ทวีตที่ปักหมุด
MTS
1.6K posts
SITUATION EXPLAINED: The Fable 5 shutdown.
@hammondcheese, chief economist at FAI, live at Manifest:
"The government introduced this 30-day review period for advanced models... Anthropic wanted to release a public version of Mythos. The government said, 'You can only do this if you can nerf these cyber vulnerability capabilities.'"
"Amazon found a pretty narrow jailbreak... the government asked Anthropic to close it. There is no solution to the jailbreak that's 100% foolproof. And so in retaliation, they threw export controls."
"It's extraordinarily unprecedented, not only export controlling you, but export controlling your ability to sell to the British."
MTS@MTSlive
SITUATION UPDATE: Anthropic confirms it is disabling Mythos and Fable 5 for all customers to comply with a US government export control directive. The company says it believes the order is a misunderstanding and is working to restore access.
English
How did a middle schooler become the world's youngest prediction markets podcaster?
@realTomBayes, live at Manifest:
"My dad showed me prediction markets in 2024 right before the presidential election, we were looking at election odds."
"In May of last year I was running for class president of my middle school, I saw some parallels there."
"I asked ChatGPT, 'Prediction market but fake money and anyone can create a market', it said, 'Manifold. You should try Manifold Markets.'"
"I made a Manifold market about, will I win my school's class president election... I did end up winning. But my school did not like at all that I made a prediction market about that."
English
Trillions of dollars will be allocated based on who wins the AI race.
@StephenGrugett, co-founder of @ManifoldMarkets & CEO of @MNX_fi:
"The new company that I'm starting is an exchange focused on AI-related markets: GPU compute prices, electricity, and AI benchmark numbers."
"Anthropic and OpenAI are about to go public... valued over a trillion dollars. The question of who will win the race to AGI will determine how trillions of capital is allocated."
"And yet the most obvious question you would have about these AI labs is... who is gonna have the best model? What rate of progress will model improvement be at over the next year, two years, 10 years?"
MNX@MNX_fi
Trillions of dollars are flowing into AI, and there is almost nowhere to trade it. There’s no exchange for hyperscalers to hedge their $100M training runs. No benchmark markets to bet on the speed of AI progress. No public markets on OpenAI despite them being worth more than Goldman Sachs. The most important sector in the global economy has less financial infrastructure than soybeans. Until now. Announcing MNX, the first exchange to allow trading across the entire AI value stack, from H100 rental prices to Nvidia shares. Trade the takeoff at mnx.fi
English
Peace through strength is the way forward for defense tech.
@alexbronzini, writer at New York Magazine, live at Manifest:
"Popular opinion in America is and has been for many years anti-interventionist."
"But if defense tech can position itself as peace through strength, we are helping America build deterrence capability, we are helping America intimidate its adversaries such that they don't want to go to war against us, that could totally be a popular proposition."
"This is somewhat the Anduril positioning... we're going to develop the most powerful autonomous weapons so that all of our enemies are intimidated into oblivion."
English
What is the most counterintuitive finding in learning science?
@tracewoodgrains, founder of @educprogress:
"One of the biggest barriers to effective learning is that often the more effective education is, the less people intuitively like it... effective learning is hard work. Effective learning is often unpleasant in a way that ineffective stuff is not."
"There's this great video from Veritasium... they show people science videos that just tell them the right thing. Everyone says, 'This was clear. This was easy. I liked it.' And then you give them a test on it, and they retained basically nothing."
"They automatically slotted it into whatever model they already have of the world... it fits with whatever mistaken conceptions they have, and they just get things wrong."
"Whereas if you say, 'This is wrong, this is wrong, and now this is right'... students say, 'That was kinda confusing. That was unpleasant.' And then you test them, and they perform significantly better."
English
MTS รีทวีตแล้ว
SITUATION EXPLAINED: Why is effective compute growing so fast?
• Hardware manufacturing improvements (Moore's Law): 1.5x per year, for six decades straight
• Investment in compute hardware: 2.5x per year, bigger than Moore's Law itself
• Combined, these two have driven ~4x per year growth in training compute for the last decade
• Algorithmic efficiency improvements add another 3x to 6x per year on top of that
@theojaffee: "Four, sixteen, sixty-four, two fifty-six, and it just keeps going and going."
MTS@MTSlive
SITUATION EXPLAINED: Effective compute for AI training is growing at roughly 10x per year. • "Effective compute" combines three things: more hardware, more spending on that hardware, and algorithms getting more efficient • Year 1: 1x. Year 2: 10x. Year 3: 100x. Year 4: 1,000x. Year 5: 10,000x • That's a 10,000x improvement in 5 years • A million x improvement in 7 years • A billion x improvement in 10 years • Epoch AI says 10x per year is on the low end of estimates, the actual rate may be higher and accelerating @theojaffee: "Do you understand how insane it is for anything to grow an order of magnitude per year?"
English
SITUATION EXPLAINED: Effective compute for AI training is growing at roughly 10x per year.
• "Effective compute" combines three things: more hardware, more spending on that hardware, and algorithms getting more efficient
• Year 1: 1x. Year 2: 10x. Year 3: 100x. Year 4: 1,000x. Year 5: 10,000x
• That's a 10,000x improvement in 5 years
• A million x improvement in 7 years
• A billion x improvement in 10 years
• Epoch AI says 10x per year is on the low end of estimates, the actual rate may be higher and accelerating
@theojaffee: "Do you understand how insane it is for anything to grow an order of magnitude per year?"
English
Every consumer product has to rethink its interface.
@andyfang, co-founder of DoorDash:
"For the first time ever, agents are more web traffic than humans, whether we like it or not, we need to figure out a way to change our interfaces."
"I used to think exactly the same way Brian Chesky did, I don't want a conversational experience in the product for consumers."
"But people's behaviors are changing, Google has publicly stated that search query lengths have only gone up over the past couple years. People are trying to speak to it like a human versus translating what's in their head into keywords."
"We've grown up doing that keyword translation for decades, we just assumed that's the most natural way to interact with computers. But now that these AIs are here, people are just trying to speak to it like a human."
@julesrosenberg
Andy Fang@andyfang
Today we’re launching Ask DoorDash — a new conversational way to search the app in your own words through chat, voice, a recipe link or photo. Ask DoorDash can build you a grocery cart ~5x faster than doing it manually. It takes a single prompt to complete your cart in under 2 minutes. In early testing, nearly half of all restaurant orders made with Ask DoorDash were from a place the customer had never ordered from before, and grocery baskets built with Ask DoorDash were over 35% larger than those without. [1/3]
English
