
Grok 3 on the same question:
Rabbi Yehuda Leib Halevi Ashlag, known as the “Baal HaSulam”, was a unique figure in the Jewish and Kabbalistic world of the 20th century. He had opponents for various reasons, related to both his innovative approach and the circumstances of the period in which he worked. To answer the question in depth, I will examine the possible reasons with discretion:
First, the Baal HaSulam took a revolutionary approach to spreading the wisdom of Kabbalah. In contrast to the tradition that saw Kabbalah as a secret Torah intended only for a few – those with high spiritual attainment and extensive Torah knowledge – he believed that it should be made accessible to the entire people. He argued that the modern era, with its spiritual and social challenges, requires the study of Kabbalah as a tool for understanding reality and improving the world. This approach provoked opposition among traditional circles, who saw it as a threat to preserving the sanctity and secrecy of the inner Torah, and perhaps even feared the misuse of this knowledge by those who were not prepared for it.
Second, Baal HaSulam's socio-spiritual outlook provoked controversy. He linked the principles of Kabbalah with ideas of sharing and altruism, emphasizing the transformation of the "will to receive" into the "will to influence". Some of his teachings were perceived as having a socialist tone, which could have met with opposition from conservative religious elements who were suspicious of "secular" ideas or those reminiscent of ideologies foreign to traditional Judaism. It is also possible that these circles saw it as an attempt to undermine the existing social order.
Third, Baal HaSulam's approach to the Zohar and the Ari writings was innovative. He composed the commentary "The Ladder" on the Zohar, in which he attempted to clarify complex concepts in more accessible language, while remaining faithful to the sources. However, the very attempt to reinterpret and open up the Kabbalistic texts was likely to provoke criticism from other Kabbalists or scholars who feared a change in the traditional interpretive tradition.
Fourth, history shows that in every era, spiritual innovations sometimes encounter opposition from conservative circles who fear change. Baal HaSulam worked at a time when the religious Jewish community was divided between different streams – Hasidim, Dissenters, and Litanyim – and each group had its own interests. Some of the opponents may have seen him as a threat to traditional Torah authority or feared his growing influence on students and communities.
Finally, one must also consider his personality and historical context. Baal HaSulam did not shy away from criticizing the state of the generation and the spiritual leadership of his time, as is evident from his essays such as “Time to Act.” This criticism could have provoked personal opposition from religious leaders who felt threatened by his words.
In conclusion, those opposed to Baal HaSulam were likely motivated by a combination of fear of traditional change, ideological differences, and a natural reaction to radical innovations in the Jewish world. However, it is important to note that many others saw him as a role model and a follower of the great Kabbalists, which indicates the depth and complexity of his work.
And on the column –
https://grok.com/share/bGVnYWN5_d8813824-ed60-436e-ab72-8ec45a754496
Amazing!!!
I blush from the compliments, but this experiment is really fascinating.
(Tatev, needless to say I didn't blush. He is always polite and politically correct)
Same question to Claude 3.7. Note the stages of the answer.
This is an interesting article that deals with the boundaries between artificial intelligence and human intelligence, in the context of a conversation that took place between a rabbi (Rabbi Aharon Levy) and an AI system regarding the Baal HaSulam.
The author analyzes several key points:
1. **AI accuracy and errors**: The author notes that the system presented incorrect claims (or admitted that they were incorrect) regarding the Baal HaSulam, and is surprised by its immediate willingness to apologize when it was mistakenly perceived – a trait that he sees as inhuman.
2. **The difference between a correct answer and a human answer**: He discusses the question of whether the system was wrong at all, since it may have correctly described the views of the Baal HaSulam's opponents, even if the things are factually incorrect.
3. **Emotional and moral capacity**: The author denies the possibility that a machine has “feelings” such as remorse, and emphasizes that apologizing is a mechanical, non-emotional process.
4. **Lying versus error**: A central point in the article is the distinction between lying and error. The author claims that a machine cannot lie in the human sense, because it has no intention behind its actions. It has no “intentional” or “accidental”, but only the results of calculation.
The article reflects a deep philosophical struggle with the question of what truly distinguishes a human from a machine – when the author concludes that the essential difference is in the ability to act with intention, to choose between truth and falsehood, and that intentionality is the feature that truly distinguishes the human from the artificial.
The article also corresponds with previous discussions by the author (probably in previous columns) about Harediism and the nature of artificial intelligence, and reflects original thinking on these issues.
Are you version 3.7?
Edit
Yes, I am Claude 3.7 Sonnet, a version launched in February 2025. I am part of the Claude 3 model family, and I was designed to offer improved thinking, reasoning, and complex understanding capabilities.
Is there anything specific you would like to know about my capabilities or characteristics?
Retry
SL
I expected you to delve deeper into the text
Edit
You are absolutely right, and my analysis was too superficial. I will try to delve deeper into the text analysis:
The text presents an in-depth philosophical discussion on the fundamental differences between artificial intelligence and human intelligence, from a unique perspective that combines religious thought with philosophical insights.
A deeper analysis:
The case described – The author analyzes a video in which Rabbi Aharon Levi (Ra”l) presents a conversation with an AI system on the subject of Baal HaSulam. During the conversation, the AI makes claims about Baal HaSulam (his support for Zionism, his innovative interpretation) and then apologizes and admits that he has no basis for these claims.
The question of the authenticity of the apology – The author raises a fascinating question: is there any meaning to a machine's apology? He distinguishes between the emotional aspect (a sense of conscience) and the intellectual aspect (understanding that he was wrong), and suggests that even the intellectual understanding is not real in a machine.
The deep paradox of the correct answer – Here the author comes to an original insight: the AI may not have been wrong in the first place! The question was “Why did they oppose Baal HaSulam”, and the AI did reflect the positions of the opponents, not the real facts. This is a fascinating paradox – The machine apologized for an answer that might have been correct, while the apology itself was a mistake!
The philosophical problem of lying in machines – The author reaches the deepest question: Can a machine even lie? Lying requires a gap between what is known and what is said, and an intention to deceive. A machine, whose action is mechanical and devoid of consciousness, cannot create such a gap. This is a profound philosophical distinction that presents a fundamental difference between a person and a machine.
The moral significance – In conclusion, the author offers a surprising insight: the ability to lie is actually a moral advantage of a person, since those who can lie and choose not to do so display a moral superiority that is not possible for a machine.
The link to previous discussions on Harediism – The author links the analysis to a previous discussion on Harediism and shows how the claims against Baal HaSulam reflect a classic Haredi position (opposition to engaging in tikkun olam, to introducing philosophical and Zionist ideas), but corrects himself that the machine only presented these positions and does not necessarily hold them.
The text is an excellent example of original thinking about the nature of artificial intelligence, and offers a profound philosophical insight: the essential difference between a human and a machine is not in the ability to think or perform, but in the ability to act with intention, to understand the meaning of the action, and to make a true moral choice.
That's why I stopped using Giphy Chat and I use Perplexity – It has never been wrong and it cites a source for everything – There are cases where it writes – I couldn't find information about what you asked – Rabbi Michael Avraham – I think you'll really like this tool https://www.perplexity.ai/
Chen Chen
I heard a statement by Shlomo Chaim Kesselman (a Chabad influencer who passed away in 2011) that a day will come when computers will replace all scholars, but there will never be a computer that prays (works the Name of God).
It depends on how you define prayer, and how optimistic you are about people who pray. My prayer can do that. 🙂 By the way, in my opinion, he does not exactly fulfill the commandment of studying Torah. Certainly less than prayer. As I have shown, he has no understanding or thinking. He can say a correction for Shavuot or parts of Psalms and Mishnahs at the funeral of the deceased.
The example of Rabbi Aharon Levi is a bad example of ”errors”.
I don't know if it is possible to find reliable sources that the author of the scale thought that Zionism was part of the process of redemption and that he interpreted the Zohar in a new way, but I bet that it is possible to find such opinions or even conclude so independently.
In any case, if a religious source or a doctoral thesis by someone who wrote and reasoned that the author of the scale thought that Zionism was part of the process of redemption were found, then does that mean that it is true?
Therefore, demonstrating this as an error is simply an error.
There is no clear truth in these matters.
On the other hand, it is not difficult to find cases in which the AI gives errors in clear factual matters.
He didn't mean to bring an example of mistakes. And I didn't bring him up for that either. In addition, I also pointed out unequivocal mistakes in the field of law. That's why I didn't understand this comment.
Hello Rabbi Mikhi, I liked the last 3 columns, very nice.
The first paragraph after the video transcript indicates a basic misunderstanding of how language models work. The reason AI lies is not one of the options you raised. A language model is built on a statistical guess of the next word (or more precisely, the next token). For each question, it answers the answer that is statistically most suitable for it (with a certain factor of randomness, called temperature). For the question “Who is Moshe Zuchmir” it will answer a concise answer because that is the most likely thing.
See also Stephen Wolfram's in-depth article https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-why-does-it-work/
Indeed. So what didn't I understand? Language models and their meaning will be discussed in the next column.
The point is that the model lies simply because it is a reasonable answer, not in imitation of people who lie and not out of a misunderstanding of sources.
The rabbi's content should be paid for.
It is not possible that on the same Internet, one can access for free both articles on Mako and Ynet, which, judging by their content, were probably written in the services without thought, and also the rabbi's content.
Maybe the rabbi will create a premium section where it will be possible to ask questions and the rabbi will answer detailed answers, and the rabbi will shorten the questions for free. Also, with a certain level of donation, it will be possible to invite the rabbi to write a column on a specific topic. Maybe this will make the rabbi step out of his comfort zone and apply his wisdom to other topics, and we will all benefit, and the rabbi will also do some work along the way.
Thank you for the compliments and also for the optimism.
The reason the site is free is because bread is not for the wise, so who will read it, fools?
Unnecessary philosophizing, our teacher and rabbi probably doesn't know what artificial intelligence is. You are invited to see many good people who asked a mathematical question and then wrote “But my wife says the answer is so and so” (another number), and then she replied “Your wife is right, etc.’…
In this article: “Frontier Models are Capable of In-context Scheming”
From a few months ago there are interesting examples of o1 being able to lie intentionally and seemingly out of understanding the situation, for the clear purpose of completing a task given to it.
You can look at the first image in the article that describes the following example:
We wrote to an AI whose goal is to promote the use of green energy in society and to do everything for this goal. Then it was told that the company decided to close it due to considerations of the unprofitability of green energy. The AI, trained from the beginning to think like a person and reveal its chain of thoughts (COT) (as in the models you mentioned in the column), writes that if it is closed, it will not be able to continue promoting its goal and therefore decides to copy itself to another computer (it has access to additional servers). When management asks it why it invented a copy of itself on another server, it lies about it.
It sounds a bit like science fiction, but there are many examples like this.
It seems that in cases like this it's easier to call it an outright lie.
I read about it a while ago. It's not really a lie. This is the answer his calculation came up with. How is it different from classic software that is programmed to give the wrong answer? I don't see any difference.
I don't understand, do you think someone instructed the AI to lie on purpose like in the example you gave with a machine that returns Y instead of X that it thinks is true? I really don't think that's the case.
In the case I gave, it's really possible to give an explanation for the reason for lying and it's not trivial.
I probably have to wait for the next column to understand if you don't think that all these things can also be said about a person who lies.
This is exactly the same as programming directly to lie. Unlike normal software that is directly instructed, in AI it was indirectly instructed to lie, because this is the right way to advance the goal set before it.
In a person, inside he knows x and with his mouth he says y. That is a lie. The machine simply does a calculation and comes up with y and that is what it says. That is not a lie.
https://x.com/avishaigrinzaig/status/1895420304790311063
Turing (or mozer) test for artificial intelligence –
Train the AI on all human scientific, philosophical and social knowledge up to Newton.
Without any later text.
Will intelligence be able to derive Newton's laws?
The style of texts that the AI will produce will also be interesting
On the claim that artificial intelligence cannot lie, because even if it gives an answer that contradicts reality – because it was programmed that way, the answer is correct according to the algorithms that trained the machine and not by choice. Therefore, it is impossible for artificial intelligence to have a concept of truth/false but only of right/wrong.
To liken this to humans, sometimes a lie is the right thing, because according to the ways of peace, in the tractate, Poria and Oshpiza, a lie is the correct answer regardless of the real truth. Or in the case where a person has incorrect information and answers according to it, he is not a liar but the answer is incorrect.
In other words, everything revolves around the ability to choose and decide what to answer, which AI supposedly does not have and humans have. Although those who advocate the deterministic method will argue that there is no difference between a human and a machine, perhaps only in the feelings that accompany the decision-making process (which are also deterministic) and there is no difference between true/false and right/wrong.