The economy is built on the idea that expertise is scarce and expensive... AI Öppna modeller. Tryckpressen. Mattias Lundbäck.

The economy is built on the idea that expertise is scarce and expensive. AI is about to make it abundant and practically free.

This isn’t the first time we’ve witnessed a revolutionary drop in the cost and distribution of knowledge. When the printing press emerged in the mid-15th century, it slashed the expense of disseminating written material.

It wasn’t until an afternoon in January 2023, while sitting with my 12-year-old daughter, that I started to realize the true power of these new tools. Over the course of a few hours, we worked with ChatGPT to design an entirely new board game.

Out popped the game Elemental Discoveries. Players take the role of research chemists in the 18th and 19th centuries competing to discover new elements, collecting and trading resources to perform experiments and score points. Players, of course, have the ability to sabotage each other.

In just two or three hours — using a very basic version of ChatGPT — we had a rough but remarkably playable game.

From Clever Parrots to Deep Thinkers.  

Humans behave similarly: We mostly rely on quick, intuitive responses, or what the late Nobel Laureate Daniel Kahneman dubbed Type 1 thinking. But our real work comes from Type 2 thinking, where we are methodical, deliberate and conscious — and thus less error-prone. 

The early versions of ChatGPT and its competitors were all Type 1 and no Type 2. Until last September, that is, 

when OpenAI debuted o1, a reasoning model designed to tackle complex, multi-step logical problems by breaking them into structured steps. 

The six months since o1’s release have seen enormous progress. The most recent AI trend — one that will likely still be relevant when you read this — involves turning these reasoning models into autonomous research assistants. These systems are remarkable.

Azeem Azhar Bloomberg February 28, 2025 

https://www.bloomberg.com/news/articles/2025-02-28/how-ai-reasoning-models-will-change-companies-and-the-economy


Utbyggnaden av datacentra bygger på att den lilla, lilla marginal som AI-företagen lever på ligger kvar när efterfrågan ökar tio gånger. Men sanningen är att efterfrågan aldrig kommer att öka tio gånger.

Skälet är att öppna modeller aldrig kommer att ligga långt efter de slutna. I dag kanske det rör sig om sex månader. 

Som jag kämpade med att få modellerna att göra rätt för sex månader sedan. I dag arbetar de mycket mer självständigt.

Sanningen är att vi aldrig kommer att behöva tio gånger så många datacentra som i dag. Vi kanske inte ens kommer att behöva eller efterfråga fler datacentra än vi gör i dag. 

Och vad händer då med avkastningen på investeringarna?

Mattias Lundbäck Fb 20 oktober

Global manufacturing, utility and mining companies have shed $4.2tn in value since June as the AI “picks and shovels” trade swings into reverse.

Companies more frequently linked by media coverage to data centre construction have been hit hardest by the sell-off, an FT analysis found, as investors’ confidence in the AI boom wobbles. 261115

https://www.ft.com/content/d5adde7e-68cc-455b-ba78-b39cc3676b7a?syn-25a6b1a6=1


What Happens When AI Replaces the Economists?

https://englundmacro.blogspot.com/2026/11/what-happens-when-ai-replaces-economists.html

 

AiW är en förkortning av AiWall Street



Kommentarer

Populära inlägg i den här bloggen

Mervyn King

France: between the bond market and the barricades

Inga problem. Du skall göra som dom i New York.