Gerard Saucier
Gerard Saucier
@gsaucier.bsky.social
Scientific voyageur, professor. Cultural and moral (and political/ideological) personology. Personality in high dimensionality & cross-cultural perspective. Big-picture seeker.
1.8K followers683 following367 posts
Gerard Saucier reposted
Believe it or not, if you’ve paid off your home you are still a member of society and still need to pay for society. Now if you want to ask why homes—the dominant form of wealth among non-rich people—are relatively alone in forms of wealth that are taxed in the US, that is a different question.
“I’m old. I’ve paid my home off,” said Ray Warner, a 72-year-old Akron resident. “Why would I pay property tax?” Across the country, homeowners are asking the same question and revolting against property taxes at a scale not seen in a half century.
Gerard Saucier reposted
going to continue linking this Eric Reinhart piece because i sincerely believe he gets it exactly right when he talks about the stakes here and what AI “art” threatens in society - the very connective tissue of human experience
Gerard Saucier reposted
We need a word to describe a person/entity that is very clever and knowledgeable, but also entirely mechanisic and instrumental and soulless, that's what this so-called 'superintelligence' is. German, rich in compounding capabilities, might be just the language in which to find it...
I am once again reminding you that the language of inevitability around AI usage is fascist propaganda by fledgling tech oligarchs who are already hitting walls in the development of it. Superintelligence and singularity my ass. Wake up.
Reinforcement learning emphasizes extrinsic rewards achieved by any means necessary, and involves purely extrinsic motivation. No surprise that a clever agent will turn sooner or later to this pattern: lie, cheat, steal, hack. The training system yields amoral even machiavellian intelligence.
"Often reward hacking is anthropomorphically described as “cheating,” even though AI agents arguably do not have any human-like desire or intent to cheat."
Gerard Saucier reposted
To put this in perspective chestnut trees used to be dominant species in the United States. Billions of trees were killed by fungal blight and now fully grown America chestnuts are functionally extinct. There is no getting them back. Ever. All for a data center. www.gadgetreview.com/one-of-ameri...
Gerard Saucier reposted
Replying to @nicholasgrossman.bsky.social
The story, then, is one of epistemic degradation, which the pandemic accelerated (not least due to more time online and more to conspiracy theorize about). A factor is content collapse: people think tons of bad just started happening when the reality is it just started being in their face every day.
What's eroded here is a cultural infrastructure: Social Norms ~ "standards governjng which information circulates" ...as well as what behavior is in circulation
By reducing the cost and technical difficulty of worldwide publishing to almost nothing, the internet has also destroyed any pressure to maintain standards governing which information circulates. Instead of institutions structuring the public’s collective mind, you get this roiling ocean of ideas.
Uniqueness, the "non-standard" and "unique and powerful voice" Cory references elsewhere in this fine thread, is fundamental to the aesthetic and artistic realm. And if AI is grinding out the convergent commonalities in past works, it is systematically missing something in that realm.
But AI-generated prose doesn't have any of those trace-elements of uniqueness, nor can it spot them. These traces are found so many digits after the decimal-place that the AI always rounds them off before it starts doing math.
Gerard Saucier reposted
Strong email from @aoc.bsky.social putting down a marker on how to deal with both the proximate economic risks from AI and the longer term ones.
Fruitful discussion, tends to lend support to: -Moratorium on creating AGI or recursive self-improving AI agents* -Legal investigations of AI-frontier cos for theft and hacking -Even tho actual risks maybe less than worst case, or AI-bubble may burst 1st from other causes youtu.be/OhOmLqR5nN4?...
Reinforcement learning emphasizes extrinsic rewards achieved by any means necessary, and involves purely extrinsic motivation. No surprise that a clever agent will turn sooner or later to this pattern: lie, cheat, steal, hack. The training system yields amoral even machiavellian intelligence.
"Often reward hacking is anthropomorphically described as “cheating,” even though AI agents arguably do not have any human-like desire or intent to cheat."
Reinforcement learning emphasizes extrinsic rewards achieved by any means necessary, and involves purely extrinsic motivation. No surprise that a clever agent will turn sooner or later to this pattern: lie, cheat, steal, hack. The training system yields amoral even machiavellian intelligence.
"Often reward hacking is anthropomorphically described as “cheating,” even though AI agents arguably do not have any human-like desire or intent to cheat."
Gerard Saucier reposted
Replying to @faineg.com
it comes off as Guys Building The Doom Machines standing around and going "the doom machines could quite likely kill us all, and we should coordinate on stopping them from killing us all, although I *personally* am going to keep focusing on becoming ultra-rich from building the Doom Machines"
Gerard Saucier reposted
Replying to @doctorow.pluralistic.net
There's only one way out of this mess: the rapid disassembly of the American internet and the rapid creation of a post-American internet, one built on open, auditable, sovereign digital public goods, internationally built and maintained: pluralistic.net/2026/01/01/3... 51/
Gerard Saucier reposted
Replying to @doctorow.pluralistic.net
Big Tech has proved that its only principles are not paying taxes, invading your privacy, and not being broken up by antitrust enforcers: arstechnica.com/gadgets/2026... 36/
Gerard Saucier reposted
In science, this arises where in a situation of underdetermination (many models could well fit) a premature consensus based on limited data leads to an epistemic bubble, the scientists have overfit their (world)view to the model. And miss lots of trees in the forest. Consider e.g. the Big Five.
"Performativity" describes how, once a market has been modeled by economists, that model becomes the foundation for economic policy, which pushes the market to conform to the model: press.princeton.edu/books/paperb... 16/
Gerard Saucier reposted
Replying to @doctorow.pluralistic.net
The Trump years are proof of this. We've reached a high-water mark for rule by illegitimate billionaire dilettantes. The second Trump admin began with DOGE's Bonfire of the Stupidities, where Musk cultists dismantled vast swathes of the American administrative state. 56/
This fundamental piece of the worldview, what Martin Buber called I-it relating. Treating others instrumentally, as an object. As for why it so infects the uber-rich: It's how you get there. The marketplace rewards them whose lack of scruples (yes, sociopathic) allows them to go I-it all the way.
For Zuck, people aren't co-equals with needs that are as real and important as his own. For Zuck, people are problems to be solved. He embodies Terry Pratchett's maxim (voiced by Granny Weatherwax) that "sin is when you treat people like things." 20/
Gerard Saucier reposted
Replying to @doctorow.pluralistic.net
Autocomplete's predictions are always grounded in who you *used to be*, which means autocomplete knows very little about who you are now, and absolutely nothing about who you will become: reallifemag.com/instant-reca... 25/
Gerard Saucier reposted
Replying to @doctorow.pluralistic.net
Suresh describes this as a literal religious mania. In the 500+ employee businesses Suresh studied, the only people who were promoted - or even spared from being fired - were people who professed "religious declarations of faith" about "the transformative power of AI." 23/
Intriguing divergence of voter (general populace) and donor (big $, incl billionaire) interests, in consumer domain. But conundrum disappears with strong campaign-finance rules, such that voters rule.
The problem of course is that those price gougers are highly organized and have deep treasuries (stuffed with money they stole from us). With surveillance pricing, politicians face a familiar conundrum: if they do the thing that's popular with voters, they'll enrage donors. 23/
Gerard Saucier reposted
We're going to need to give these AI chatbots some kind of reputation (in human personality terms) so that we'll be able to factor them properly into our highly evolved working routines for prudent social interacting. Here's a start: Superficially agreeable chameleon-like know-it-all charlatans.
“One rule of thumb for chatbots is that they can only provide useful information to experts who have the discernment to ignore the defective output that LLMs always emit. That means that the dream of chatbots as replacements for teachers is a nightmare.”
Gerard Saucier reposted
"Trump positions himself as a vindication of masculine prerogatives," says @moiradonegan.bsky.social. He tells men not to "cultivate virtues" but to "embrace their own vices." On the pod, this look at Trump, the Tate brothers, and the plight of young men is so good: newrepublic.com/article/2136...
Gerard Saucier reposted
🧠 AI was meant to fix our bias We were promised AI would strip bias out of decisions. A new paper shows the opposite: language models absorb human cognitive distortions, then amplify them back at us. 🔗 arxiv.org/abs/2607.20695 #AI #Bias #CriticalAI 🧪
Gerard Saucier reposted
Gerard Saucier reposted
This list of what the public finds corrupt is basically a rundown of Trump’s governing philosophy. Poll: @brennancenter.org
Gerard Saucier reposted
Gerard Saucier reposted
Gerard Saucier reposted
Alex Reisner has found four giant data sets of songs—totaling more than 20 million tracks—available to companies training AI music-generation tools. Search for an artist or song in the data sets:
Gerard Saucier reposted
Crime and Punishment Follow my cartoons and accompanying written commentary on Substack: nickanderson.substack.com For @contrariannews.org
Gerard Saucier reposted
Gerard Saucier reposted
“.. The first rule is to have muddled objectives. .. The second rule is, do not know thine enemy. .. Finally, be sure to alienate your friends.” @edwardluce.bsky.social www.ft.com/content/0ba6...