Op-ed
Warning Shots from Meta
Meta has opened up to using content from private messages to 'improve AI.' That's just the beginning of what's to come.
·Anders Eidesvik·3 min
Archive
28 articles
Analysis
Uhm, Has Anyone Noticed That Claude Is Resorting to Blackmail?
New research shows that language models will resort to blackmail when they feel threatened.
·Anders Eidesvik·2 min
Op-ed
A Week with ChatGPT-5. What Now?
Is GPT-5 proof that the scaling laws have finally hit a wall? Or is OpenAI courting the mainstream?
·Anders Eidesvik·5 min
Langsikt Newsletter
The AI Horizon: April
OpenAI launches its most capable (and cunning) models yet. A new report suggests how AI could take over the world by 2027. Nvidia produces 'original' Tom and Jerry episodes from scratch. Here are some of the most important AI stories you missed while you were away for Easter!
·Anders Eidesvik·4 min
Analysis
Hva er greia med Hugging Face?
Alt du trenger å vite
·Anders Eidesvik·3 min
Op-ed
KI-fortjener smartere kritikk
At modellene bommer på nisjeinfo om en norsk skribent er like overraskende som at Yr.no av og til bommer på værmeldingen i Astana.
·Anders Eidesvik·2 min
Langsikt Newsletter
The AI Horizon: May
Google launches AlphaEvolve, the new Pope warns about AI, and Altman sets the stage for an AI race with China. AI development thunders on. Here are some of the most important stories since last time.
·Anders Eidesvik·5 min
Research
AI Models' Ability to Perform Tasks Doubles Every Seven Months
If the trend continues, we will see AI models capable of performing an entire day's work by 2026-27 and month-long projects by the end of 2030.
·Anders Eidesvik·2 min