"The pattern in every case is the same: a 5 to 15% consumer defection layered on top of a regulatory wave that the companies didn’t see coming because their internal polling told them they were fine."
Unfortunately we have somehow untethered consumer/voter sentiment to what government actually does. Regulation may not save us.
Ooo small efficient models wouldn't justify the billions and billions of infra spend. Nor would they make for good marketing content like how "dangerous" this new model is that we're gonna hold it back. Haha but I agree with your premise here!
Efficiency in AI models is improving fast, but it’s being outpaced by scale, meaning gains often get absorbed by even bigger deployments rather than lower total cost
The modern world is all about wading through poorly throught through regulation - forms, checks, pop ups, confirms, notary signatures etc - ie Europe this means we can't actually do any work, can't build a house or a road or a power plant and so most people don't bother. Hence no GDP growth 😂 I can't wait to see how they regulate AI
It seems quite likely to me that over a period of ten years, and for the expenditure of around £100million, one could have a research and development efort that reduced the size and cost of AI systems by a factor of a thousand: that is, a factor of 2 each year, pretty much what DeepSeek did over five years for a factor of 30. It's a matter of some surprise that nobody in an industry with trillions to play with did not choose to risk 0.01% of that to reduce their trillions to billions.
Really appreciate the timeline with the links to the papers. A post to bookmark!
"The pattern in every case is the same: a 5 to 15% consumer defection layered on top of a regulatory wave that the companies didn’t see coming because their internal polling told them they were fine."
Unfortunately we have somehow untethered consumer/voter sentiment to what government actually does. Regulation may not save us.
Ain't that a depressing truth. Projects like Token Offset could become pretty important if that becomes reality
https://tokenoffset.com/
Seeing the timeline like this makes you aware how much happens in how little time 🙈 Great piece!
Ooo small efficient models wouldn't justify the billions and billions of infra spend. Nor would they make for good marketing content like how "dangerous" this new model is that we're gonna hold it back. Haha but I agree with your premise here!
That's the tech sector for you, though. Every efficiency gain just becomes permission to scale up further.
Efficiency in AI models is improving fast, but it’s being outpaced by scale, meaning gains often get absorbed by even bigger deployments rather than lower total cost
The modern world is all about wading through poorly throught through regulation - forms, checks, pop ups, confirms, notary signatures etc - ie Europe this means we can't actually do any work, can't build a house or a road or a power plant and so most people don't bother. Hence no GDP growth 😂 I can't wait to see how they regulate AI
Every industry that bet on consumer goodwill outrunning a regulatory wave eventually found out the hard way which one moves faster.
Read this sir https://theblackswanfiles.substack.com/p/why-the-rest-of-the-world-is-not?r=837k2&utm_medium=ios
It seems quite likely to me that over a period of ten years, and for the expenditure of around £100million, one could have a research and development efort that reduced the size and cost of AI systems by a factor of a thousand: that is, a factor of 2 each year, pretty much what DeepSeek did over five years for a factor of 30. It's a matter of some surprise that nobody in an industry with trillions to play with did not choose to risk 0.01% of that to reduce their trillions to billions.
You’d think! Maybe someone is and its a big secret