Stephen Klein

Most of us say we want to think for ourselves. But our behavior suggests we’d rather let machines do the thinking.

@thought_harbor Indeed

@stephenbklein The media should focus on real facts.

The media continues to report on AI companies as if their press releases are financial analysis.

The number one use case for Generative AI is not business, it's personal therapy and companionship. According to Harvard Business Review, 800 million weekly users sharing their secrets, fears, hopes and dreams with a bot. Which means humanoid robots probably won't be much different. Everyone keeps talking about folding laundry. Factory floors. Fencing practice. I'm going to bet that the "drinking buddy" is more likely.

@BigBrainBizness Because thinking differently is hard, but thinking the opposite is dangerous.

When asked a great question at MIT in 1992, Steve Jobs sat in silence for 19 seconds before giving his most honest answer.

@justalexoki It is flattering, you're a creator now, you're a founder now, you're 10x and so on..

@jeffreytucker Maybe because people are being flattered and frightened at the same time.

I've been puzzled as to why AI has been met with such resounding public opposition. Theory: it's not the same as other technological advances that make manual operations easier or offer a new means of deploying resources to achieve human ends. AI seems to target what it means to be human: thinking, knowing, creating, judging. The perception is that fighting AI means fighting for our dignity as a unique species.

@yacineMTB No one knows anymore. Here's the deal. If you are a good thinker and writer, it will always show, and if you are not, no matter what, that becomes more obvious than ever

LLMs are so ludicrously bad at doing anything novel or making any decisions. It's remarkable. I can't believe how far we've gotten on making something useful without actually.. solving it

@tunguz Why are we so easy to be played?

Anthropic really takes us for idiots.

If thinking is getting easier and faster, then you are thinking less.

@ValerioCapraro Many of us are beginning to reaffirm something we almost forgot. Thinking still matters. Most of us knew that. Some forgot.

Students without access to LLMs are 2 to 8 times more creative than students with access. That is the finding of a new paper comparing 2,200 college admissions essays written by humans before ChatGPT with essays generated by GPT-4. The key point is not individual creativity. GPT-4 can write well, sometimes better than individual students. The problem is collective creativity. Each new human essay added new semantic territory. New ideas. New angles. New experiences. New combinations. Each new GPT-4 essay added much less. The authors call this the diversity growth rate: how much novelty each additional text contributes to the collective pool of ideas. Humans kept expanding the pool. GPT-4 made the pool converge. Even when the authors pushed GPT-4 to be more creative, changed parameters, or used chain-of-thought prompting, the homogenizing effect remained. This is the real danger of AI in education. Not that students will write worse. That everyone will write the same. * Full paper in the first reply

@TrustScore_1 When they were raising their private funds, they had developed an AI that was too dangerous to release. They played to the investors. Oh and then their round close and they released it.

Five companies own the intelligence & chips & access—Nvidia, Google, OpenAI, Anthropic, SpaceX A few hundred people own the companies. The rest of us own the consequences. They get the trillions. We get the layoffs. Still think this was built for you? This time they don’t need team, they have automation tools. And they are shouting now Ai won’t take your jobs because they don’t want us to know.

@veryeet No, they can’t be trusted, and they will tell you that. If one is a lazy and careless person, it is the perfect solution.

I fucking hate AI because the fuck you mean you sold out our children’s futures just so you wouldn’t have to write an email

@morganlinton AI labs face massive variable costs. We aren't looking at software companies with 80% gross margins. We are looking at hyper-expensive, high-intensity digital utilities selling computing power at a massive discount to manufacture growth.

Are we really living in a world where we're starting to brag about how many tokens we're using? This is purely a measure of how much money you are paying to AI labs. It does not show if you're writing good code, building useful products, etc. Shouldn't people brag about how many people are actively using the products they build? Or at least some metrics around how well the things they're building are doing? I mean, I kinda feel like this has to be said right?

@GaryMarcus @tomfriedman They are saying this for their IPO, they warn (for those who want excitement), and they talk about a pause (for those who might be concerned). There is something for everybody!

Anthropic played the media like a fiddle. From “untold catastrophe” to “check our latest model”, in two months and a day 🙄 (cc @tomfriedman)

@DataChaz What worries me most is slowly eroding our children's (and ours) ability to think critically. It feels like we are moving toward a place where we are all complacent and very gullible.

accurate

@JonhernandezIA I hope too many people don't get hurt. He will be fine, to say the least!

Elon Musk says Starlink could put something close to an Nvidia GB300 rack in orbit — 72 GPUs, ~150kW peak power, and terabit laser links. The important part is not just compute in space. It is compute that can talk to other racks, Starlink satellites, and the ground with only a few milliseconds of latency. “Orbital AI starts looking less like science fiction, and more like infrastructure.”

@cgtwts The sad state of some of humanity is what I call the "stampede dilemma." Even if some of these CEOs wanted to stop, they'd get crushed from behind.

Companies spent months pushing AI adoption. Now they’re trying to stop people from using it.

@jimstewartson The sad state of some of humanity is what I call the "stampede dilemma." Even if they wanted to stop, they'd get crushed from behind.

HARD RULE: Everything Anthropic says is marketing bullshit. This is nonsense. They absolutely can’t slow down or they will die. This is a cover story for the models slowing down all by themselves. They have stolen all the data there is to steal. It’s diminishing returns now.

@0xsachi As Steve Jobs famously described technology, a bicycle for the mind. Not a wheelchair.

During the age of ai, First principles thinking is critical

@GaryMarcus We are looking at hyper-expensive, high-intensity digital utilities selling computing power at a massive discount to manufacture growth.

I have said exactly this before and still believe it to be true

@Polymarket Indeed, they would say, AI labs face massive variable costs. Every complex reasoning query requires a dedicated burst of a $30,000 Nvidia GPU running at maximum wattage.

JUST IN: OpenAI says the world may need a way to coordinate “slowing frontier development when needed.”