The Ledger
Every statement that passed quotation and attribution checks. Mix any filter with any other: certainty 1/5, debate potential 5/5, or both at once.
why aren't all 6 resolved? a statement only gets an assessment when the public record can support or contradict it. opinions and what-ifs never can, and 0 checkable ones are still open, waiting for their date. predictions held up or didn't; assertions are supported or contradicted. on every card: ▮▮▮▮▮ certainty · ▮▮▮▮▮ debate potential. speakers are clickable
David Friedberg predicts GPT-4 training costs will drop 100x within 18 months
“If they spent four hundred million dollars in the last couple of years, you could probably assume that doing the same trading exercise could be done for five to ten million dollars, 18 months from now, to generate the same model.”
Altman: Concentrating entire company on single AI models works best
“These systems are so complicated and concentrating bets are so important. Like one You know, at the time, before it was, like, obvious to do this, you have, like, DeepMind or whatever has all these different teams doing all these different things, and they're …”
Allison: Tencent AI rivals GPT-4 in science but fails on political queries
“Tencent's AI machine is a pretty good competitor for GPT-IV in the science or math, but if you ask a question about, you know how do freedom-centered societies perform, it can't answer that question, because the thought of Xi Jinping says this.”
Friedberg: GPT-4o outperforms GPT-4 on benchmarks but trails Claude 3 Opus
“And you can see the scorecard here that GPT four Oh actually outperforms GPT four. And so this goes counter to some of the narrative that in order to get some of the performance improvements and speed improvements they got in four Oh, that they actually made t…”