Sam Altman's Predictions
Where OpenAI may be headed.
I start a lot of videos I never finish and I figured this thirty two minute presentation might last all of ninety seconds. I was hooked right at the start and listened to it all.
“There is a sort of like a new protocol for the future of the internet where things get federated and broken down to like much smaller components and agents are like constantly exposing and using different tools” - Sam Altman
The source here, Matthew Berman, earned a follow and two more full video views after that one. The agent angle got me started, Altman hooked me long enough to wade through it, and at the end I felt … enlightened. Berman’s work scratches my itch in this area.
As I’m drafting this for Thursday the 15th, late on the 14th, he’s getting another thirteen minutes of my time. This is pretty unusual for me, he’s just joined Peter Zeihan, Anders Puck Nielsen, and Tony from China Update - someone for whom I am likely to start viewing all new content as it comes available.
This video is REALLY intriguing - it explores the … almost counseling … that’s required for ChatGPT’s offering to go from creation to general availability. Listen to the wording - they’re not tuning for data so much as they aer for personality. And Berman makes this accessible, at least for me, in a way that no other commentator has.
I probably catch at least a dozen clips this size daily, and offhand there are four major categories.
News & policy, heavy “anywhere but the U.S.” bias on sources.
Nature & science, a mix of the big well known channels and intriguing niches.
Military history & systems, biased for WWII, but often about new directions.
Technology from chemistry/physics to finished systems & software.
This Substack has been in need of a refresh since the 2024 election and this may be a formative change. I nominally have an AI project in the works in the form of the third stage of Shall We Play A Game? What I’m hearing from Berman, even just these first few videos, makes it easier for me to put what I think I’m seeing in context of overall changes in this area.
And if you’d like some nightmare fuel … deceptive, self-copying, system breaking for the sake of cheating AI activity. In the hands of corporate psychopaths.
Nervous yet? If not, you haven’t understood what’s happening here …