Tessera
Folgen

Chris K

@thekensta.bsky.social

scruffy looking ERD herder

19 Posts126 Follower282 Folgt
Chris K@thekensta.bsky.social · last wk.

I don’t know the yes-or-no, but how all of this pans out, if AI is a software or infrastructure or other business and if there is anything close to the multi trillion dollar market it demands and who is going to pay that … is SO massively missed in the polarisation around a single vocal journalist

Ben Collins@bencollins.bsky.social · last wk.

But that’s… not the key question. “Is this internecine squabble about something Ed said once?” isn’t the question. The question is “is any of this sustainable, why have we folded the entire economy into this gambit, and is this how we should be spending our money?” Ed is right that the answer is no.

Chris K@thekensta.bsky.social · 4w ago

Mildly infuriating state of vendor wrappers and harnesses - how can this not run the most basic operations? They need to be better in their domain than the model provider

Screenshot showing output of Dbt Clouds AI wizard being unable to perform a simple operation of listing tags in the dbt project
Alejandra Caraballo@esqueer.net · last mo.

This is right. The issue is that LLMs are at best a $50 billion/year revenue type business and they've invested in it like it's going to be $500 billion/year in revenue or more. The mismatch is potentially trillions in capex overspend which is enough to cause a deep recession when the bubble pops.

Better Things Are Possible@internethippo.bsky.social · last mo.

I think it's closer to the truth that LLMs *can* save time on *some tasks* which can be valuable *if* you monitor them and check their output. But that doesn't get you a trillion dollars so they have to squint into the middle distance and say stupid sci fi bullshit instead

Hagen Blix@hagenblix.bsky.social · last mo.

We need to rephrase the whole discussion in terms of hazardous products, instead of "minds". And then we need to make sure that there's liability. A model hacks someone? Put a C-suite guy in jail. A model produces sexual harassment material? Put a C-suite guy in jail. Watch how fast things get safe

Adrienne LaFrance@adriennelaf.bsky.social · last mo.

New just now in The Atlantic: David Robinson resigned this week. He was among the longest-tenured employees at OpenAI—and oversaw safety reports on 12 frontier launches. He is very worried: “The time for trial and error is over.” You can read his essay here: www.theatlantic.com/technology/2...

Micah@rincewind.run · 2mo ago

I may still have more hours in this than any other video game - my parents were the “no violent video games” type and so I had no consoles or shooters but between this and Master of Orion 2 I played the absolute shit out of 4x

Sid Meier’s Civ 2
Retro Blast 👻@retroblastpod.com · 2mo ago

What's your "origin" video game? Most gamers have a lot of favorite games, sure, but there's always one that basically 𝘥𝘦𝘧𝘪𝘯𝘦𝘴 you. I'm talking about the game that flipped that little switch in your brain and made into a life-long gamer. So, which game is 𝙮𝙤𝙪𝙧𝙨?

Adam Becker@adambecker.bsky.social · 2mo ago

@aoc.bsky.social just put out a great statement on AI. Refreshing to see a major politician on the left nail this. "Silicon Valley billionaires want us to believe that Al robots will destroy humanity. But the biggest threat to humanity are the tech oligarchs themselves." 🎯

Charles, Last weekend, Dario Amodei, the CEO of Anthropic, one of the largest Al companies in the world, publicly asked his industry to slow down Al development. Shortly after, Sam Altman, the CEO of OpenAl, followed suit. Then Elon Musk. This all came days after an Anthropic researcher quit, citing his belief that this technology could wipe out humanity and that Al companies are "gambling with our lives." For years, people have warned that generative Al could slip out of our direct control. Those safety concerns must be taken seriously. And this is why I have advocated for strict safety standards on how Al is built and deployed, public oversight over this technology, and using existing laws to hold these companies accountable when they cause harm. But I want to draw your attention to another reality: Right now, the U.S. economy and our financial futures are riding on Silicon Valley's bet that Al will soon be immensely profitable and pay off. And that bet isn't looking good. The formula for a profitable business isn't complicated: make more money than you spend. But the biggest Al companies aren't doing that. They are spending hundreds of billions per year on data centers, chips, and other costs while their ability to make revenues that exceed their already-immense sunk costs are unclear. Experts estimate that these Al companies will require another $600 billion of debt just to financethemselves in 2026 alone. That is more than the entire budgets of the U.S. Departments of Justice, Transportation, & Education combined. Meanwhile, despite almost every single one of these companies not turning a profit & having sunk billions of investor dollars in development of data centers & other infrastructure, Al companies are also on a spending spree of hundreds of millions of dollars in lobbying & dark money in our elections. It would be one thing if we were discussing a few private companies with a poor business model. But last year, virtually all stock market gains - 80% - rested entirely on just a handful of Al companies. That tells us two things: 1. Nearly the entire U.S. economy outside the Al industry is relatively stagnating and faring poorly & the gains from these few companies are disguising it; & 2. Millions of people's retirements - including their pensions, 401ks, & investments - are potentially exposed to this risk. That's why these Al companies' dubious financial positions matter to working people. These companies argue that debt is necessary. They'll tell you that building transformative technologies requires upfront capital, & that massive borrowing today will pay off in generating unprecedented prosperity tomorrow. But this is not the typical debt that startups take on while they build out a company. We are talking about billions going on trillions of dollars leveraged with no indication that these companies will be able to recoup these funds profitably. And they are increasingly looking to the government as their source of cash. In an effort to hide their debt, tech companies have been relying on "circular financing" where tech giants like Google, Amazon, & Microsoft give billions to Al startups like OpenAl & Anthropic, only for those startups to hand that money right back to buy Al processing power. In other words, these Al companies are luring in new investors by making it look like their financials are healthier than they really are.What's even worse is that retirement & pension funds are key to supporting this illusion. When the Al companies issue bonds to raise more money, your pension funds & retirement accounts are some of the biggest buyers. When Musk's Al business went public, it became part of index funds where millions of Americans invest our savings. Many Americans don't even know that their retirement investments are tied up in these companies at all. From Big Tech's perspective, that's a feature, not a bug. By tying our futures to their reckless gambling, Al billionaires are trying to make themselves too big to fail. They think they can take such dangerous risks because we will bail them out if things go wrong. Sound familiar? This scene is dangerously reminiscent of the rampant speculation, corruption, & broken incentives leading up to the 2008 financial crisis that plunged millions of working class Americans into financial chaos that we never fully recovered from. It is no accident that these warnings from Altman & Amodei come months-if not weeks-before these companies were meant to go public. When companies go public, they have to be more transparent about how they are doing financially. And after a waiting period, they typically have to let private investors-like their own employees-sell their shares. The New York Times recently confirmed that OpenAl has postponed going public for this very reason. They are in no rush for the truth about their finances to come out. But the moment these oligarchs began to seek our public tax dollars & public contracts while exposing their companies to public pension funds & investments, they opened themselves to public accountability & scrutiny. Silicon Valley billionaires have rigged the game & are attempting to shift their financial risk to working people: If they're right about Al's potential, they reap the profits. If they're wrong & the bubble bursts, they are setting up the American people to pick up the tab. Heads, they win; tails, we lose.We can stop this, and we shouldn't play their game. Together we can address both the economic and safety threats of unrestricted Al development and prevent financial chaos and stabilize market risk. We must make it clear now that should the Al bubble pop, Congress should not bail out Big Tech. And across our federal agencies, we have the authorities and tools to hold Al companies and their leaders accountable for releasing dangerous, unvetted, and defective products. We must not wait. Silicon Valley billionaires want us to believe that Al robots will destroy humanity. But the biggest threat to humanity are the tech oligarchs themselves. It's time we hold them accountable. Thank you for reading. In solidarity, Alexandria
Courtney Milan@courtneymilan.com · 2mo ago

This is such a great thread because it tells you instantly how to answer the supposedly intractable regulation problem: liability, potentially joint liability, for the person who set up the AI system.

Marco Rogers@polotek.bsky.social · 2mo ago

If I was just a little more condescending, I'd start a web series where I sit down with someone who thinks LLMs are inherently harmful, help them install one on their laptop, and then we sit and wait for it to do something harmful on it's own. 30 minutes of uncomfortable silence. "Any moment now..."

Nicholas Grossman@nicholasgrossman.bsky.social · 3mo ago

Evidence keeps accumulating that LLMs are a powerfully useful technology for people who already know what they're talking about, and a dangerous outsourcing of cognitive function for people who don't. This can occur within a single individual, who knows a lot about some things and not about others.

Nicholas Grossman@nicholasgrossman.bsky.social · 3mo ago

Students who get good grades on homework tend to get good grades on tests. But a recent study out of China shows a divergence. The old correlation holds for students who didn't use AI on homework, but it's the opposite for those who did. Using AI = best homework grades, worst test grades

Natasha Devon@natashadevon.bsky.social · 3mo ago

I didn’t discuss Jason Arday on my show for this exact reason. If he had been white and behaved unprofessionally it would have been handled internally by his employers. As it is, it became clear that certain commentators wouldn’t be satisfied until he was dead www.bbc.co.uk/news/article...

Jason Arday: Former Cambridge university professor found dead

Jason Arday was found "unresponsive" at an address in Battersea, south London, on Friday afternoon, officials said.

bbc.co.uk

Chris K@thekensta.bsky.social · 3mo ago

Enjoyed this. As a layperson it was really educational on technical mechanisms of economic fallout, things I wouldn’t think about if not for this intersection. I feel it nicely says “these bits are wrong” without extrapolating to “your diagnosis is off - AI funding is secure - and EZ is an idiot”

Nathan Tankus@nathantankus.bsky.social · 3mo ago

NEW @crisesnotes.bsky.social PIECE: In this piece I critique a recent piece from @edzitron.com comparing the AI(LLM) boom to the "Subprime" Mortgage Crisis (and implicitly the GFC of 2007-2009). The ? is:"How big of a macrofinancial threat is the AI(LLM) Boom?" www.crisesnotes.com/sigh-no-ed-z...

James McLeod@jamespmcleod.ca · 4mo ago

It's safe to assume that all major LLMs are scraping Bluesky data to train future models. As an open protocol, Bluesky is one of the few sources of fresh human-generated content that anybody can gobble up. So all this is probably going in the models. There are submarines in the Great Lakes.