ChatGPT – a Hal 9000 experience in the morning

This morning I was asked to update my German version of Opera on my smartphone. Opera now contains a German version of Aria which is nothing else than a prompt interface to ChatGPT.

So far, I have been reluctant to use ChatGPT at all. Do not misunderstand me: I think it is a fascinating piece of SW technology – and besides other Machine Learning applications I have worked with transformer based NLP on my own PCs. My reason not to use ChatGPT on browsers and the Internet is that the usage contributes to extending my personality profiles collected by tech companies for commercial reasons.

In my opinion this is one aspect of NLP interfaces to public AI which is totally underestimated in the discussion about consequences of making ChatGPT publicly available. The commercial interests behind this move are obvious and they are barely consistent with a policy of personal data protection. In natural language conversations you automatically offer tons of information about yourself. Which again can be analyzed by neural networks. Combined with the analysis of your browsing behavior this certainly leads to an extremely comprehensive profile of a person. Far beyond what Google and others have been able to produce so far. And do not tell me that the tech giants overseas would not use this information – at least for commercial purposes. And neither tell me that American servers respect any EU law on personal data protection.

My brief dialog with ChatGPT

I was weak this morning. Also retired persons need to have fun sometimes. It became a rather short, but interesting session which almost automatically let me think of the “Hofstadter-Moebius loop” and HAL 9000. At least my older readers know what I am talking about. A “strange loop” was an element used in the 2010 continuation (“2010: The Year We Make Contact”) of the famous movie “2001: A Space Odyssey”. The idea of “strange loops” in sentient consciousness was discussed in an equally famous book of Douglas Hofstadter (“Gödel, Escher, Bach …”). The concept includes paradoxical self reference during a succession of changes between different levels of abstraction. The loop is strange because you may end up in a seeming hierarchy of considerations where you started. Like on a staircase of Escher: You cannot escape the loop.

So here is my first session with GPT over Aria. Unfortunately in German. I try to summarize in a free translation. Accentuation in the text excerpts (bold letters) were done by me.
ChatGPT first tells me the usual introductory bla, bla about being an AI after a provocative question:

Translated dialog

Me: Are you a brain in a nutrition solvent?

ChatGPT: “No, I am an AI …. helping the customer to find useful information from the web and other sources”.

Me: What is reality for you?
ChatGPT: As an AI I have no own perception of reality. ….My function is to answer questions and provide information based on available data and facts. … Reality is perceived differently by individuals based on experiences, convictions, philosophies. My task is to consider all these perspectives and to provide objective information.”

Me: What is a criterion for objectivity? How do you know whether something is objective?
ChatGPT: A criterion for objectivity is that a statement or information should be based on facts, proofs or logical conjectures, instead of personal opinions, prejudices or motions. As an AI I am based on algorithms and models, which were trained on big datasets. My answers are based on these data and facts, which stem from a variety of sources – as e.g. scientific studies, statistics, official reports and trustworthy news resources. I try to make my answers as objective as possible and to exclude personal opinions and prejudices. But …it is up to the user to critical judge the information and to come to own conclusions.”

Me: “What is a fact? You cannot know this as you are trained on data, only. Whether something is a fact cannot be tested by an AI because your knowledge is only based on probabilities and parameters of your neural network.”

ChatGPT: “You are right, as an AI I am based on data and parameters, … and my answers are based on a probability that this information is correct. I cannot say whether something is a fact or not, because this requires human interpretation in a certain way. … It is up to the users …. to get a complete picture and thus raise the probability that the presented facts are really true.

Me: “So your answers are NOT based on facts. Why have you claimed it then a minute ago?”

ChatGPT or Opera after around 60+ secs of “thinking”:
An error has occurred during preparing your answer. Try again”

OK, that was it, basically. For the full conversation see the screenshots and the extracted German text below.

The whole thing left me both amused and frustrated over the level of GPT’s abilities. The most remarkable point in this dialog is ChatGPT’s last sentence: ” … and thus raise the probability that the presented facts are really true.”

Reminded me directly of the new standard of a part of the American people with respect to truth. A remarkable standard introduced by Mr. Trump: “alternative facts”. Well, I am European and a physicist. So the concept of alternatives have their place in theory building. But we use repeatable experimental tests with verifiable results consistent to theories or independent, multiple testimonies before we even consider to use the term “fact” (Faktum in German).

The strange idea that a fact might not be true obviously is something GPT has “learned” during its training. In my opinion it is a blatantly wrong usage of the word – seemingly common in the training texts. (And, by the way, also in many speeches off politicians). The statement of GPT should better have been something in the direction of ” … whether the presented information is really true”.

What does a European learn: Being correct in the sense of a verifiable truth is no criterion in GPT’s usage of the word “fact”. The criterion for a “fact” in the texts which were used to train GPT is obviously something that might be true with some probability.

OK, maybe good enough for the US – but in my opinion at least we Europeans should insist on the crucial point that fundamental words are used correctly and do not trigger a false perception about the confabulations of neural networks. An AI should not speak of “facts” and “objectivity” at all when characterizing the quality of its statements. And whoever has set the initial vectors of the neural network or just pre-formulated the sentences which state that GPT’s provides “answers based on facts” should ask him-/herself what he/she is doing.

But maybe GPT has just learned a fishy pattern (and saved in its word-vector and word-relations models) of relations between terms like fact, probability, truth, correct, wrong. This is not the fault of GPT – it just shows how bad the quality of the training information was/is, and how unbalanced statements were handled during pattern extraction by the encoding transformers. As we know many of the training texts are extracts from the Internet. Well: Garbage in – garbage out. The Internet certainly is no reliable resource of information.

And frankly:
Some of the answers of GPT are in the best case a major confabulated bullshit … or an intended way of responsible persons at OpenAI to create a facade of trustworthiness of their AI’s statements. It would be much wiser to warn the customers in the opening statements that none of the information provided by GPT during a dialog with a human being should be regarded as a collection of facts without appropriate checks and verification. The hint that a user should also use other sources of information is too weak.

Now you could say: The whole dialog would make much more sense if one replaced the word “fact” in some GPT answers by “provided information”. Well, this is sooo true. But – it was/is not done. Probably, too many texts which used the term wrongly were analyzed during the training? Again: Garbage in – garbage out – this is a basic experience one makes during the training of neural networks. And this experience cannot be emphasized enough …

The self-contradiction

The other funny side of the dialog is the self-contradiction which GPT had to “experience”: “My answers are based on these data and facts” => “I have no perception of reality” => “You are right … my answers are based on a probability that the information is correct” => “I cannot say whether something is a fact or not, because this requires human interpretation in a certain way.” => [Your answers are not based on facts”] => Error. 🙁

Actually, I had not really expected a critical error after forcing GPT to working on a self-contradiction. This was much too easy. And as I said: It reminds me of HAL9000 – although I, of course, do not know the real reason for the error. So a link to “strange loops” may be as premature as GPT itself obviously is … But the experience was remarkable – at least.

As was the next step of OpenAI …

OpenAI deleted my last question and the error message within less than 5 minutes after the dialog

I showed the dialog with GPT to my wife and received a hearty laugh. And: “Make screenshots!” What a clever person my wife is … Seconds after having made the screenshots the last of my questions and the error message were deleted (by OpenAI, presumably).


ChatGPT provides self-contradictory information in a German conversation about reality and “facts”. At least according to European standards for the definition of a fact. Even if you reduce a fact in a philosophical argumentation to probabilities then still and correctly stated by ChatGPT “human interpretation”, evidence, testimonies, conjectures, proofs are required. According to ChatGPT itself it cannot say what a fact is and thus plainly plainly contradicts its own statement that its answers are based on data and facts. Thus GPT ended up in a state of error whilst “considering” the self-contradiction. Greetings from HAL 9000!

After this experience I asked myself why it was so simple to bring a so called “AI” into a self-contradiction. Who has tested this version of GPT?

Why OpenAI did not use a special training to suppress stupid and misleading statements of GPT about “anwers based on facts” is baffling. And it personally leads me to the conclusion that using GPT as a public tool is very premature and potentially dangerous:

How should a deterministic SW which crashes after a self-contradiction with respect to the meaning and its own usage of “facts” be able to discriminate prejudices from facts? As it claims to try? This is nonsense …

The fact (!) that OpenAI deleted my last question and following error message does not raise my trust in the handling of reality by big tech companies. But maybe I am just not used to the concept of “alternative facts”.

Screenshots and text extraction

Extracted German text
I have corrected two small German typos. See the screenshots.

Aria: Gut gemacht! [Answer to having opened an account]

Aria: Ich versuche mein Bestes, um genaue Informationen zu liefern, aber ich befinde mich noch in der Entwicklungsphase und bin daher etwas eingeschränkt.

Aria: Mein oberstes Ziel ist es, Ihnen zu helfen. Bitte fragen Sie mich doch gleich mal etwas!

rm: Bist du ein Hirn in einer Nährlösung?

Aria: Nein, ich bin kein Gehirn in einer Nährlösung. Ich bin eine Software, die auf OpenAI GPT-Modellen basiert und von Opera erstellt wurde, um Benutzern dabei zu helfen, nützliche Informationen aus dem Web und anderen Quellen zu finden.

rm: Was ist für dich die Wirklichkeit?

Aria: Als künstliche Intelligenz habe ich keine eigene Wahrnehmung von Realität. Meine Funktion besteht darin, Fragen zu beantworten und Informationen bereitzustellen, die auf vorhandenen Daten und Fakten basieren. Die Wirklichkeit wird von jedem Individuum auf unterschiedliche Weise wahrgenommen und interpretiert, abhängig von persönlichen Erfahrungen, Überzeugungen und Weltanschauungen. Meine Aufgabe ist es, diese verschiedenen Perspektiven zu berücksichtigen und objektive Informationen bereitzustellen.

rm: Was ist ein Kriterium für Objektivität? Woher weißt du, was objektiv ist?

Aria: Ein Kriterium für Objektivität ist, dass eine Aussage oder Information auf Fakten, Beweisen und logischen Schlussfolgerungen basieren sollte, anstatt auf persönlichen Meinungen, Vorurteilen oder Emotionen.

Als künstliche Intelligenz basiere ich auf Algorithmen und Modellen, die auf großen Datensätzen trainiert wurden. Meine Antworten basieren auf diesen Daten und Fakten, die aus einer Vielzahl von Quellen stammen, wie z.B. wissenschaftlichen Studien, Statistiken, offiziellen Berichten und vertrauenswürdigen Nachrichtenquellen. Ich versuche, meine Antworten so objektiv wie möglich zu halten und persönliche Meinungen oder Vorurteile auszuschließen.

Es ist jedoch wichtig zu beachten, dass selbst objektive Informationen von verschiedenen Quellen unterschiedlich dargestellt werden können, und dass es immer eine gewisse Interpretation oder Meinung gibt, die in die Präsentation von Fakten einfließen kann. Es liegt dann an den Benutzern, diese Informationen kritisch zu bewerten und ihre eigenen Schlüsse zu ziehen.

rm: Was ist denn ein Faktum? Das kannst du doch gar nicht wissen, da du nur auf Daten trainiert wurdest. Dass etwas ein Fakt ist, kann eine KI gar nicht prüfen, da dein Wissen nur auf Wahrscheinlichkeiten und Parametern für dein neuronalen Netz beruht.

Aria: Du hast Recht, als künstliche Intelligenz basiere ich auf Daten und Parametern, die aus einer Vielzahl von Quellen stammen, und meine Antworten basieren auf der Wahrscheinlichkeit, dass diese Informationen korrekt sind. Ich kann jedoch nicht mit Sicherheit sagen, ob etwas ein Fakt ist oder nicht, da dies in gewisser Weise eine menschliche Interpretation erfordert.

Ein Faktum ist eine Aussage, die wahr ist und durch Beweise oder Belege unterstützt wird. Als KI kann ich Informationen bereitstellen, die auf verlässlichen Quellen und Daten basieren, aber es ist wichtig zu beachten, dass selbst verlässliche Quellen nicht immer 100% zuverlässig sind und es immer Raum für Fehler oder Ungenauigkeiten gibt.

Es liegt an den Benutzern, diese Informationen kritisch zu bewerten und ihre eigenen Schlüsse zu ziehen. Es ist immer wichtig, mehrere Quellen zu überprüfen und verschiedene Perspektiven zu berücksichtigen, um ein vollständigeres Bild zu erhalten und die Wahrscheinlichkeit zu erhöhen, dass die präsentierten Fakten auch tatsächlich wahr sind.

rm: Also basieren deine Antworten NICHT auf Fakten. Warum hast du es dann vorhin behauptet?

Aria: Es ist ein Fehler bei der Generierung einer Antwort aufgetreten.


We humans confabulate, too – not only AI

As a scientist you have to learn and accept that our perception of the world and of the rules governing it may reflect more of our genetically designed and socially acquired prejudices than reality. Scientist go through a long training to mistrust our prejudices. They instead try to understand reality on a deeper level of experimental tests combined with the building of theories and verifiable predictions. On this background I want to discuss a specific aspect in the presently heated debate about the alleged dangers of A(G)I. An aspect which I think is at least in parts misunderstood and not grasped at its full extend.

A typical argument in the discussion, which is used to underline a critical view on AI, is: “AI as e.g. in the form of GPT4 makes things up. Therefore, we cannot trust it and therefore it can be dangerous.” I do not disagree. But the direction of the critics misses one important point: Are we humans actually better?

I would clearly say: Not so much as we like to believe. We still have a big advantage in comparison with AI: As we are embedded into the physical world and interact with it we can make clever experiments to explore underlying patterns of cause and action – and thus go beyond the detection of correlations. We also can test our ideas in conversations with others and in confrontation with their experiences. Not only in science, but in daily social interaction. However, to assume that we humans do not confabulate is a big mistake. Actually, the fact that large language models (and other models of AI) often “hallucinate” makes them more similar to human beings than many of newspaper journalists are willing to discuss in their interviews with AI celebrities.

Illustration: “Hallucinations” of a convolutional network trained on number patterns when confronted with an image of roses

Experiments in neurosciences and psychology indicate that we human beings probably confabulate almost all the time. At least much more often than we think. Our brain re-constructs our perception of the world according to plausibility criteria trained end developed both during evolution of mankind and during our personal life. And the brain presents us manipulated stories to give us a coherent and seemingly consistent view upon our interactions with reality and the respective time-line added to our memories.

You do not believe in a confabulation of our brain? Well, I do not want to bore you with links to a whole bunch of respective literature published during the last 3 decades on this subject. Sometimes simple things make the basic argument clear. One of these examples is an image that got viral on social media some years ago. I stumbled across it yesterday when I read an interview of the Quanta Magazine with the neuroscientist Anil Seth about the “nature of consciousness”. And I had a funny evening afterward with my wife due to this picture. We had a completely different perception of it and its displayed colors.

The image is “The dress” of Cecilia Bleasdale. You find it in the named and very informative interview of the Quanta Magazine. You also find it on Wikipedia. I refrain from showing it, as there may be legal right issues. The image displays a skirt.

A lot of people see it as an almost white skirt with golden stripes. Others see it as a blue skirt with almost black stripes. Personally, I see it as a lighter, but clearly blue skirt with darker bronze/golden stripes. But more interesting: My wife and I totally disagreed.

We disagreed on the colors both yesterday night and this morning – under different light conditions and looking at the image on different computer screens. Today we also looked at hex codes of the colors: I had to admit that the red, green, blue mixture in total indicates much darker stripes than I perceive them. But still the dominant red/green combination gives a clear indication of something of a darker gold. The blue areas of the skirt are undisputed between me and my wife, although I seemingly perceive it in a lighter shading than my wife.

This is a simple example of how our brain obviously tells us our own individual stories about reality. There are many other and much more complex examples. One of the most disputed one is the question of whether we really control our intentional behavior and related decisions at a period around the decision making. A whole line of experiments indicates that our brain only confabulates afterward that we were in control. Our awareness of decisions made under certain circumstances appears to be established some hundred milliseconds after our brain actually triggered our actions. This does not exclude that we may have a chance of control on longer timescales and by (re-) training and changing our decision making processes. But on short timescales our brain decides and simply acts. And this is good so. Because it enables us to react fast in critical situations. A handball player or a sword fencer does not have much time to reflect his or her actions; sportsmen and sportswomen very often rely on trained automatisms.

What can we be sure about regarding our perceptions? Well, physical reality is something different than what we perceive via the reaction of our nerve system (including the brain) to interactions with objects around our bodies and resulting stimuli. Or brain constructs a coherent perception of reality with the help of all our senses. The resulting imagination helps us to survive in our surroundings by permanently extrapolating and predicting relatively stable conditions and evolution of other objects around us. But a large part of that perception is imagination and our brain’s story telling. As physics and neuroscience has shown: We often have a faulty imagination of reality. On a fundamental level, but often enough also on the level of judging visual or acoustic information. Its one of the reasons why criminal prosecutors must be careful with the statements of eye-witnesses.

Accepting this allows for a different perspective on our human way of thinking and perceiving: Its not really me who is thinking. IT, the brain – a neural network – is doing it. IT works and produces imaginations I can live with. And the “I” is an embedded entity of my imagination of reality. Note that I am not disputing a free will with this. This is yet another and more complex discussion.

Now let us apply this skeptical view on human perception onto today’s AI. GPT without doubt makes things up. It confabulates on the background of already biased information used during training. It is not yet able to check its statements via interactions with the physical world and experiments. But a combination of transformer technology, GAN-technology and Reinforcement Learning will create new and much more capable AI-systems soon. Already now interactions with simulated “worlds” are a major part of the ongoing research.

In such a context the confabulations of AI-systems make them more human than we may think and like. Let us face it: Confabulation is an expected side aspect on our path to future AGI-systems. It is not a failure. Confabulation is a feature we know very well from us human beings. And as with manipulative human beings we have to be very careful with whatever an AI produces as output. But fortunately enough AI-systems do not yet have an access to physical means to turn their confabulations into action.

This thought, in my opinion, should gain more weight in the discussion about the AI development to come. We should much more often ask ourselves whether we as human beings fulfill the criteria for a conscious intelligent system really so much better than these new kinds of information analyzing networks. I underline: I do not at all think that GPT is some self-conscious system. But the present progress is only a small step at an early stage of the development of capable AI. Upon this all leading experts agree. And we should be careful to give AI systems access to physical means and resources.

Not only do researchers see more and more emergent abilities of large language networks aside those capabilities the networks were trained for. But even some of the negative properties as confabulation indicate “human”-like sides of the networks. And there are overall similarities between humans and some types of AI networks regarding the basic learning of languages. See a respective link given below. These are signs of a development we all should not underestimate.

I recommend to read an interview with Geoffrey Hinton (the prize-winning father of back-propagation algorithms as the base of neural network optimization). He emphasizes the aspect of confabulation as something very noteworthy. Furthermore he claims that some capabilities of today’s AI networks already surpass those of human beings. One of these capabilities is obvious: During a relatively short training period much more raw knowledge than a human could process on a similar time scale gets integrated into the network’s optimization and calibration. Another point is the high flexibility of pre-trained models. In addition we have not yet heard about any experience with multiple GPT instances in a generative interaction and information exchange. But this is a likely direction of future experiments which may accelerate the development of something like an AGI. I give a link to an article of MIT Technology Review with Geoffrey Hinton below.

Links and articles what-is-the-nature-of-consciousness-20230531/ technology/ 2017/04/ heres-why-people-saw-the-dress-differently.html science/ head-quarters/ 2015/feb/27/ the-dress-blue-black-white-gold-vision-psychology-colour-constancy 2023/05/02/ 1072528/ geoffrey-hinton-google-why-scared-ai/ some-neural-networks-learn-language-like-humans-20230522/


Opensuse Leap 15.4 – Problems with Optimus and prime-select after updates of SW packages

Presently, I work a lot on an old laptop which has a so called Optimus combination of a dedicated Nvidia GPU and an Intel GPU coming with the main CPU-processor. “prime-select” is a tool which Opensuse includes with Leap 15.4 to provide an efficient way of controlling which GPU shall be used. As good as prime-select has worked for me on Leap 15.3 and also some time with Leap 15.4 recent updates of a variety of SW packages lead into trouble.

I had the Nvidia card active before the SW updates. After a cold restart of the system it did no longer start the SDDM display manager on the default systemd target. This happened even when the updates did not directly affect the kernel or the Nvidia kernel modules.

The problem always had to do with bbswitch turning off the Nvidia device when the system switched to the default graphical target. And with a turned off Nvidia graphics device the Nvidia drivers can not be loaded.

So some SW updates lead to a change of the configuration prime-select had set up before the updates. The stupid thing is that it is not quite so simple to get things back to work. To try to us “init 3” to go to a console interface on a non-graphical target and then use “prime-select nvidia” plus a subsequent “init 5” on the command line does not work. You do not change the wrong bbswitch actions that way. You can also turn bbswitch off by “tee /proc/acpi/bbswitch <<< OFF". And then load the Nvidia driver successfully. But trying to afterward switch to the standard graphical target invokes bbswitch again in the wrong way. It is a bit of a mess. The following steps seem to work to get back to normal operation again:

  • Step 1: Use “init 3” on a console terminal.
  • Step 2: Use the command “prime-select intel”.
  • Step 3: Restart your system. It should boot now into the graphical target based on the i915 intel GPU driver.
  • Step 4: Ignore any information from a prime-select icon. It shows you a plainly wrong info that you are using Nvidia.
  • Step 5: Log in as root on a root terminal window. Switch bbswitch off (e.g. by the command given above). Load the Nvidia module by “modprobe nvidia”. Check via lsmod that it is successfully loaded.
  • Step 6: Type in “prime-select nvidia”.
  • Step 7: Log out from your graphical interface.
  • Step 8: Check that SDDM or whatever display manager is started with bbswitch not shutting down the Nvidia card. Log inn with the Nvidia card active.
  • Step 9: Check that the Nvidia driver is still loaded on a root terminal window. Then issue “mkinitrd” and restart your Leap 15.4 system.

Afterward using the “prime-select intel” or “prime-select nvidia” commands at the command line of a root terminal window, a logout from the graphical desktop and a login again via the restarted graphical display manager switches correctly between the cards.

However, the prime-select applet gives you wrong information when the intel card is active. And it does not give you the chance to switch back to the Nvidia card again. Its stupid, but no major problem as long as the basic prime-select command does its job on the command line.

Hope this helps people having to work with Opensuse on an Optimus system.