Yang May 8
monkϵyborg 🦾🐵

All those people saying “I was skeptical of AI, until I tried Claude and I was amazed that it could tell me this and this and this” sound just like

“I was skeptical of psychics, but then Madame Fortuna told me something only my dead wife would know!”

Human perception has certain well-documented flaws that confidence men prey upon, and you are not immune from these

Apr 29 at 8:16pmWeb
Show threadTheosoreassApr 29
@monkeyborg what did madame fortuna say?did she mention the sprouts of wrath?
Show threadmonkϵyborg 🦾🐵Apr 30
@Theosoreass Oh yes, totally
Show threadSashinApr 29
@monkeyborg Where have these flaws been documented? I want to read about them!
Show threadChookMother 🇦🇺🦘Apr 30
@sashin @monkeyborg There's an article here: https://biolecta.com/articles/exploring-psychic-tricks-analysis/
If you then search some of the terms used, such as "cognitive bias" you'll find more information.
Show threadmonkϵyborg 🦾🐵Apr 30

@anne_twain @sashin Are you trying to be ironic by linking an AI-generated article?

EDIT: Folks, please no more replies to this. It was a simple mistake. Piling on after the mistake has been called is not a good look. You may have caught this one, but you’re not too smart to fall for AI, and if you think you are, the original post was about you.

Show threadChookMother 🇦🇺🦘May 1
@monkeyborg @sashin No I'm trying to be helpful to Sashin who wants to learn more about the subject. The article has a named author and author bio.
Show threadmonkϵyborg 🦾🐵May 1

@anne_twain @sashin I apologize, I intended no disrespect.

I don't believe that's a human-written article. My day job involves editing articles for a university magazine. The writing style here immediately felt off to me, much like an AI-generated image feels off. There is too much imitation of writing formalisms, too little specific information.

I could be wrong. But I can't find a Linda Tran at Berkeley. Even if there were, why would a physicist be writing an article on human psychology?

Show threadChookMother 🇦🇺🦘May 2
@monkeyborg @sashin Thank you for your apology, though I don't blame you at all. I look at it this way - we were living in the Information Age, now we're moving into what I call the Age of Bewilderment, when we are constantly uncertain about the credibility of what we see, hear and read. Even if one person is certain, they'll meet with others who are not certain. Stopping to debate every possibility will just be exhausting, don't you think?
Show threadschratzeMay 5
@anne_twain @monkeyborg @sashin but that is definitely 100% an AI-generated webpage. There's not much to debate here.
Show threadChookMother 🇦🇺🦘May 5
@schratze @monkeyborg @sashin Like I said, exhausting.
Show thread🍉 max frühschütz – нет войнеMay 6
@anne_twain @schratze @monkeyborg @sashin also the logo of the web page and all the images are generated. the whole page and all its contents might be, actually
Show threadChookMother 🇦🇺🦘May 6
@mfru Would you untag me from this conversation please? I'm not interested.
Show thread🍉 max frühschütz – нет войнеMay 6
@anne_twain it is kind of relevant to point out LLM generated resources in a thread that informs about LLMs, don't you think? even more so if you are the one presenting information that is, with a high degree of confidence, generated.
Show threadmirthMay 5
@monkeyborg @anne_twain @sashin Agreed, additionally it would be unusual to use the "Prof." title as opposed to "Dr.". The bio reads like a field from a database of prompts.
Show threadvarx/socialMay 6

@monkeyborg @anne_twain @sashin Also, you can look at the other articles that the "author" wrote: https://biolecta.com/authors/prof.-linda-tran/

- "Diverse Adaptations of Land and Sea Animals"
- "Mathematics in Programming: Essential Insights"
- "Understanding Dark Matter: A Deep Dive"

Unlikely that this is all one, real person.

Show threadChookMother 🇦🇺🦘May 6
@varx @monkeyborg @sashin Would you untag me from this thread please? I'm not interested.
Show threadmathewMay 6

@monkeyborg @anne_twain @sashin I have a slop detection app, and it rates the article as slop with 90% confidence. It rates the article's image as slop with 78% confidence.

There are three LLM-isms just in the first paragraph:
1. "In examining A it is crucial to B"
2. "they are not just A but B"
3. "draw on X, Y and Z" (rule of 3)

Show threadChookMother 🇦🇺🦘May 6
@mathew Would you untag me from this conversation please. I'm not interested.
Show threadchrottkatze "off-by-one"May 6

@mathew dont fucking use slop detectors.

while that article may be slop and i despise slop as much as the next person, these slop detectors also cannot be accurate.

Show threadmathewMay 6
@schrottkatze That’s why it provides a percentage, and why people test implementations against a sample text corpus. It’s also why I performed a human analysis — which also cannot be accurate.
I refuse to give in and say that we can’t detect slop completely accurately so we should give up trying. That’s exactly what the LLM proponents want.
Show threadChookMother 🇦🇺🦘Apr 30

@monkeyborg Excellent observation.
Others who exploit these human vulnerabilities are politicians and advertisers. Only the most vigilant avoid getting conned. Now we must summon even more vigilance. It's exhausting.

#AI #ASI #AGI

Show threadMatt GoddenApr 30
@monkeyborg "I was skeptical of the quality of the products of child slave labour, but let me tell you, these new Nikes..."
Show threadjohanApr 30
@monkeyborg that’s true in the sense that statistic probability tells you Madame Fortuna is sometimes right. The problem is when you start making life choices based on those predictions
Show threadTheolodianApr 30
@monkeyborg this is a pretty good summary of my concerns. LLMs are patronising and sycophantic. With a bit of interaction I can get sophisticated results that I really like. But there’s this nagging feeling that it’s just figuring out what I want to hear…
Show threadMastodon MigrationMay 1

@monkeyborg

I was a mediocre coder until I started using Claude, and my productivity went through the roof.

Show threadNeo EhproqueMay 5
@mastodonmigration @monkeyborg so, why should I not fire you and hire a 16 year old
kid to use Claude instead?
Show threadDaveMay 5

@mastodonmigration @monkeyborg

How did you measure that productivity?

(I assume your post is sarcastic)

Show threadSamuelJohnsonMay 1

@monkeyborg only to people incapable of any serious discrimination, like Joseph Weizenbaum's secretary who asked him to leave the room so she could converse in private with Eliza, the first chatbot, which she had seen him working on.

Claude routinely tells me things I don't already know and which are immediately verifiable and it routinely tells me things are wrong, out of date, inaccurate.

Only a credulous fool would trust a probabilistic algorithm as an oracle. Who knew?

Show threadmonkϵyborg 🦾🐵May 1
@samueljohnson While I agree with the kernel of what you're saying, I don't see any need to insult people who may have gotten taken in — or to imagine that people who possess better-than-average critical thinking skills are immune.
Show threadSamuelJohnsonMay 1
@monkeyborg I am merely very weary of the endless sweeping generalizations about AI we're increasingly subjected to. No insult intended, presumably any more than in your unfortunate invocation of spirit medium based fraud.
Show threadjabsonikMay 5
@monkeyborg "I was sceptical of drugs, but then I tried heroin.”
Show threadGroßlesbe Ranja ١٣١٢May 5
@monkeyborg
Except that I'd still trust a psychic a million times more than an llm.
Show threadThe Sleight Doctor 🃏🍉May 5

@monkeyborg

Just in case you or anyone reading this missed @baldur 's excellent 2023 article on this very topic, here it is. 👇🏼

https://softwarecrisis.dev/letters/llmentalist/

It's well worth a read!

The LLMentalist Effect: how chat-based Large Language Models rep…

The new era of tech seems to be built on superstitious behaviour

Out of the Software Crisis
Show threadghosttieMay 5
@monkeyborg I was skeptical of AI, until I tried Claude and it told me there were two r's in strawberry
Show threadred_rooster May 10

@monkeyborg

Likewise with me #ai #lechat by #mistal also an #europe made AI.

For catching point for all #catlovers
Le chat doed not mean the obvious, (actually: le bavardage) but the the tomcat.