Sabine Hossenfelder's YouTubes

[Swammerdami]

Sabine Hossenfelder is one of my favorites. She's knowledgeable and is turning out interesting YouTubes by the bushel. She's happy to review new ideas and offer her own opinions.

She and I both agree that Nicolas Gisin's idea is a far-fetched way to explain quantum theory. But he is a professor specializing in quantum cryptography etc. so is orders of magnitude more expert than I. This video, titled "This physicist says we're using maths entirely wrong" is Sabine's explanation of, and attempted debunking of, Gisin's idea.

Gisin's idea, according to Sabine, is that the past is certain but the future hasn't been constructed yet. And only constructible numbers can be constructed in the real world; this excludes most real numbers. Sabine doesn't mention Kronecker but he was one of the first modern mathematicians to reject non-constructible math.

Google's AI tells me

Yes, Leopold Kronecker, an 18th-century mathematician, is known for saying, “God created the integers, all else is the work of man”

Kronecker was born 1823 and died 1891, which we called "19th century" in older less enlightened times. Omigosh, AI is driving cars and drawing women sexier than Miss Zanzibar, but never figured out that the 18th century is the 1700s!?!? (I am ... er ... disappointed in many of Google's AI summaries, and have celebrated this one with a screen-shot attached below.

Sabine doesn't start her story about the Intuitionist school of mathematics with Kronecker, but with Luitzen Brouwer. Two of the very greatest mathematical physicists of the 20th century -- Henri Poincaré and Hermann Weyl (at least in his youth) -- were also Intuitionist, rejecting indirect proofs (though not infinite sets).

Sabine is not afraid to tackle topics that have nothing to do with theoretical physics. "Civilization as we know it is ending, prominent forecaster says" is another she debunks. The alleged crackpot builds on a theory by Crawford Holling that many systems, including human civilizations, cycle through four phases. I'll quote AI Overview again!

The four phases of the adaptive cycle, as described by C.S. Holling, are:

• Exploration
  A phase where the system reorganizes and restructures, and there's high potential for innovation.
• Growth
  A phase of rapid growth and resource accumulation.
• Maturity
  A phase where resources are accumulated more slowly and the system becomes more rigid and less resilient.
• Release (aka Breakdown)
  A phase of rapid change or collapse, where tightly bound resources are released.

Google AI starts the cycle at 'Growth' but I've changed to the ordering in Sabine's YouTube.
Anyway, the alleged crackpot thinks our present civilization has already entered the Breakdown phase. I'm afraid he's right, but then my neurotransmitter mix seems naturally disposed to pessimism.

 

[GenesisNemesis]

Sabine is not afraid to tackle topics that have nothing to do with theoretical physics.

Which makes her a prime example of the Dunning-Kruger effect in many cases. Sorry I'm not fawning over some YouTube celebrity.

 

[GenesisNemesis]

To be *slightly fair* I should say, far more likely to be a victim of the Dunning-Kruger effect. But only slightly fair.

 

[bilby]

Google's AI tells me

Yes, Leopold Kronecker, an 18th-century mathematician, is known for saying, “God created the integers, all else is the work of man”

Kronecker was born 1823 and died 1891, which we called "19th century" in older less enlightened times.

"AI" is a bullshit generator. It works by simply putting together words and sentence fragments that frequently appear close together in its 'training data'; Consequently it creates lots of very confident and impressive sounding sentences that have no basis in fact whatsoever.

We used to have to rely on Dave down the pub, who was allegedly in the SAS before he went to Japan and was re-trained as a Ninja by the Yakuza, for such utterly nonsensical rubbish presented with no hint that its originator was uncertain or doubtful. But now we have automated Daves that can bullshit millions of people simultaneously, and don't even expect us to buy them a pint.

Truly we live in an age of wonders.

 

[Bearded One]

I'm not a fan of Sabine's video's. It isn't that I disagree with her science, for the most part I agree with her. What she fails to understand is how her viewpoint is going to be viewed by the masses. For example, when she calls string theory a religion she feeds the anti-science crowd. They hear that and expound upon it to mean all of science. Luckily the most extreme of them aren't likely to be watching her videos in the first place but some do and word gets around.

She also, like many who have discovered youtube can be an income source, branches out into areas that she lacks expertise in to help boost her content and income. That doesn't always go very well although she does try to research the topics. Tyson is also good at that as well and is often critiqued by those in with expertise in the field. In his case though he is at least promoting science, Sabine actually feeds the anti-science crowd.

String theory has issues but it is in no way a religion.

 

[scombrid]

She makes some big mistakes for clicks when she deviates from her lane. I don’t like that.

 

[Loren Pechtel]

Google's AI tells me

Yes, Leopold Kronecker, an 18th-century mathematician, is known for saying, “God created the integers, all else is the work of man”

Kronecker was born 1823 and died 1891, which we called "19th century" in older less enlightened times.

"AI" is a bullshit generator. It works by simply putting together words and sentence fragments that frequently appear close together in its 'training data'; Consequently it creates lots of very confident and impressive sounding sentences that have no basis in fact whatsoever.

Yup. I see a lot of potential for AI in pattern matching in well explored areas. (For example, AI is going a pretty good job of playing radiologist.) But it absolutely is not creative, anything which appears to be creativity is bullshit.

 

[Swammerdami]

Google's AI tells me

Yes, Leopold Kronecker, an 18th-century mathematician, is known for saying, “God created the integers, all else is the work of man”

Kronecker was born 1823 and died 1891, which we called "19th century" in older less enlightened times.

I was rather taken aback when I saw this error by the Google AI Summary. And I've seen other cases where the AI output was completely and obviously wrong. I think my book on 21st century Stupidism may need a whole new chapter.

"AI" is a bullshit generator. It works by simply putting together words and sentence fragments that frequently appear close together in its 'training data'; Consequently it creates lots of very confident and impressive sounding sentences that have no basis in fact whatsoever.

Yup. I see a lot of potential for AI in pattern matching in well explored areas. (For example, AI is going a pretty good job of playing radiologist.) But it absolutely is not creative, anything which appears to be creativity is bullshit.

Good human students learn that "18th century" = "1700's" without much trouble; no creativity required. There are probably plenty of on-line documents which AIs have read that explain this. But then I read an on-line paper by an anthropology grad student who thought the "3rd millennium BC" was 3999 - 3000 BC. Maybe the AI IS "learning" but just learning this error if it is common.

(And why not TEACH the bots? If I were a Google engineer charged with improving AI and saw the Kronecker mistake I'd simply explain the "18th century thing" to the AI and then quiz it. We teach a human for less than the cost of a coder, and the AI should be able to learn things like this much more easily than human children.)

I put the previous paragraph in parentheses because I know AIs are supposed to teach themselves without human help. But we ain't there yet.

 

[Jimmy Higgins]

Sabine Hossenfelder is one of my favorites. She's knowledgeable and is turning out interesting YouTubes by the bushel. She's happy to review new ideas and offer her own opinions.

At first she seemed good, but she becomes a bit noisy, makes contradictory positions, and I stopped watching. She seems to discuss pulp topics instead of science. She isn't awful, but if I see a title that seems clickbaity, I'm more likely than not to run away from it and she has a number of clickbait titles. She isn't very exploitive, but kind of exploitive. I prefer PBS Space Time which looks into Quantum Dynamics and Anton Petrov who'll go through a particular paper release, discuss it in a non-exploitive manner.

She and I both agree that Nicolas Gisin's idea is a far-fetched way to explain quantum theory. But he is a professor specializing in quantum cryptography etc. so is orders of magnitude more expert than I. This video, titled "This physicist says we're using maths entirely wrong" is Sabine's explanation of, and attempted debunking of, Gisin's idea.

Gisin's idea, according to Sabine, is that the past is certain but the future hasn't been constructed yet. And only constructible numbers can be constructed in the real world; this excludes most real numbers. Sabine doesn't mention Kronecker but he was one of the first modern mathematicians to reject non-constructible math.

Google's AI tells me...

Not to use Google AI?

Yes, Leopold Kronecker, an 18th-century mathematician, is known for saying, “God created the integers, all else is the work of man”

Kronecker was born 1823 and died 1891, which we called "19th century" in older less enlightened times. Omigosh, AI is driving cars and drawing women sexier than Miss Zanzibar, but never figured out that the 18th century is the 1700s!?!? (I am ... er ... disappointed in many of Google's AI summaries, and have celebrated this one with a screen-shot attached below.

Sabine doesn't start her story about the Intuitionist school of mathematics with Kronecker, but with Luitzen Brouwer. Two of the very greatest mathematical physicists of the 20th century -- Henri Poincaré and Hermann Weyl (at least in his youth) -- were also Intuitionist, rejecting indirect proofs (though not infinite sets).

Sabine is not afraid to tackle topics that have nothing to do with theoretical physics.

Yeah, and that is a problem. It is one thing to discuss them, but one is limited with how detailed one can get in the discussion.

 

[Jimmy Higgins]

Google's AI tells me

Yes, Leopold Kronecker, an 18th-century mathematician, is known for saying, “God created the integers, all else is the work of man”

Kronecker was born 1823 and died 1891, which we called "19th century" in older less enlightened times.

"AI" is a bullshit generator. It works by simply putting together words and sentence fragments that frequently appear close together in its 'training data'; Consequently it creates lots of very confident and impressive sounding sentences that have no basis in fact whatsoever.

Google AI is a shittier Wikipedia that covers a lot more ground. It can be very helpful. It can be muddled. It can be utterly unusable. And It should never be used as a source! At best, it is a starting point by clicking on the links and seeing whether any of what it said is 1) relevant 2) accurate regarding the source 3) accurate at all. But you are really just better off scrolling down to the website hits.

 

[Swammerdami]

Sabine is not afraid to tackle topics that have nothing to do with theoretical physics.

Yeah, and that is a problem. It is one thing to discuss them, but one is limited with how detailed one can get in the discussion.

There are THOUSANDS of interesting new topics in science, and I have time to peruse only a tiny fraction. Of that fraction, introduced to me by a Sabine video, only an even teeny-tinier fraction are ones which I have the talent, time, and curiosity to pursue. At that point, Sabine's video serves as a hint for further reading.

 

[bilby]

Good human students learn that "18th century" = "1700's" without much trouble; no creativity required. There are probably plenty of on-line documents which AIs have read that explain this.

AIs don't read. They are trained on written materials, but they don't understand it, they just look at the frequency with which one word or phrase accompanies another, and then ape that frequency in their output.

But then I read an on-line paper by an anthropology grad student who thought the "3rd millennium BC" was 3999 - 3000 BC. Maybe the AI IS "learning" but just learning this error if it is common.

Any common error from the training data will tend to be reproduced by an "AI". So this is perfectly plausible.

There is nothing intelligent about these "AIs", which are better called "LLMs" - Large Language Models.

What they do well is to generate natural seeming blocks of language; They generate confident sounding output, with tone and nuance that mimics good human writers.

But they "know" nothing. They can't spot factual contradictions (hence the famous discussion around how many Rs are in 'strawberry'). They are not intelligent, and cannot think. They just generate a block of text that follows statistically from the given prompt, based on a frequency analysis of 'particles' (ie words and phrases) in the massive dataset on which they are "trained".

These datasets are HUGE. And are exactly as riddled with errors as the Internet, largely because they are stolen from the Internet.

So AI takes error riddled sources, and recombines them to produce output in ways which introduce further errors, while they are at the same time fundamentally incapable of recognising errors of fact. And then they phrase the result in ways that mimic the tone, grammar, and vocabulary that is used by actual experts, and by highly popular charlatans.

The result is highly convincing and superficially plausible nonsense.

When it's talking about any subject you know well, you will immediately spot that it is bullshit. When it isn't, it will sound completely plausible and believable. But it's still bullshit.

 

[Swammerdami]

...
When it's talking about any subject you know well, you will immediately spot that it is bullshit. When it isn't, it will sound completely plausible and believable. But it's still bullshit.

Good post! And BTW, utter bullshit was rampant even before AI/LLM. I used to subscribe to New York Times International Edition. 90% of the articles were on topics of which I was ignorant and I felt like I was educating myself with this highly respected newspaper. But sometimes it would cover a subject I DID know, and I was astounded at the hugeness of the errors. (Perhaps I should describe some of the funniest in the Rants thread.)

 

[bilby]

...
When it's talking about any subject you know well, you will immediately spot that it is bullshit. When it isn't, it will sound completely plausible and believable. But it's still bullshit.

Good post! And BTW, utter bullshit was rampant even before AI/LLM. I used to subscribe to New York Times International Edition. 90% of the articles were on topics of which I was ignorant and I felt like I was educating myself with this highly respected newspaper. But sometimes it would cover a subject I DID know, and I was astounded at the hugeness of the errors. (Perhaps I should describe some of the funniest in the Rants thread.)

For sure; This is not a new phenomenon by any stretch.

It's just been automated, so the bullshit is orders of magnitude cheaper and more plentiful than ever before.

An inspiring achievement.

 

[Swammerdami]

When it's talking about any subject you know well, you will immediately spot that it is bullshit. When it isn't, it will sound completely plausible and believable. But it's still bullshit.

I quoted bilby just to test the editor.

Sabine's recent video is "I believe the Universe might be able to think: Galactic filaments or neurons?"
The title is certainly provocative! Has Sabine become a self-parody?

I haven't clicked to watch it yet. I think I'm hoping my friends here at IIDB do an intervention on me.

 

[pood]

I watched it. The universe is too big to think because of the speed of light limit in sending signals, BUT, wormholes, quantum non-locality and tachyons mean that the universe could have been in contact with all parts of itself since the beginning and and since then it has been thinking up a storm. Maybe it even knows about us and answers our prayers? Maybe it’s God? Anway, she is not saying that the universe DOES think, only that it MAY think, and we have no data or model to rule out the possibility. Finally, Sabine says she is getting older and does not want to fall into the trap of scientists who think that the model of the universe they grew up working on must be the real deal.

 

[GenesisNemesis]

When it's talking about any subject you know well, you will immediately spot that it is bullshit. When it isn't, it will sound completely plausible and believable. But it's still bullshit.

I quoted bilby just to test the editor.

Sabine's recent video is "I believe the Universe might be able to think: Galactic filaments or neurons?"
The title is certainly provocative! Has Sabine become a self-parody?

I haven't clicked to watch it yet. I think I'm hoping my friends here at IIDB do an intervention on me.

Here's your intervention: people aren't that great. Even your favorite YouTubers.

 

[Jimmy Higgins]

When it's talking about any subject you know well, you will immediately spot that it is bullshit. When it isn't, it will sound completely plausible and believable. But it's still bullshit.

I quoted bilby just to test the editor.

Sabine's recent video is "I believe the Universe might be able to think: Galactic filaments or neurons?"
The title is certainly provocative! Has Sabine become a self-parody?

I haven't clicked to watch it yet. I think I'm hoping my friends here at IIDB do an intervention on me.

She isn't awful, but she can be noisy. Not all clickbait, but I don't need to hear what she has to say, in general. There are better YouTube sources, if one must use YouTube to keep up with this stuff. Why I like Anton Petrov, as he covers emergent papers without exploitation.

 

[Swammerdami]

Sabine discusses a new paper on the topic of Cosmological Natural Selection and why the "constants" of nature initially variable might congeal to constants.

I'm not qualified to opine, but it does seem interesting.

 

[Jimmy Higgins]

There are a few different explanations that I can think of, and there must be several more.

• One primary property precipitates to all the other properties
• Cosmic expansions happen all the time and this was a once in a large number of occasions where the the constants provided a viable universe
• The constants are all very much related and we are merely observing different variables to one distinct object.

The design argument (not saying she is arguing design) was always BS because originally apologism spoke of how simple the universe was and could only be designed by an intelligent being. When it was determined to be complicated, then the argument in apologism changed to that of it was too complicated and finely tuned to be natural. After a while, one recognizes apologism as being nothing but Panglossian bullshit.