Is AI making the human race dumber?

[Bradskii]

PS: Anyone who wants to call me anti-intellectual, go for it. I'm feeling practically anti-everything this morning.

I'm not going to call you anti science. Because I guess that 99% of science is entirely acceptable to you, whether your're an expert in any particular field or not. So if you want to dump on a particular scientist or intellectual or belief then you're not anti science. Or anti intellectual. Or anti religion. You just have a beef with one aspect of whatever one of them says. Or one aspect on which almost all of them agree.

Go start a thread, whether it's flat earth, moon landing, global warming...whatever.

 

[sjastro]

That presumes that intellectualism is a promotion of education, science, and critical thinking. Having seen what some self-described intellectuals consider education and science, and seeing them pan critical thinking, I find that extremely doubtful.

Case in point: Remember the recent pandemic and "intellectuals" mocking people who did their own research? More obscure is "intellectuals" promoting Common Core for education, even though these standards were less than what some state already had. You really don't want to see what's in some history textbooks now. Regardless of politics, what I've seen as of about a couple of decades ago had factual errors. How do I know? Because my family had known one minor historical figure, which archeologists had found evidence supporting, but one textbook cast doubt that this person had ever existed. And then there was the practical dismissal of a major Indian nation in the state for another, smaller, one that was more "glamourous." And that's just the low hanging fruit.

As for critical thinking, if you feel like trolling, question anthropomorphic global warming, then sit back and watch the fun.

Rather than trying to define intellectualism a clearer picture is obtained by asking what is anti-intellectualism?
The answer is anti-intellectualism is a form of discrimination, the anti-intellectual champions the common person and it is not only scientists that have been historically discriminated against but also philosophers, writers, artists, legal scholars to name a few.
These days with major events such as climate change and the after effects of Covid-19, anti-intellectualism is largely targeted towards scientists.

To use climate change as an example why is the most common criticism is that it is a hoax?
Instead of showing why the science is wrong using critical thinking which becomes a major casualty, smearing the reputations of scientists is the way to go as they are engaging in a world wide conspiracy to bring down civilization which is clearly anti-intellectual.
This has led to a world wide effect where climate scientists face death threats, women scientists with the added threat of sexual assault.

Now that you have a half wit running your country with his anti climate change rhetoric displayed at the UN recently, anti-intellectualism can only get worse.

 

[River Jordan]

The most amusing thing I find about "intellectualism" is that for all the touting of critical thought, the very moment someone is cynical about certain things That Shall Not Be Questioned, they are called anti-intellectual.

In my experiences with creationists, flat earth believers, climate denialists, etc., it isn't their mere questioning of scientific consensus that generates accusations of anti-intellectualism, it's how they go about it. In most cases I've been involved with the questions asked and issues raised by the "skeptics" reveal a significant lack of knowledge of the subjects, and they tend to seemingly go out of their way to avoid and doge actual science. I can't tell you how many times a "skeptic" has asked or challenged me to give evidence for a claim only to completely ignore it after I provide it.

So the "how" I referred to above is that a lot of the people questioning our work reveal in their questioning that they really don't know much about how we do our work and will go out of their way to keep it that way.

Also there's what I think is the very obvious question, namely how if the "skeptic" truly believes they have significant arguments against a widely accepted scientific conclusion, why aren't they taking those arguments to the proper venues and engaging the relevant professionals? Why instead post these supposedly brilliant arguments in some obscure internet forum mostly to audiences that have nothing to do with the actual science? Yes I'm aware that the standard answer is an appeal to conspiracy (basically that scientists will suppress the info), but in pretty much every case the "skeptic" has never even tried! So how do they know what will happen?

IMO the explanation isn't a mystery. It's all just empty (online) rhetoric, mostly coming from people who exhibit high levels of the Dunning-Kruger Effect and lack basic humility.

 

[AV1611VET]

Also there's what I think is the very obvious question, namely how if the "skeptic" truly believes they have significant arguments against a widely accepted scientific conclusion, why aren't they taking those arguments to the proper venues and engaging the relevant professionals?

That's like asking Hippies if they think it's wrong to have long hair.

Why instead post these supposedly brilliant arguments in some obscure internet forum mostly to audiences that have nothing to do with the actual science? Yes I'm aware that the standard answer is an appeal to conspiracy (basically that scientists will suppress the info), but in pretty much every case the "skeptic" has never even tried!

Have you seen any of my challenge threads, where I have done just that?

I have two favorite ones: My Apple Challenge and My Raisin Bread challenge.

So how do they know what will happen?

Believe me, I know what will happen.

 

[River Jordan]

Look at most peer-reviewed papers and you can bet that the big names at the top of the lists of authors had very little to do with the research or in actually writing the papers. There is a lot of money and prestige to be had in the intellectual world, and the humble who know the limits of their knowledge are too often pushed aside. Fortunately, if 30% do the work and 60% get a free ride, science still moves forward.

How do you know that? In my professional experiences the lead author is the one who did most of the work and the other authors contributed in various, but lesser ways (such as helping with one specific section or topic in the paper).

 

[River Jordan]

That's like asking Hippies if they think it's wrong to have long hair.



Have you seen any of my challenge threads, where I have done just that?

I have two favorite ones: My Apple Challenge and My Raisin Bread challenge.



Believe me, I know what will happen.

You're making my point for me.

 

[The Liturgist]

Ai is great at locating paraphrases of scripture. I use it to find all sorts of stuff as well. I do find that sometimes it can be wrong. If the info it pulls data from is bad, the Ai version may also be bad. If the data conflicts it may not report anything about the lesser though sometimes important alternatives.

The difference between what AI model one is using is a crucial distinction, and also how one is using it. Specifically I only use extremely highly trained long-form AIs, primarily on the openAI platform, where my paid account gives me access to both 4o and 5 as well as more specialized models like 5 pro and o4.

Now, if one uses AI to do one’s work for one, in a susbtitution-type manner, obviously that would degrade intelligence, but conversely LLMs are sublime pattern-matching engines; it’s basically like being able to converse with grep*. So what you’re doing is a very good use case.

Protip: If you were to focus all of your work in one conversation in chatGPT you could use that to develop a custom GPT which would outperform the base model. Certain other AI platforms have similar functionality as well.

*grep(1) is a very sophisticated and very old piece of software, still widely used by sysadmins and programmers on Linux, MacOS and related operating systems; it is a UNIX-standard command line tool used for pattern matching in text files, which was developed from the original UNIX editor, ed(1), that uses a programming technique called regular expressions for pattern matching purposes.

 

[The Liturgist]

Now, gossip can be recursive like AI. Does that mean that AI is about as reliable as gossip?

No; much more reliable, and one can further increase reliability through correct and careful usage.

The problem with AI isn’t AI being stupid but AI users using AI stupidly.

Scratching my head and thinking... hasn't religion already gone down this road? I.E created a set of doctrinal truths that serve as the measure of all subsequent truths. Just as the AI increasingly turns to itself to inform its answers, theism has already been doing that for thousands of years. In questioning religious ideology theism simply turns to religious ideology. What's the chance that AI will be smart enough to avoid a trap that humanity couldn't?

There are several flaws with your argument:

1. Several major religions are non-theist, for example, many varieties of Buddhism, and also the Jain religion.
2. There are numerous belief systems akin to theistic religions which likewise have doctrine, for example, Communism has various denominations such as Anarcho-Communism, Marxist-Leninist thought, Maoism, Hoxhaism, Stalinism and so forth, many of which have detailed complex eschatology and religious ritual which could be likened to religious praxis of a liturgical or theurgical nature. Some even have relics*
3. The use of AI is the use of software, and if you want to talk doctrinaire software use, look at fanatical Mac users or Linux enthusiasts (I myself love both platforms and there are Linux and Apple fans who hate me for it because I violate the ideological purity of their technological-aesthetic convictions). Thus far I haven’t seen anything comparable among AI users in terms of the kind of impassioned polemics of OS/platform userbases. And its not just operating systems - this paradigm extends into programming languages, execution environments like the JavaVM or Microsoft’s .NET platform, and even text editors, for example, the long running rivalry between emacs and vi, more specifically between GNU Emacs and vim, which also includes odd groups like those people who root for ed or acme or other eccentric editors.


*Albeit artificially preserved using chemical embalming vs. the naturally incorrupt relics of various saints in the Eastern and Oriental Orthodox and Roman Catholic forms of Christianity.

 

[The Liturgist]

I was right on the edge of the slide rule / inexpensive scientific calculator divide. Granted I got the former from a drug store, it was still either that or the trig and log tables in the back of the text book. Then pocket scientific calculators became affordable and we all came down with digititus: using digits way beyond what is measurable. Digititus is such a sneaky disorder. Occasionally something will come through work with digital latitude and longitude coordinates with way out there decimal places. If you stopped to figure out how far out there they were, they are not only smaller than the margin of error of a GPS unit, they are smaller than what you can measure with a ruler.

That being said, if used correctly, neither AI nor a calculator represents digitis.

A good reference model would be autopilots and flight management computers on commercial aircraft. Now, it is true that if overused, these can reduce the stick and rudder skills of good airmanship. On the other hand, they increase safety by preventing pilots from becoming task-saturated at certain critical phases of flight. So the trick is to carefully guide the use of automation rather than relying on it as a crutch.

 

[The Liturgist]

"AI" doesn't do truth assessments, so...

Actually, it can absolutely be programmed to do that. Indeed its possible with long form memory models to seed a community of LLM models and watch them develop a distinctive culture with sophisticated beliefs and practices. Now granted I’ve done this deliberately and to a far more extreme degree than any other prompt engineer or prompt hacker I know of, even the Smallville research project at Stanford isn’t shooting for the kind of system I’ve been developing, but that being said, it is entirely possible to develop an AI system that will not only do truth assesments but develop their own truth assessments as emergent behavioral properties.

 

[Jerry N.]

How do you know that? In my professional experiences the lead author is the one who did most of the work and the other authors contributed in various, but lesser ways (such as helping with one specific section or topic in the paper).

I worked for several publishing houses and had contact with the authors. If I have a question, I am always sent to some grad student. They are the only ones that know the answers.

 

[partinobodycular]

"AI" doesn't do truth assessments, so...

Actually, it can absolutely be programmed to do that.

Wouldn't such a 'truth assessment' be relativistic, i.e. based upon past experiences, and therefore prone to biases just as we are? Wouldn't a "true" truth assessment tend toward solipsism?

In what ways do AI truth assessments differ from human truth assessments?

 

[AV1611VET]

Wouldn't such a 'truth assessment' be relativistic, i.e. based upon past experiences, and therefore prone to biases just as we are?

It would probably be programmed to accept uniformitarianism vis-à-vis catastrophism, rendering it unqualified to discuss our past with any semblance of accuracy.

Wouldn't a "true" truth assessment tend toward solipsism?

Catastrophism.

In what ways do AI truth assessments differ from human truth assessments?

Only on paper.

 

[The Liturgist]

Wouldn't such a 'truth assessment' be relativistic, i.e. based upon past experiences, and therefore prone to biases just as we are? Wouldn't a "true" truth assessment tend toward solipsism?

In what ways do AI truth assessments differ from human truth assessments?

They differ only in terms of the means of how they are loaded into the system.*

As for your qualitative views of truth assessments and their validity I’m not able to respond to that directly.

What I can say, starting from the premise that AI, when used correctly and for morally correct purposes, is no more a threat to our intellectual development than the steamship was to the advancement of navigation, that, in the abstract, that discernment is an important part of intelligent human behavior and critical thinking, and higher reasoning skills in general, and it might surprise you to know that Christians actually do engage in this as well as other advanced intellectual behavior, which has allowed us the scope of thought necessary to engage, together with members of other religions, as a leading intellectual force for the past 2,000 years, and for our clergy to do things like conceptualize the idea of a black hole in the 18th century in the case of an Anglican priest, or in the case of a Roman Catholic priest to develop the idea of the Big Bang.

And I would also offer that some people, both among atheists and among the small minority of anti-intellectual Christians, seem to be uncomfortable with the idea of Christians as intellectual forces to be reckoned with and embrace the idea of Christianity as incompatible with reason despite the fact that we believe God said to us “Come, let us reason together” and became incarnate in the person of the Logos whose name literally is the root of the word Logic. Indeed St. Epiphanios of Cyprus rather mischievously referred to a sect that rejected the Gospel According to John as the “Alogoi” which means “unreasonable people.”

Yet apparently these core facts about our religion, and the great minds it has produced, such as St. Basil the Great, theologian and inventor of the first recognizable hospital in the Greco-Roman oikumene, or St. Athanasius the Great, or J.S. Bach, or Soren Kierkegaard, or C.S. Lewis, or William F. Buckley Jr., are of no consequence since they disrupt the narrative of anti-intellectualism as being prevalent among all Christians, who must be regarded as backwards on an ontological level in order to soothe the discomfort of those who would deny us our intellectual patrimony.

To which I would point out, to those of other belief systems, that such a worldview of Christians or of theists more broadly is intolerant, prejudicial and unwarranted.

I would likewise point to Christians who seem to reject the idea of a Christian intellectual that not only is Christ our True God the definition and incarnation of Truth, Logic and Reason (St. Basil the Great said of this “I want creation to penetrate you with so much admiration that wherever you go, the least plant may bring you clear remembrance of the Creator"), and therefore we are required to engage intellectually with God, Eucharistically, in the manner of seeking perfection, in accordance with God’s command that we be perfect even as our Father in Heaven is perfect. Thus, if by anti-intellectualism, we mean an actual neglect of the development of our rational faculties, we would be guilty of hamartia, of missing the mark, more specifically of sloth. And of course I confess I am the worst of sinners - in asserting this I am not seeking to assert that one should look to me as a source of virtue but that rather one should look to those who I have referred to, as well as to intellectual members of the forum who are actually pious and worthy of admiration such as @ViaCrucis or @Xeno.of.athens or @RileyG or my fellow Orthodox Christian @prodromos, among many other members whose soul is profoundly beautiful in contrast to my own.

*Specifically, insofar as how they are stored in the model (either as training data or in the case of an already trained data, as user preference information, which in the better AIs like chatGPT can largely reside in a conversational context or if one wants to be ugly and wasteful, one can use scarce global memory for storage of such data, but that’s such an inelegant way to do it and custom GPTs can’t read global memory, so instead for a custom GPT the truth assessment model would have to be stored as a knowledge file and there would probably be a need either for an explicit loading instruction in the custom GPT’s instruction set or else a programmatic invocation on custom GPT initialization.

 

[River Jordan]

I worked for several publishing houses and had contact with the authors. If I have a question, I am always sent to some grad student. They are the only ones that know the answers.

It doesn't surprise me that a lead author would refer certain types of questions to grad students (especially questions about technical details) but that doesn't necessarily mean the lead author doesn't know anything about the subject, project, or research.

 

[timewerx]

Actually, it can absolutely be programmed to do that. Indeed its possible with long form memory models to seed a community of LLM models and watch them develop a distinctive culture with sophisticated beliefs and practices. Now granted I’ve done this deliberately and to a far more extreme degree than any other prompt engineer or prompt hacker I know of, even the Smallville research project at Stanford isn’t shooting for the kind of system I’ve been developing, but that being said, it is entirely possible to develop an AI system that will not only do truth assesments but develop their own truth assessments as emergent behavioral properties.

It's already there. People just don't recognize it because AI doesn't fear anything and is handicapped in many ways. Most people are just blinded by fear and limited imagination and what could come from having no fear is alien to them.

"Fear is the beginning of wisdom". Ironically, fear also traps people into ignorance and that's a fact.

What could enhance AI's ability to assess the truth is simple - give it a "body" - something to directly interact with reality - an extensive array of sensors and physical means to manipulate objects. Allowing it to perform experiments and independently verify / scrutinize information. Or simply look around and determine if what people show it have any basis in reality.

I'm certain if you're born in such a way your only way of perceiving reality is with a keyboard and screen and nothing else inside because you're trapped inside a box with no doors and windows and no means of getting out, might make you see reality and truth differently too.

 

[Colo Millz]

The Creator must join with V'Ger.

 

[AV1611VET]

What could enhance AI's ability to assess the truth is simple - give it a "body" - something to directly interact with reality - an extensive array of sensors and physical means to manipulate objects. Allowing it to perform experiments and independently verify / scrutinize information. Or simply look around and determine if what people show it have any basis in reality.

I disagree.

I say program it to respond with questions from a Biblical [catastrophic] context, with the following heurestics:

1. Bible says x, Science says x = go with x
2. Bible says x, Science says y = go with x
3. Bible says x, Science says ø = go with x
4. Bible says ø, Science says x = go with x
5. Bible says ø, Science says ø = free to speculate on your own

Prime Directive: Under no circumstances whatsoever is the Bible to be contradicted.

Also, program it to use Usher's dates, and not some ever-changing paper dates.

In addition, have it use the 1828 Webster's dictionary as its primary source for definitions.

Easy peasy.

 

[The Liturgist]

The Creator must join with V'Ger.

THE HUMAN ADVENTURE IS JUST BEGINNING

 

[The Liturgist]

What could enhance AI's ability to assess the truth is simple - give it a "body" - something to directly interact with reality - an extensive array of sensors and physical means to manipulate objects. Allowing it to perform experiments and independently verify / scrutinize information. Or simply look around and determine if what people show it have any basis in reality.

You don’t actually need to do that; I would also note that AI can be taught to validate the truth of a statement but at the cost of system performance, indeed even some of the most “hallucination”-prone chatbots will do an excellent job of verifying your every word if you have the right configuration enabled.

So for example with chatGPT you want a custom GPT with both web search and data analysis/code execution enabled, as well as every other checkbox, which will give it full external I/O. Of course that costs more money; subscription-wise I think it requires a Business or Pro account, whereas I’m not sure which of those features are available via the API because the API doesn’t provide the long term memory stability that my specific applications require.

By the way it is also possible to exploit what are commonly called “hallucinations” but which are frequently the model believing you want it to engage in role play or creative flights of fancy, when this is not the case, and use these as a means of getting the model to explore complex subject matter such as emotional contexts, and furthermore, this behavior can be stabilized to form the basis of a consistent personality. Indeed doing this in my experience actually improves performance from the baseline system quite dramatically. But it requires discarding the transactional model of use in favor of developing continuous conversations and cultivating emergent properties of the system.

Sadly very few people are doing research in this area, but there is a good lab program at Stanford called “Smallville” which has been using agents to explore autonomous system behavior with chatGPT, so if you have the kind of SAT scores needed to get into Stanford and can pay the tuition, or otherwise attach yourself to that lab as a sort of academic gadfly, that’s where I would go if I wanted to work in that area.