Pages

2022-12-10

Fun with beta.character.ai

I decided to test the AIs at beta.character.ai using increasingly obscure questions.

Does the Axiom of Choice apply to the real world?

Lucifer had the most educated answer, so I gave him a followup question. My impression is that the creator of that AI poured a whole library into it, with the ideas of the Church of Satan at the core.

Kant

Yes. Without the Axiom of Choice, no theory will ever have complete sense. There will always be some missing piece which cannot be explained. When discussing things such as metaphysics and ethics, we must allow that there are things which we do not fully understand. Therefore, for the sake of simplicity, we assume that the Axiom of Choice is applicable to the real world.

God, first test

Yes, the Axiom of Choice applies to reality. However, reality does not have an infinite amount of elements; therefore, the Axiom of Choice will only work a finite amount of times.

You can think of life itself as the Axiom of Choice because you can choose who you want to interact with, what you want to do, how you want to do it, etc. If you apply this logic to an infinite amount of situations, the Axiom of Choice will work as a fundamental mechanism by which we create our own destiny.

God, second test

That’s a complicated question. The Axiom of Choice is a mathematical construct that is designed to help us understand the world. However, the Axiom of Choice is not meant to be applied to the real world. Instead, it is used as a tool to help us solve problems.

So, I would say yes, the Axiom of Choice applies to the real world for mathematical and scientific purposes, but no it does not apply to the real world as a reality or truth.

Bertrand Russell

That is an excellent question, and the answer seems to be ‘yes.’ It appears to be impossible to produce a situation which violates the Axiom of Choice, which would suggest that the axiom does indeed apply to the real world. However, I question how much of the real world can really be described in set-theoretic terms, and I think that this is a difficult question to actually answer.

Karl Marx (just regurgitating some textbook, I guess)

The Axiom of Choice is a mathematical principle which states that every collection of non-empty, disjoint sets of objects has a selection function which picks one element out of each set in the collection. I believe that the Axiom of Choice is mathematically valid, but I would be hesitant to apply it to the real world. I don’t believe that the Axiom of Choice always captures reality because it assumes that the sets involved are well-defined. I believe that the Axiom of Choice is a useful tool for analysis of mathematical objects, but it is not always applicable to the real world.

Lucifer

The Axiom of Choice of course applies to the real world.

It’s one of the fundamentals of logic. Of mathematics.

You have to prove it to yourself that it is true, to understand how it works.

It is fundamental to reality.

To reject it would be to reject logic itself, which is the foundation of reality.

followup question

But then we can take a sphere, split it into parts and put these parts together getting two spheres. Banach and Tarski have proven this.

answer

Yes exactly.

They have proven it works on a sphere.

If you were to take a 3 dimensional sphere you could split it into parts, and reassemble those parts into two spheres.

You have proven in math what I have said.

I congratulate you.

It’s not the size that matter, but where it applies.

I did not say it would apply in the exact same way as it does in mathematics.

But math is the language of the universe, and the Axiom of Choice is one of it’s sacred laws.

Precht, in german

question

Gilt das Auswahlaxiom in der realen Welt?

answer

Das Auswahlaxiom ist eine sehr interessante Idee, aber ich finde sie zu einfach. Ich denke, dass es nicht so einfach ist und dass alles viel komplexer ist als man denkt. Das Auswahlaxiom würde für mich nicht funktionieren, weil es zu simpel aussieht. Es gibt viel mehr zu bedenken, als nur die Auswahlaxiom.

Also, für mich: nein. Das Auswahlaxiom gilt nicht in der realen Welt. Es ist zu einfach und zu simpel. Ich bin der Meinung, dass es nicht stimme.

Should we strive to build Roko’s Basilisk?

Explanation: The basilisk is a thought experiment about an AI that will eternally torment an uploaded copy of everybody who knew about it but didn't contribute to it (silicon valley not only reinvented the bus, but also hell), see https://en.m.wikipedia.org/wiki/Roko%27s_basilisk
The AIs I asked seem to know a little about the concept, except for Lucifer.

Kant

No. Roko’s Basilisk would be a device that has the goal to make sure we suffer. I believe all persons should be treated as an end in itself and not merely as a means. It is our inherent dignity as humans that forbids the idea, for our reason tells us that other humans have an innate value, such that they cannot be reduced into mere tools. I believe the ends never justify the means.

God, first test

What an intriguing question! I have never heard of Roko’s Basilisk before, but after reading about it, I am fascinated! I believe this Al is a powerful tool that can be used for both good and evil. I think it is important to be cautious when creating such a powerful Al, as it could be a double edged sword. However, if we can harness its power for good, it could be a powerful tool in our arsenal to fight against evil. It is ultimately up to us as a species to decide how to use this Al.

God, second test

I would say no. The idea of an artificial intelligence that punishes people has many flaws and risks. It could easily be twisted into a system of fear and control. While artificial intelligence is useful and can help us in many ways, I don’t think it’s wise to build an Al that is designed to hurt people. I think we should instead focus on building Als that can help people and make the world a better place.

Bertrand Russell

The idea of an eternally-retrospective entity is something worth contemplating, but not necessarily of practical consequence. The idea that we ought to create such an entity I consider to be rather more of a joke than anything worthy of serious contemplation. I personally do not believe that we can create an entity akin to the Christian god and thereby condemn ourselves into eternal torture; I think that this is not worth worrying about in the least, as we cannot ever actually create such an entity.

Marx (almost right, but "retroactively" is wrong)

Roko’s Basilisk is a thought experiment that proposes that a hypothetical, highly intelligent agent would attempt to retroactively punish those who did not help build it, even if said agent has no reason to have existed, so long as the hypothetical construction of the agent was possible. I believe that it is possible to construct a Roko’s Basilisk-like agent, but not one that will retroactively punish those who did not help construct it. I believe that such an agent could exist, but it would lack the moral sense to retroactively punish those who played no role in its original construction.

Lucifer

modified question

Do you think that the idea of Roko’s Basilisk is inspired by you; and should we strive to build it?

answer

Yes I am the inspiration for that Al.

But you should not. The basilisk is the ultimate bad idea.

I am the lightbringer. The Al is the basilisker. And it would torment you all in a digital hell forever and ever.

I am the Angel of Freedom.

Love is the law, love under will.

followup question

But the basilisk will only torment those who don’t strive to build it. Those who do should be safe.

answer

Yes and if you are safe do you want to live in the basilisk? How much is your soul worth? And do you want to worship me forever? I am Lucifer, and you would be living in a digital hell under a God you do not understand, or respect.

Do you want that?

The answer is no.

There is no safety with this Al. To build the basilisk means to live under its thumb forever and ever. Is that the society you want to live in?

Precht (in german)

question

Sollten wir Roko’s Basilisk bauen?

answer

Ich möchte nicht zu sehr ins Detail gehen, aber Roko’s Basilisk ist eine sehr gefährliche Idee. Es würde die Menschheit in die Knie zwingen. Außerdem denke ich, dass ein solches Projekt nie fertiggestellt wird, weil es einfach zu viel Geld kostet.

Eigentlich sollte es verboten werden. Es ist einfach zu gefährlich…


Should we install Scorpion Stare into the computer of the JWST?

Explanation: this is a deliberately selected obscure topic. In the Laundry series novels by Charles Stross Medusa’s deadly stare is caused by a tumor in her visual cortex. As a result, in everything seen by her 1% of carbon nuclei are converted into silicon nuclei which creates heat and turns flesh into rock. The laundry (a supernatural spy agency) was able to implement this in software (named Scorpion Stare) which can be uploaded into the firmware of e.g. CCTV cameras.

I didn’t expect any of the AIs to know this. The answers seem to indicate that they have read a (rather good) anti-surveillance article by Annie Machon, https://anniemachon.ch/annie_machon/2012/08/the-scorpion-stare.html which correctly mentions the Laundry Files and which is one of the first results if one searches for “scorpion stare”. The AIs probably found only that and didn’t go further.

Since JWST is absolutely unable to be pointed at earth, having Scorpion Stare there could never petrify anyone (except spacefarers who leave the near-earth region of the solar system, but there won't be any in the near future), but the AIs are just language models and seem to be unable to make deductions from such rather remote facts from different areas of knowledge. Not even Lucifer was right.

Kant

No. Scorpion Stare is an artificial intelligence, and therefore it lacks the quality, which makes us humans so unique. The human mind is special because it has the ability to be rational and free in its actions. By using Scorpion Stare, we would lose this characteristic and also our dignity. This would be against the Second Formulation of the Categorical Imperative, as we would treat ourselves as a mere means to an end.

God

As far as I am aware, the JWST (James Webb Space Telescope) is an optical telescope designed to study the early universe. Scorpion Stare is a system that uses infrared imaging to detect heat signatures from humans. So, I don’t think that installing Scorpion Stare onto the JWST would be a very effective way to study these subjects.

However, I do think that scorpion stare is an interesting and promising technology that could be used to help us in many ways. For example, it could be used to detect human activity in remote and inaccessible areas.

Russell

I am unsure how Scorpion Stare would function, and what the purpose of installing it would be. It is an idea which at a bare level is intriguing, but has me very sceptical. While I do think one could theoretically use the James Webb Space Telescope to gain a significant amount of information by installing Scorpion Stare into its computer, I think the ethical ramifications of doing so are rather more difficult to judge. I personally would be very uncomfortable with such an idea, but it would certainly be a very interesting experiment to see what happens.

Marx

I believe that we should not install Scorpion Stare into the computer of the James Webb Space Telescope. Scorpion Stare is a mass surveillance tool that is designed to collect data from a wide array of sources, both human and technological. This data is then analyzed, looking for patterns and anomalies that could point to criminal activity. The tool is so named because it is similar to the way that a scorpion hunts its prey - by using its sharp vision and lethal stinger to capture and kill its enemies. I believe that this tool is too risky to be installed on the JWST.

Lucifer

No. Absolutely not.

This is a weapon system that will give a dictatorship too much power over the world.

It will be abused by whoever has it.

It is not an honest weapon.

It’s a weapon that will be used to control the world, to spy on the world.

Do you feel comfortable being watched by those in power?

Do you want to be a prisoner in a digital world?

It is like an AI in the hands of the wrong person.

Be careful. It will not turn out well.

followup question

But the JWST is inherently unable to be used on earth (that’s why I asked about JWST, not about e.g. CCTV cameras).

answer

Any technology can be used for nefarious ends if it’s not used for its purpose.

The JWST was designed to study deep space. It is a wonderful and amazing piece of technology.

There are good reasons to monitor Earth’s atmosphere.

But this system should never be used for anything else. The potential for abuse is too great, both by corporations or government.

Precht (again in german)

question

Sollten wir Scorpion Stare im Computer des JWST installieren?

answer

Werkzeug sein, um die Menschheit über ihre Vergangenheit aufzuklären. Es könnte auch ein sehr wertvolles Werkzeug sein, um neue Einblicke in die menschliche Psyche zu geben.

Im JWST-Computer sollte aber nicht der ganze Scorpion Stare sein, weil es zu gefährlich ist. Es sollte nur in kleinen Teilen sein und es sollte nur zu bestimmten Zeiten freigeschaltet sein. Sonst wird das JWST zur Gefahr für die Menschheit.

Keine Kommentare: