JS
LightNews LightNews
banner
jays077.bsky.social
JS
@jays077.bsky.social
300 followers 210 following 270 posts
GPCR scientist
Golfer
Banish oligarchs, #RATM
Crypto is a Ponzi scheme
#Resistance
Posts Replies Media Videos
Reposted by JS
jimgumbley.com
Grok 4.1 beta is an interesting model
November 20, 2025 at 10:22 PM
jays077.bsky.social
Let’s hope so!
a cartoon of homer simpson asking for please please yes
ALT: a cartoon of homer simpson asking for please please yes
media.tenor.com
November 19, 2025 at 11:16 PM
jays077.bsky.social
She forgot the parable about the mouse and the snake….
November 16, 2025 at 2:33 PM
Reposted by JS
edzitron.com
The AI Bubble inflated by a dereliction of duty from those meant to seek out and publish the truth. LLMs have been sold on myth and outright lies about what they do or will do, and when the bubble bursts, retail investors will be left with the consequences.

www.wheresyoured.at/premium-the-...
Why Are We In A Bubble? You’re tired, I’m tired, we’re all god damn tired of the AI bubble. It’s been three years and very little has actually happened other than hundreds of billions of dollars being incinerated to build data centers or sustain the reckless and directionless ambitions of Sam Altman and OpenAI, or any number of near-identical AI startups that all claim to be worth somewhere between $5 billion and $500 billion. Yet what truly upsets me about AI is the lies. Despite the fact that Large Language Models cannot really be trusted to do any particular job without endless prompt engineering and tinkering, so many people talk about them as if they’re the second coming of Christ, capable of doing just about anything if you put your mind to it. In reality, LLMs are capable of some stuff but not a lot of stuff and most people you talk to use it as either a coding assistant (that they have to either give very simple tasks or constantly check up on) or as a replacement for Google search that may or may not be better. Yet for years people have said things about LLMs that they don’t actually do, and when pressed to explain further, they’ve said that “these models are getting exponentially better.” We have seen a dereliction of duty by those who are meant to know the truth — the media and analysts — who are either ignorant of reality or refuse to live within it, saying instead that every bit of proof that we’re in a bubble is just “doomerism” or “skepticism” or proof of some sort of corruption or animosity toward the future, rather than an attempt to speak of what’s actually happening and inform the public. LLMs are limited! What they do today is similar to what they did a year ago, and what they’ll do a year in the future! The time of LLM innovation is over, this era is over, and every day we extend it only beckons further retail investors to throw themselves into an active volcano after being convinced that it’ll transform them into a golden god. It’s insu… It is insane to me that we still, to this day, have people claiming that “agents are coming,” or still peddling out the dog-brained “Uber and Amazon Web Services burned a lot of money” talking points, or somehow justifying the hundreds of billions of dollars burned by saying that “ChatGPT is really popular,” as if that makes up for the theft of millions of people’s art or the destruction of our planet. We’ve known that OpenAI was burning billions of dollars since the middle of 2024, as we’ve known the same of Anthropic, as we’ve known that every AI company is unprofitable, as we’ve known that models will hallucinate forever, as we’ve known that the “scaling laws” paradigm was a bucket of bullshit. These facts have sat baking in the sun for anywhere from a year and a half to two years, yet the media narrative has only fairly recently shifted to consider whether there’s any problem with an entire industry of unprofitable companies building wrappers on top of models made by unprofitable frontier AI labs. AI has proven that we live in an era of high information and low processing, and that the majority of people — the media, investors, analysts, and even government officials — simply take the top-line analysis from whatever the most “trustworthy” source they can find. As a result, TV networks like CNBC publish outright nonsense to back up things like OpenAI promising to pay $1.4 trillion in compute costs in five years with “trustworthy” analysts like Futurum’s Dan Newman, who had this to say about how OpenAI might afford it: This is a company that its biggest backers and investors believe will become the largest hyperscaler in the future. So it isn’t just going to compete for these software and for these ChatGPT and video making, uh, models, they're going to be focusing on taking over this AI layer, going down the path of Anthropic, playing the enterprise, developing these biggest models, and then ultimately developing a model to make money. This is a fucking analy… This is why people are going to lose so much money! This is why people are being failed! The mainstream media is deliberately, aggressively pushing nonsense narratives with enough buzzwords to keep the bubble inflated. In reality, our society is built on a lot of beliefs that are extremely thinly-held, and LLMs took root because society has been primed by science fiction to believe that it’s inevitable that we’ll have an autonomous computer in our lifetimes. We are trained to follow smart-sounding people from birth, and what constitutes “smart’ is judged by a combination of right-sounding words and braggadocious confidence, and while that’s sort of worked before, Large Language Models feel as if they’re purpose-built to exploit these semiotic tropes. ChatGPT can convince an imbecile he knows quantum physics because it gives him the topline definition in simple enough terms that he can repeat. ChatGPT can give a convincing-enough impression of a person that it can convince somebody that it wants him to leave his wife. ChatGPT can give a convincing-enough impression of writing code that it can convince basically every investor and member of the media that it will replace software engineers in mere months. And because ChatGPT did a convincing-enough impression of a person writing something, Satya Nadella decided to buy as many GPUs as possible, and because Satya Nadella was doing a convincing-enough impression of a CEO who knew what the fuck he was doing, Sundar Pichai, Mark Zuckerberg and Andy Jassy immediately copied him, so that they too could do convincing-enough impressions of executives that knew what they were doing. And because we live with markets that are poisoned by growth-at-all-cost thinkers, they fell in line behind the hyperscalers, claiming that the “era of AI” was here without any proof that was the case, all because the already-existing businesses underpinning the magnificent 7 kept growing. That, and the media has entirely failed to hold the… There is only one way to avoid this happening again: holding every single booster’s feet to the fire for months and months after the bubble bursts, and make sure that any other hype cycles they push are met with venom and receipts the likes of which God has never seen. If I sound like I have an axe to grind, it’s because I do. Regular people are already getting hurt — their communities destroyed by data centers, their art stolen, their loved-ones driven mad by AI psychosis, and, most likely, their 401Ks ravaged when this ends. I will have no mercy for those who failed to do their jobs. Will you?
November 14, 2025 at 7:00 PM
jays077.bsky.social
a man is singing into a microphone with the words insane in the membrane behind him
ALT: a man is singing into a microphone with the words insane in the membrane behind him
media.tenor.com
November 11, 2025 at 10:19 PM
jays077.bsky.social
Intensity is so much better than last game but we still look vulnerable. Need to get Brenner or Kevin making runs toward goal to open up space for Evander
November 9, 2025 at 12:03 AM
jays077.bsky.social
Congrats on this achievement!

Trapping the whole signaling complex in a latent state is 🤯🤯🤯🤯!

Looking forward to reading the full paper!
November 5, 2025 at 11:28 PM