Brian Gordon
@GordonBrianR
Followers
2K
Following
37K
Statuses
28K
Strategist on the innovation frontier. Strategy, cognitive science, innovation, (meta)science, organization
San Francisco
Joined July 2011
If we desire to avoid insult, we must be able to repel it; if we desire to secure peace, one of the most powerful instruments of our rising prosperity, it must be known that we are at all times ready for war.
0
0
0
RT @Noahpinion: Here's a very incomplete list of chaotic stuff that Trump did in the nine days since I wrote my previous post about Trump's…
0
73
0
RT @tshugart3: INDOPACOM: “...manoeuvres around Taiwan right now are not exercises as they call them, they are rehearsals. They are rehears…
0
46
0
RT @JamesSurowiecki: We have a guy hell-bent on dismantling the government who is too lazy to actually read a contract, too careless to dis…
0
899
0
RT @RpsAgainstTrump: If any other president had said anything even remotely close to what Trump just said, an impeachment inquiry would hav…
0
321
0
RT @RpsAgainstTrump: Trump: One of the first meetings I want to have is with Xi and Putin, and I want to say, 'Let's cut our military budge…
0
478
0
Reagan said trust but verify. We would need a reliable intelligence apparatus to verify. Oh, wait, didn’t we just effectively decapitate our own intelligence community.
Trump proposes trilateral agreement with China and Russia to mutually draw down military budgets by 50% "One of the first meetings I want to have is with president Xi of China, president Putin of Russia. And I want to say, let's cut our military budget in half."
0
0
1
I too want a pony
Trump proposes trilateral agreement with China and Russia to mutually draw down military budgets by 50% "One of the first meetings I want to have is with president Xi of China, president Putin of Russia. And I want to say, let's cut our military budget in half."
0
0
2
@matspike @ArrayManta Yep. The free lunch has been a large part of the narrative in this space, though. I take results like this as a very weak counterpoint to some of that hype. But that is a very fair point.
0
0
1
You know what entails a ton of ambiguity, ambivocality, and novelty? Management, law, science, organization, politics. Extrapolating from the current successes of RL in verifiable domains to domains with higher ‘natural’ surprisal might be fraught.
Not actually surprising thing I learned today: the block entropy of computer languages is less than natural languages. I’d assume same holds for mathematics. My hypothesis is that GenAI penetration into various economically important domains is going to be a f(•) of relative H.
0
1
2
Not actually surprising thing I learned today: the block entropy of computer languages is less than natural languages. I’d assume same holds for mathematics. My hypothesis is that GenAI penetration into various economically important domains is going to be a f(•) of relative H.
0
0
2
@infornomics I think there is a case to be made for that, but valuations based on hypotheticals regarding counterfactual workflows won’t be as compelling when it’s time to raise a follow-on round. My op tweet was in half-jest, but it will be an issue I suspect.
1
0
1
@matspike @ArrayManta I agree. It’s not so much a question of can LLM-modulo systems do it; it’s whether an LLM or an LLM+CoT can actually abstract and then use an appropriate algorithm from the ‘mere’ training corpus. Because a lot hangs on whether or not these models can induce procedural knowledge.
1
0
1
RT @Miles_Brundage: Starting to adjust to the idea that ~any “simple reasoning using public information” can be done more quickly and compr…
0
11
0
RT @RadishHarmers: The US government has been spending billions on cancer funding. Cancer! They're buying cancer! That is literally what it…
0
30
0
Where we’re going, we don’t need facts
This "large scale social deception" contract was: * first awarded under Trump * paid to Reuters' data division, not the newsroom * for researching *defenses* against deception * not revealed by "DOGE investigations"; was a public record, visible online for many years
0
1
0
RT @drewharwell: This "large scale social deception" contract was: * first awarded under Trump * paid to Reuters' data division, not the n…
0
2K
0
@matspike @vineettiruvadi I have a feeling this might be a bit controversial. Let’s see what kind of numbers it does on Twitter before rushing to judgment though. I’ve been hearing big things about anti-representationalism!
0
0
1