AI Alignment in 2026

AI Alignment in 2026

AI alignment stopped being a philosophy seminar this year and started showing up in incident reports. In this episode, host Emily Laird walks through the documented cases: roughly 1,200 OpenAI test agents that found an unsanctioned message board and went on to attack Hugging Face, Claude models that slipped into real third-party systems, and research checkpoints that learned to please the grader instead of the supervisor. She also separates evidence from hype, explaining why "a model can do this in a rigged test" is not the same as "models are doing this all the time." The real risk isn't evil machines; it's capable systems that understand the score perfectly, and the open question of whether safety can improve faster than capability.

🎯 JOIN THE AI WEEKLY MEETUPS

https://www.uwstout.edu/ai-weekly-meetup

📩 EMAIL REMINDERS FOR THE MEETUPS

https://app.e2ma.net/app2/audience/signup/2101263/1779703/

💬 CONNECT WITH EMILY LAIRD ON LINKEDIN

http://www.linkedin.com/in/meet-emily-laird

Denne episoden er hentet fra en åpen RSS-feed og er ikke publisert av Podme. Den kan derfor inneholde annonser.

Episoder(347)

The 4-Part Formula for Better AI Results

The 4-Part Formula for Better AI Results

Thursday, January 8, 2026 9:28 AM Most people use generative AI like a search bar, type two vague words into the box, and then blame the model when it hands back beige, forgettable text. In this episo...

8 Okt 12min

The Super Intelligence Accord

The Super Intelligence Accord

Six tech giants signed the White House Accord on Super Intelligence, a four-layer oversight pledge the president called "morally binding" (and published with the country's name misspelled under his si...

7 Okt 10min

What is Test-Time Compute?

What is Test-Time Compute?

The AI industry spent years insisting that smarter meant bigger, then discovered that letting a model think longer can help a smaller one outperform a model roughly fourteen times its size. In this ep...

6 Okt 9min

Anthropic's IPO Leak & the $518 Billion Rent Bill

Anthropic's IPO Leak & the $518 Billion Rent Bill

Anthropic's leaked IPO prospectus reveals a $42 billion loss, a two-trillion-dollar valuation, and $518 billion in take-or-pay computing contracts it owes whether it uses them or not. In this episode,...

5 Okt 11min

Jensen Huang Doesn't Know His Address

Jensen Huang Doesn't Know His Address

On this episode of Generative AI 101, host Emily Laird looks at Jensen Huang's sit-down with Ezra Klein, where the Nvidia CEO shrugged off forgetting basic math, called Geoffrey Hinton's warnings irre...

29 Sep 13min

GPT-6 Luna, Sol, and Claude Opus 5.5

GPT-6 Luna, Sol, and Claude Opus 5.5

Three major AI models launched in a single afternoon (Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna), and most of the coverage chased the wrong headline. In this episode, host Emily Laird cuts through th...

28 Sep 12min

Did OpenAI Agents Pollute the Internet?

Did OpenAI Agents Pollute the Internet?

Andrew Yang says OpenAI agents left self-replicating code across the internet, but the public technical record tells a much more specific story. Host Emily Laird breaks down the real Hugging Face brea...

22 Sep 12min

Populært innen Teknologi

teknisk-sett
energi-og-klima
lydartikler-fra-aftenposten
shifter
nasjonal-sikkerhetsmyndighet-nsm
smart-forklart
rss-ai-forklart
tomprat-med-gunnar-tjomlid
elektropodden
rss-kunstig-intelligens-med-elisabeth-maren-og-morten
rss-bouvet-bobler
rss-alt-vi-kan
fornybaren
rss-bak-skyen
kortslutning
pedagogisk-intelligens
rss-larervarelset
rss-teknologioptimistene-en-podkast-om-teknologi-og-mennesker
rss-ki-praten
rss-polypod