Post by @thomasfuchs@hachyderm.io

LLMs are really interesting as a technology because of how polarizing people think about them.

There’s three main very vocal groups: “AI boosters”, “AI doomers”, and what I would call “never AI” people.

Now obviously the doomers and boosters are very much worse than that third group, however all of these groups show at least some cult-like behavior and fallacious arguments.

I’d urge you, if you tend to be critical of “AI” (I’m too!), to not succumb to sanctimonious fanaticism about it like the doomers and boosters are.

0 Boosts

0 Likes

Anthony

@abucci@buc.ci 4 hr. ago

@thomasfuchs@hachyderm.io There is more than one rational, reasoned stance one could take that leads to a "never AI" position.

Cults have charismatic leaders whose dictates are followed more or less unquestioningly. Who's the "never AI" leader?

Sorry but this part of your post doesn't pass a smell test.

Thomas Fuchs

@thomasfuchs@hachyderm.io 3 hr. ago

I said “cult-like behavior”; also cults do not require a leader.

alihan_banan

@alihan_banan@mastodon.world 4 hr. ago

@thomasfuchs I somehow find myself sharing a position with the Pope Leo. Although I'm not sure whether he is critical or never ai type of person, he still advocates for humanity

Stuart :progress_pride:

@slowe@mastodon.me.uk 6 hr. ago

@thomasfuchs There are more than three. Your thread here demonstrates another very common one: telling off those who say they don't want "AI" (the marketing term that is generally used to conflate LLMs with all other methods) by insisting that, actually, LLMs are interesting with positive (sometimes imagined or hoped for) use cases and that the "no AI" folks are bad people for not being supportive of those. Every position, including that one, can be presented as "cult-like".

Stuart :progress_pride:

@slowe@mastodon.me.uk 6 hr. ago

@thomasfuchs It isn't that surprising that there is a commenter in your thread telling people off for being against "AI" who also talks in terms of "thinking sand". He says he is against the bad uses and bad people (who control almost the entire endeavour) but that open/local/good will win out in the end. With no solid plan for how that massive change in power will happen other than a call to inevitability (of market forces?).

Stuart :progress_pride:

@slowe@mastodon.me.uk 6 hr. ago

@thomasfuchs There's definitely an element of presenting "there are good uses of 'AI'" as a "reasonable" or "rational" position that then gets sort-of ignored as a position itself. It makes itself a "default". I don't really see it act against power.

Thomas Fuchs

@thomasfuchs@hachyderm.io 3 hr. ago

It’s IMO mainly boosters who say this with the “it’s inevitable” claim or with hypotheticals like “it will cure cancer” that reek to me of religious-like fervor. (Both are just different types of fallacies.)

I think it’s completely evitable (of course we have a choice) and it won’t ever cure cancer.

My gist from studying the technology and the people hawking it is that while it’s a tool that can speed up certain things (e.g. making software), it’s really not applicable or worse, actively detrimental, to almost anything it’s sold for.

It’s also possible to use it responsibly without creating harm (there are truly open models [only licensed data] that you can run locally, with very little power consumption). Of course few people do this.

Rich Puchalsky ⩜⃝

@richpuchalsky@mastodon.social 11 hr. ago

@thomasfuchs

Putting "never AI" people in with the other two is a standard liberal move: it goes: there are various political cult-like people, fascists, and anti-fascists.

Not everyone who has come to a strong conclusion is a sanctimonious fanatic. Some are more correct than the people carefully preserving their nuanced middle position.

David B. :SetouchiExplorer:

@David@setouchi.social 11 hr. ago

@thomasfuchs The planet is burning. Data centers are accelerating the process. People controlling LLMs are fine with fascism, they're fine with billions of people dying (which is the outcome of the planet burning, not their AI fantasies).

I don't think it's being a fanatic to reject the thing altogether and be vehement against it.

Is it a being a fanatic to be 100% against nuclear weapons? Is it being a fanatic to be against slavery?

Sean

@seanlinsley@mastodon.social 11 hr. ago

@David @thomasfuchs practically every action that people take in their daily lives is accelerating climate change. You can't isolate AI from capitalism as uniquely evil; similar emotional logic led governments to abandon nuclear power and in much more real terms cause this inevitable climate crisis decades ago. Are you going to maintain this position when AI is used to cure incurable diseases?

David B. :SetouchiExplorer:

@David@setouchi.social 11 hr. ago

@seanlinsley @thomasfuchs 1. LLMs will never cure any disease.

2. You do understand there is a different between being stuck in a practice that has been imposed by a system decades ago (for example, driving) and a new practice that has no added value to our lives and that is not indissociable from society yet, quite the contrary.
For most people it's impossible to stop driving. For all of us, it's easy to not start or to stop using LLMs.
⤵️

Sean

@seanlinsley@mastodon.social 11 hr. ago

@David modern machine learning has been used in drug research for years now. Medicine is an absolutely massive computational space; larger than most any other technical field. The exact possibilities of how atoms interact with the body is incalculable. Humans will never solve the truly difficult problems on their own. Computers, with both deterministic algorithms and generative algorithms are non-negotiable

1+ more replies

David B. :SetouchiExplorer:

@David@setouchi.social 11 hr. ago

@seanlinsley @thomasfuchs Data centers don't just emit large amounts of CO2, they also deprive communities access to water. And climate is just one issue. How about the concentration of human knowledge into the hands of the oligarchy, the destruction of the web, the monopolization of RAM, which will make computers and phones unaffordable to most people (watch the next step being OpenAI selling ChatGPT terminals to lock people in even more) and the list goes on. I'm sure you know it.

1+ more replies

Thomas Fuchs

@thomasfuchs@hachyderm.io 11 hr. ago

One of the best ways to avoid descending into fallacious proselytizing is to thoroughly understand the theory of the technology (this requires some pretty deep understanding of computer science concepts and maths) and the applied uses of it (needs a grasp of philosophy and psychology, not so much because of sand generating words but to understand the humans behind it and those who are using it)

Thomas Fuchs

@thomasfuchs@hachyderm.io 11 hr. ago

Personally I have read A LOT about all of the above, I’ve always found “AI” fascinating (way before it was in the current LLM hype phase). And psychology and philosophy is IMO required to make good* software.

*both in a quality and moral sense

Thomas Fuchs

@thomasfuchs@hachyderm.io 11 hr. ago

The best single book about all of this remains “Computer power and human reason: From judgment to calculation.” by Joseph Weizenbaum.

It’s 50 years old this year, and obviously doesn’t cover LLMs specifically, but the concepts apply.

(You can easily find a PDF of it online.)

Albert Cardona

@albertcardona@mathstodon.xyz 15 min. ago

@thomasfuchs

Weizenbaum's opening is quite something:

«Science promised man power. But, as so often happens when people are seduced by promises of power... the price actually paid is servitude and impotence.»
—Joseph Weizenbaum

#ComputerScience

1 more reply

Ivo Limmen

@ivolimmen@mastodon.nl 4 hr. ago

@thomasfuchs https://archive.org/details/computerpowerhum0000weiz_v0i3

View

1+ more replies