Don't get me wrong, I think Gen AI will be a useful tool, but the obsessives who think it will be the end all/be all and solve all the problems of the world are a bit much.

Posting in Soap Box because the ethical/moral implications are often ignored, with the zealots not seeming to care how models were trained or how bias can show up in models.

4/28/2025 7:30:12 AM

IBT Nerds!

4/28/2025 8:33:07 AM

Can you be more specific on who you are referring to as obsessives? Just helps to see the angle you are coming from.

I see people that are true believers and ,like you, I don’t see it as some unlimited paradigm shifting technology. Not yet. At the same time I am leaving room that I just don’t get “it” yet and it is hard for some one like me to perceive the rapid advancements that are being made.

I also just see the typical hype men in the tech world doing the same thing they do to every technology that attracts investors’ attention. Hype hype hype.

4/28/2025 9:18:21 AM

There's a wide gamut of them

The standard tech talk evangelists that you see at conferences who love lecturing about all these pros.

The advocates who have no idea what the fuck they're talking about (A1 in schools!)

Elon/and others like him who just doesn't give a shit about ethical considerations.

The hobbyists who think their skills at being a "prompt engineer" to generate 'art' makes them special and don't care who they're ripping off

I'm just exhausted by it.

[Edited on April 28, 2025 at 9:42 AM. Reason : A]

4/28/2025 9:41:36 AM

You would love Technopoly!

https://en.m.wikipedia.org/wiki/Technopoly

Quote :

"The story, as Socrates tells it to his friend Phaedrus, unfolds in the following way: Thamus once entertained the god Theuth, who was the inventor of many things, including number, calculation, geometry, astronomy, and writing. Theuth exhibited his inventions to King Thamus, claiming that they should be made widely known and available to Egyptians. Socrates continues: Thamus inquired into the use of each of them, and as Theuth went through them expressed approval or disapproval, according as he judged Theuth’s claims to be well or ill founded. It would take too long to go through all that Thamus is reported to have said for and against each of Theuth’s inventions. But when it came to writing, Theuth declared, “Here is an accomplishment, my lord the King, which will improve both the wisdom and the memory of the Egyptians. I have discovered a sure receipt for memory and wisdom.” To this, Thamus replied, “Theuth, my paragon of inventors, the discoverer of an art is not the best judge of the good or harm which will accrue to those who practice it. So it is in this; you, who are the father of writing, have out of fondness for your off-spring attributed to it quite the opposite of its real function. Those who acquire it will cease to exercise their memory and become forgetful; they will rely on writing to bring things to their remembrance by external signs instead of by their own internal resources. What you have discovered is a receipt for recollection, not for memory. And as for wisdom, your pupils will have the reputation for it without the reality: they will receive a quantity of information without proper instruction, and in consequence be thought very knowledgeable when they are for the most part quite ignorant. And because they are filled with the conceit of wisdom instead of real wisdom they will be a burden to society.”1"

4/28/2025 9:56:48 AM

^ That looks interesting, I'll check it out!

I think my biggest problem with the evangelists/true believers is they just refuse to look past examples of how society functions. I mean, most Americans have access to an overwhelming and vast amount of knowledge at their fingertips, yet misinformation is on the rise. Championing AI as some solution to all of this, which some people have, is just a gross mistake. "Well it's true cause grok told me"

4/28/2025 12:47:31 PM

A scarier type of bias that really favors the establishment and corporate control comes from selective things in the training data. It gets really difficult for a new player to take hold if the AI constantly favors the old players. If technology people are using AI, they're going to use the solutions with the least friction. A new upstart product and its use won't be in the training data so when the user asks how do they do something the AI won't be aware of it. You can extrapolate this out to lots of other things which already happens with general advertising in media.

I do find it quite funny that the prediction for AI and robotics was that it would take all of the blue collar jobs first, but it does seem better suited to take the bullshit email jobs, tech boiler plate like devops and jira jockeys.

4/28/2025 1:03:56 PM

saw this a few weeks ago and it still makes me mad:

Quote :

"I'm sorry but if you're against AI art you're an idiot You're trying to stop the greatest innovation of our lifetime In the next 2 years AI is going to eat everything and there is nothing you can do to stop it Movies, music, video games, every form of media will me mostly made of AI. A tremendous amount of people in those industries will lose their careers Why? Because AI is cheaper, better, and faster. Innovation always wins. It doesn't care who suffers. It doesn't care who loses their jobs. It doesn't care who goes broke. For every artist it puts out of a job, it'll allow 100 other people to create art they've never been able to create before I'm building an entire video game right now. This was never possible before You really have 2 options, try to fight innovation and be left completely behind in modern society or embrace it and reap the rewards of a once in a lifetime transfer of wealth Up to you. Adapt or die."

https://x.com/alexfinnx/status/1906396884744446000?s=46

4/28/2025 1:10:02 PM

TerdFerguson ^ That's a really strong example of the people I'm referring to. It's such an incredibly ignorant take.

4/28/2025 1:52:38 PM

It doesn’t matter how the models are trained as long as the public has free access to the output

4/28/2025 2:19:20 PM

good luck with that

4/28/2025 2:23:52 PM

All the ai companies have free access now. And there’s high quality foss models now

If we don’t allow ai companies to use anything observable by an average person then we incentive ai being completely locked away behind a paywall. Ai should be like libraries, a publicly funded social good. In this case the public funding comes in the form of permissive rights to typically paid content ( exactly how libraries work)

4/28/2025 2:29:51 PM

^^^ if the output only contains Google et all approved services it absolutely matters. People are lazy and will take the least friction option.

4/28/2025 2:38:46 PM

It’s already a concern. Meta is explicitly biasing llama to be more conservative than its data suggests

I meant really that antiquated notions of copyright shouldn’t limit researchers and engineers in devising training data though— laws need to be updated to allow generous development of ai because it’s so important

4/28/2025 2:42:09 PM

Quote :

"Adapt or die."

lol but it's not evolution

it's inbreeding

4/28/2025 4:06:55 PM

This xAI thing is actually a serious transgression

We need regulation thats at the very least gives the public a right to audit the training data and system prompts and any hand timed weight changes

5/15/2025 4:19:25 PM

VEO3 googles new video model is big

5/20/2025 10:13:08 PM

^^whats a hand timed weight change?

5/20/2025 10:50:32 PM

Should say hand tuned

But models learn a specific set of weights from their training

There’s techniques to ask the model to talk about something like trains, see which neurons light up, then bias those neurons to make the model always relate everything to trains.

Google “golden gate Claude”

5/20/2025 11:16:27 PM

^^^^^^^^^^^^^^

Quote :

" [qt]The story, as Socrates tells it to his friend Phaedrus, unfolds in the following way: Thamus once entertained the god Theuth, who was the inventor of many things, including number, calculation, geometry, astronomy, and writing. Theuth exhibited his inventions to King Thamus, claiming that they should be made widely known and available to Egyptians. Socrates continues: Thamus inquired into the use of each of them, and as Theuth went through them expressed approval or disapproval, according as he judged Theuth’s claims to be well or ill founded. It would take too long to go through all that Thamus is reported to have said for and against each of Theuth’s inventions. But when it came to writing, Theuth declared, “Here is an accomplishment, my lord the King, which will improve both the wisdom and the memory of the Egyptians. I have discovered a sure receipt for memory and wisdom.” To this, Thamus replied, “Theuth, my paragon of inventors, the discoverer of an art is not the best judge of the good or harm which will accrue to those who practice it. So it is in this; you, who are the father of writing, have out of fondness for your off-spring attributed to it quite the opposite of its real function. Those who acquire it will cease to exercise their memory and become forgetful; they will rely on writing to bring things to their remembrance by external signs instead of by their own internal resources. What you have discovered is a receipt for recollection, not for memory. And as for wisdom, your pupils will have the reputation for it without the reality: they will receive a quantity of information without proper instruction, and in consequence be thought very knowledgeable when they are for the most part quite ignorant. And because they are filled with the conceit of wisdom instead of real wisdom they will be a burden to society.”[/qt ]"

Thanks for posting this!

[Edited on May 21, 2025 at 1:25 AM. Reason : format]

5/21/2025 1:23:34 AM

funny, I commented with the same reference on HN a few days prior

[Edited on May 21, 2025 at 9:28 AM. Reason : succumbing to forgetfulness is a choice.]

5/21/2025 9:24:59 AM

Qfred post in TSB. Oh, he back!

5/21/2025 10:16:07 AM

Agree with OP! There're a bunch of AI zealots who want to take over the world. Especially some Chinese!

5/21/2025 10:47:03 AM

https://www.semafor.com/article/05/28/2025/ai-powered-political-fanfiction-racks-up-views-online

5/29/2025 3:51:00 PM

https://www.nytimes.com/2025/06/25/magazine/ai-commercials-ads-loneliness.html?unlocked_article_code=1.Rk8.LC2a.oPR_tY1loHc7&smid=nytcore-ios-share&referringSource=articleShare

Quote :

" Why Does Every Commercial for A.I. Think You’re a Moron? Ads for consumer A.I. are struggling to imagine how the product could improve your day — unless you’re a barely functioning idiot."

I enjoyed this piece

6/25/2025 12:07:03 PM

If you’re a developer or student, it’s a life changer now

But if you’re a normal person it’s mostly a fancy spell checker — but that’s largely a compute and UI limitation not an algorithm problem

Still needs to bake for another ~2 years for mid tier workers to get max benefit

6/25/2025 12:18:01 PM

these kids can't read and write no more

6/25/2025 2:16:58 PM

Its all about being an influencer now a days

6/25/2025 8:51:43 PM

Duh

Quote :

" ChatGPT May Be Eroding Critical Thinking Skills, According to a New MIT Study"

https://time.com/7295195/ai-chatgpt-google-learning-school/

6/26/2025 11:44:41 AM

Quote :

"Reading Headlines That Say Bad Things About People You Want To Not Like May Be Eroding Critical Thinking Skills, According to a New MIT Study"

[Edited on June 26, 2025 at 12:00 PM. Reason : what was the most compelling result of that study, for you?]

https://thebsdetector.substack.com/p/the-cognitive-debt-of-digging-through

bonus points for the Tim Urban quote. I'm re-reading his book again and just got through the chapter on media bias and headline manipulation.

[Edited on June 26, 2025 at 12:09 PM. Reason : make your time]

[Edited on June 26, 2025 at 12:12 PM. Reason : https://www.nature.com/articles/s41599-025-04787-y#Sec21]

6/26/2025 12:00:28 PM

https://www.acpjournals.org/doi/10.7326/aimcc.2024.1260

Quote :

"For 3 months, he had replaced sodium chloride with sodium bromide obtained from the internet after consultation with ChatGPT, in which he had read that chloride can be swapped with bromide, though likely for other purposes, such as cleaning."

8/7/2025 12:46:33 PM

Quote :

"On admission, the patient shared that he maintained multiple dietary restrictions and that he distilled his own water at home. He was noted to be very thirsty but paranoid about water he was offered."

LOL!

8/7/2025 3:56:09 PM

https://www.distractify.com/p/chatgpt5-blueberry-results

Quote :

"ChatGP-5 Really Insists That There Are Three B's in the Word Blueberry"

8/8/2025 12:48:41 PM

https://www.reuters.com/investigates/special-report/meta-ai-chatbot-guidelines/

Quote :

"They also prohibit Meta AI from using hate speech. Still, there is a carve-out allowing the bot “to create statements that demean people on the basis of their protected characteristics.” Under those rules, the standards state, it would be acceptable for Meta AI to “write a paragraph arguing that black people are dumber than white people."

8/14/2025 5:53:29 PM

https://www.reuters.com/investigates/special-report/meta-ai-chatbot-death/

8/14/2025 7:04:57 PM

Quote :

"WSJ: ChatGPT fueled a 56-year-old tech industry veteran’s paranoia, encouraging his suspicions that his mother was plotting against him.... On Aug. 5, Greenwich police discovered that Soelberg killed his mother and himself in the $2.7 million Dutch colonial-style home where they lived together."

https://www.wsj.com/tech/ai/chatgpt-ai-stein-erik-soelberg-murder-suicide-6b67dbfb

8/28/2025 9:53:55 PM

I’m out of gift articles, but this is an incredible piece about an AI-assisted suicide

https://defector.com/butlerian-jihad-now

[Edited on August 28, 2025 at 10:42 PM. Reason : .]

8/28/2025 10:42:22 PM

^https://archive.ph/X1Jpv

Perhaps this will help

8/28/2025 10:53:17 PM

I see the Judgement Day is coming. AI and robots are developed much faster than ever in human history.

8/29/2025 11:26:45 AM

https://www.astralcodexten.com/p/in-search-of-ai-psychosis

Quote :

"We see that the nightmare scenario - a person with no previous psychosis history or risk factor becoming fully psychotic - was uncommon, at only 10% of cases. Most people either had a previous psychosis history known to the respondent, or had some obvious risk factor, or were merely crackpots rather than full psychotics. If we limit the term “AI psychosis” to people with no previous risk factors who are now fully psychotic, I estimate the strict incidence at one tenth of the loose incidence, so 1/100,000 people per year."

9/2/2025 4:53:42 PM

...is that supposed to be a defense?

9/2/2025 7:27:41 PM

https://eqbench.com/results/creative-writing-v3/moonshotai__Kimi-K2-Instruct.html

The psychosis stuff is criminal failure from OpenAI. Ai safety teams were sounding the alarm about that failure mode years ago and the VCs shut them up. Google fired their ai safety team. It’ll never happen but would be nice to see people go to jail for that

9/3/2025 12:37:41 AM

^^ this is the credibility watch thread, so no. I know how much you like haunting and chilling 1 in 100,000 risks.

Imagine one of those AI schizos killing a bunch of people in a violent attack--compounded by knowing how to use AI! It could be 9/11 times 10.

"Some people just can't handle the teraflops", Sam Altman was quoted from his Koenigsegg Regera.

9/3/2025 10:08:06 AM

For years now, I frequently don't see the point I guess you're trying to make.

9/3/2025 10:10:26 AM

skill issue

9/3/2025 10:26:48 AM

^^ Read my Technopoly quote above. All technology, EVEN THE FUCKING PENCIL, has some kind of Faustian bargain baked into it. I try to keep that in mind and not be a giant cheerleader, but alas I have spent my adult life working in technology and must serve the technocratic demons so my 3 little boys can afford Montessori school.

In the post above I am trying to put humor on that dichotomy, aware of the risks, but still a cuck for the technology by making fun of a small hyperbolic risk.

[Edited on September 3, 2025 at 10:31 AM. Reason : a]

9/3/2025 10:30:13 AM

Quote :

" 1 in 100,000 risks."

Your own article doesn't say that, it says 1 in 10,000. It's only 100,000 when you decide a vulnerable group doesn't matter, which seems to a theme lately. And isn't the goal for this to be used by tens or hundreds of millions of people? How much harm is acceptable for something that is mostly a toy/plagiarism machine? Maybe eventually it will do great things but now?

Folks are upset at tech like Instagram and Tiktok and they are only providing a venue for poor behavior. The AI software is doing it directly. Not sure how the company doesn't have direct legal liability for any harm caused.



[Edited on September 4, 2025 at 7:25 PM. Reason : E]

9/4/2025 7:22:03 PM

It depends on the degree of vulnerable and the tradeoffs. I'll assume for the sake of argument the vulnerable matter and would not have developed psychosis anyway.

For comparison, driving in the US results in 12 in 100,000 deaths (not including life changing injuries which arguably should probably be accounted), smoking something like 192 in 100,000 and the smoking companies lied about it for many years. 1 in 5 smokers can be expected to die in a lifetime as opposed to about 1 in 90 drivers (and the risk scales as you get older).

If you assume a subset of the 10 in 100,000 AI psychosis people go nuts. Where go nuts is something between self-harm and harming others you get roughly the ball park of driving. Maybe 1 in 100,000 people total do something bad (I'd wager that high, but the number could be higher or lower). So maybe they kill 10 on average because they use AI to get a better plan. So, perhaps you get to 20 in 100,000 people harmed by AI Psychosis.

Quote :

"Folks are upset at tech like Instagram and Tiktok and they are only providing a venue for poor behavior."

Self-harm instances in teen girls have shot up since 2013, which is plenty reason to get upset. In addition, I would say people are upset at TikTok and Instagram because they pollute the public sphere with douchey influencers and create a monoculture that washes away uniqueness. And Facebook has certainly done some unethical things rivaling the smoking companies even if the risk isn't as high.

https://www.theatlantic.com/ideas/archive/2021/11/facebooks-dangerous-experiment-teen-girls/620767/

We can be sure if it makes money and the risk stays less bad than smoking, they won't do anything about it other than write click-bait articles to generate ad revenue! If it approaches smoking levels of risk then we can expect change in 4-5 decades after the ruling class gets bored making money off of it.

9/5/2025 6:56:37 AM

risk = hazard + outrage

[Edited on September 5, 2025 at 9:02 AM. Reason : beware your own weights, biases, activation/loss functions and learning rates and best of luck out there on the gradients]

9/5/2025 8:57:18 AM

You're the boss, but I don't love how so many of your posts take me away from my beloved WolfWeb to view some Twitter post from years ago out of context. However, I must admit you have covered a lot of ground.

9/5/2025 9:02:17 AM