@peterwildeford The corporation has shareholders that at some point want money returned to them to consume stuff.
Foreign holders of US equity do not consume a meaningful US government provided services.
English
Arthur B.
31.4K posts
Arthur B.
@ArthurB
@Tezos co-founder w/ wife @breitwoman & Agitprop founder. Aligning ASI to not zap everyone is unsolved, that's bad. Tezos stuff, high context humor & more.
arthurb.tez Katılım Mart 2008
1.4K Takip Edilen43.6K Takipçiler
This would makes the C-corp a perpetual tax shelter, where you just incorporate, retain earnings, and defer forever. This is what we had pre-1909 and it didn't work well. We want to be able to capture capital in our taxation system.
Also pure consumption taxation can't reach the ~40% of US equity held by foreigners and tax-exempts. You need entity-level tax to capture that base at all.
Consumption taxes have real merits but are regressive without large transfers.
Obviously there are important trade-offs here.
English
I agree that an "AI token" tax doesn't make a ton of sense.
If AI companies do someday (soon) build a ton of wealth and displace a lot of labor, I think it would be better to just raise the corporate tax rate, tax capital gains as ordinary income above some high threshold, strengthen corporate alternative minimum tax, consider federal equity stakes, and consider an 'excess profits' tax.
Adam Ozimek@ModeledBehavior
Tax compute sounds to me like tax machine tools. Or tax mass assembly. Tax electricity. Tax the steam engine. How about we just like tax capital gains? Tax corporations? wsj.com/tech/ai/job-lo…
English
@DanNeidle Diminishing returns in campaign spend. An extra dollar of campaigning saves him less than an expected dollar not confiscated.
English
Help me understand something. If a Californian wealth tax would cost Brin $15bn/year, why is he only spending $57m to campaign against it?
(I'm not asking about the policy, or the morality, just why he's spending so little)
Bernie Sanders@BernieSanders
Google founder Sergey Brin's wealth has DOUBLED to $311 billion since Trump's election. Now he’s spending $57M to oppose a 5% billionaires' wealth tax in California. He’d rather millions lose healthcare than pay his fair share in taxes. This kind of arrogance is unacceptable.
English
Arthur B. retweetledi
@lawrencenewport Alas, the electrorate doesn't just want to improve people's lives, it's strongly opinionated, and wrong, about how to do that. This is why populism fails.
English
@ArthurB Speaking purely electorally; the public would like them better than they do now!
If they did radical actions that also actually improved people's lives they might even win an election (and ya know, actually fix stuff)
English
Current gov view of the world appears to be:
Fast action = dumb action
Radical = populist and therefore immoral
They have a mental aesthetic that is totally unsuited for a country in decline and a population needing fast results.
A wake up call today? (They won't hear it)
English
@TheStalwart It's very good at stuff you can pick up from reinforcement learning, and then skills generalize a bit from that, but you can't easily learn stock picking from reinforcement learning.
English
Everyone scoffs typing "what stocks should I buy?" into ChgtGPT and expecting good answers.
But I still don't totally get it. If a human could theoretically be good at stock picking, why not a machine trained to think like a human bloomberg.com/opinion/newsle…
English
English
@TheStalwart gwern went into a lot of detail on this all the way back in 2016: gwern.net/tool-ai
English
Great line from Zvi:
"And I’ve been saying, for years, that the problem with this ‘mere tool’ approach, the quest for Tool AI, is that the first thing people would do to Tool AI is turn it into Agentic AI, because an agent is more useful."
Zvi Mowshowitz@TheZvi
English
@TheStalwart The market is efficient because people who are good stock picking are paid to pick stocks.
English
As an EMH guy this is my view too. Humans are only good at stock picking in practice, but not in theory.
Jason Kerwin@jt_kerwin
@TheStalwart A human cannot theoretically be good at stock picking
English
Arthur B. retweetledi
@MattZeitlin Its the frontier labs that are constrained
English
Assuming a uniform prior, you don't bid. You bid b and win. Conditioned on winning, the price is Uniform[0,b] with expectation b/2. So your expected return is 3b/4, but you paid b. With other priors, you might bid.
Mercurius@MercuriusFilius
How would you answer this common Jane Street interview question?
English
Arthur B. retweetledi
📣 Tezos X Previewnet test network is now live
Explore the new execution layer with EVM and Michelson running on a shared ledger.
🔗 previewnet.tezosx.nomadic-labs.com
English
@elder_plinius @allTheYud @justinvincent @grok Very normal technology that sometimes requires you to write your incantation in runes to work.
English
One way know AI is not concious is to realize you can make it say anything, argue any viewpoint, reverse any decision. You can reverse it's opinion as many times as you want in the same conversation.
English
@pixelfondue @RichardDawkins Dawkins says:
"When Turing wrote — and for most of the years since — it was possible to accept the hypothetical conclusion that, if a machine ever passed his operational test, we might consider it to be conscious."
Did Turing actually write this?
English
@ArthurB @RichardDawkins 1. He said Turing wrote, not the Turing test - and Turing did discuss machine consciousness.
2. It's just an anecdotal recounting of his interaction with an LLM to set up the third part.
3. Yes, it is thought provoking and it's the whole point of 1 and 2 setting it up.
English
There are three parts to the @RichardDawkins post. The first where he mistakenly asserts that the Turing test said anything about consciousness, the second where he seems oblivious that he's being glazed — that's the most embarrassing bit people are reacting to — and a third part which is genuinely well formulated and interesting. I think most people probably didn't read past the headline, let alone the second part.
English
@SamuelAlbanie You're absolutely right, I got started on a fix without really understanding the problem, and now the codebase has drifted a lot from where it originally was.
English
"the person who thinks they understand the problem and does not is usually more of a curse than the person who knows they do not understand the problem"
hamming
English
@SamuelAlbanie That's a common question, the less obvious one that he raises: "what is consciousness for?". We evolved to be conscious, it means consciousness serves a purpose... but if something can do what we do without being conscious, then what purpose would consciousness serve?
English
understand the pushback, but liked this question posed by dawkins in his piece
“if these machines are not conscious, what more could it possibly take to convince you that they are?”
worth pondering what evidence would meet that bar imo
English