Why are we treating it as a derivative, and what's the slope? Do we have proof it's not zero? Let's say Johnny and Kareem both study poli-sci at Harvard. When Johnny graduates with his gentlemanly C's, his dad puts in a word with his golf buddy and it's off to Wall Street. Kareem shoots his resume off through online portals but no one bites. He goes to law school and takes on even more debt. Johnny's sister, Sally, passes the bar on her third try and is hired by a white-shoe law firm. Et cetera.
When you combine the fraction of "hooked" admits with the number of seats affected by affirmative action, you get something between 35%-43% (depending on if you discount hooked admits that were extremely qualified anyway). I think this is an interesting way to frame it. Left to its own devices, Harvard would only devote around 60% of its undergraduate program to simply educating very very bright students. The other 40% is/was for les vieux riches, athletes, generally connected kids, and racial diversity.
While I'm sure it varies by school, I suspect you will find a similar dynamic at many other elite private schools. The undergraduate program is going to be much less important to a top-tier research university, and consequently the admissions board can go nuts with other priorities. When even the runners-up (on a merit basis) are quite strong you can go quite far indeed before anyone would notice a slip in standards.
People have different definitions of progress. I have found that people who are "progressive" on one axis can often be quite conservative on another. Look at the SF Bay Area. While it is quite progressive in the political-ideology sense, we oppose construction that would cause literal progress in the material conditions of the citizenry. "Manhattanization" has been a word used for decades to oppose the thought of densifying SF. My neighbors here in North Bay come out in arms to oppose light bollards on a public footpath. We cannot even progress our footpaths. Rather than build a larger, more inclusive, and cheaper city, you will find countless proponents for rent control - a solution to the question of, "how can I use the law to keep my apartment cheap while refusing to accommodate any more people in the city?"
You are seeing this in this thread. I doubt anyone likes to be described as contra-progress. But nevertheless people would rather conserve the current night sky than see it transmute into a shimmering sea of a million artificial satellites. It's not really obvious to me why one state should be preferable to the other.
It is not human waves, but they have been on the attack for quite some time now relying on small infantry units while obtaining very little ground. What does that tell you about what's happening to those infantry?
I'm curious what the state of alignment research is. My gut says this is basically impossible. People have different moral frameworks. Each individual probably has an inconsistent moral framework. Even granting perfect consistency, applying these typically requires some knowledge of reality. And these LLM / harness combos are turing complete.
So you don't know what it should do, you may not even know what you would do, you don't necessarily know what's happening, and can't predict what will happen. How do you align that?
Seems like these overly sensitive filters are responding to this difficulty.
Yeah and Anthropic is a... dividual consisting of founders, staff, and shareholders, and must comply with various governments ultimately deriving their values from billions of people.
Nobody at Microsoft knows what a good OS is, or what a good dev toolchain is, or what a good videocall app is, or what a good AR headset is, or...
I'm being harsh. I'm sure that there are plenty of people in the org that have taste. And they have a lot of users. But MS is where products go to die (i.e. enshittify).
When you have monopolistic or near-monopolistic control over platforms and other key pieces of software such as office suites, taste is irrelevant since people are still buying licenses out of necessity. Now, Microsoft is capable of writing impressive, tasteful software; I have fond memories of the Windows 95-2000 era, and I also enjoyed Office 97 and Visual Basic 6.0. Even Internet Explorer wasn't bad when it was competing against Netscape Navigator. The problem is Microsoft has a tendency of getting complacent and lazy once acquiring dominance, which breeds bad software.
Apple is more tasteful, IMO, but Apple also has a tendency to abuse its users, taking the attitude, "What are you going to do? Switch to Windows or Android?" Adobe has taken a similar attitude since the Creative Suite became subscription-only. "What are you going to do? The GIMP? Seriously?"
I believe there's some data that they will have different responses if the names signify different cultural / race / gender affiliations. Here be dragons.
Even if it's deterministic that doesn't mean it isn't arbitrary. I can achieve determinism at any temperature by saving the seed. But that wouldn't make rejects feel much better knowing that if a bit was flipped in an arbitrary seed they would be scored differently.
Of course it's arguable. You make it sound like founders perform some jedi mind-trick to take money from others. Here's what actually happens. Investors put in initial money because it's a win-win (they get an expected return, founders get starting capital). Employees join because it's a win-win (they get a salary, health, equity, other perks; founders get a workforce). Customers pay cash because it's a win-win (they get a product or service they want, the business gets money). At no point is someone being held down and forced to hand money to someone else.
Health is not a perk but an inelastic demand: a threat to withhold health is a threat of physical harm, and a negotiation in which one party's physical health is on the line is quantitatively but not qualitatively different from a negotiation held with a gun to that party's head.
I do not understand your statement, maybe you can elaborate. If you are saying there should be a public option for healthcare, I happen to agree. Then we can have the standard discussions on how the government ought to raise funds for it. If you are saying that by negotiating terms of employment, any employer is intrinsically engaged in violence, that stance is pretty out there.
If those terms include the potential for predictable harms like lack of healthcare or housing if an agreement isn't reached, then yes, I think that is indeed an engagement in violence.
Now I'm not saying that the employer is necessarily morally culpable here — I'm sure most employers would like nothing more than to not have to worry about their employees' healthcare, and certainly I doubt many people enjoy having the ability to take it away. But it doesn't change the fact that it's impossible to have a real negotiation when inelastic demands are (potentially) unmet. Someone under threat of losing health insurance or housing is negotiating under duress, contrary to the comment I replied to.
Under this principle no human has ever been able to consent to anything in the history of the world. Certainly 99.99% of humans.
This would also imply that the best thing ethically is not to give people goods in exchange for labor because the simple act of interaction with them puts their housing and food needs under your responsibility.
No human can _100%_ consent to anything (… probably: free will is tricky). Coercion is a continuum, not a binary.
I don't really think that companies (or other parties in trades) bear moral responsibility for this inherently — a company that accepted every job applicant to try to meet their inelastic demands wouldn't last long, so the company itself is also under some duress even if it might like to. Trying to assign blame for complex distributed problems isn't really that simple. Your example in particular is a trolley problem, and I (personally) don't believe that pulling the lever makes you more culpable than deliberately choosing not to pull the lever.
But regardless of your chosen ethics, my point is pragmatic — while it's not correct to say that people take jobs only because they are under duress, it's also not correct to base arguments on them acting on their own free will based on their personal preferences. UBI experiments show significant changes in employee behaviour when inelastic demands are guaranteed to be met and negotiations pertain only to elastic quantities.
This effect is very much not limited to monopolies, though it's certainly easiest to see there. There's no step change from monopoly to competitive marketplace though. If you believe it's the company's moral duty to provide e.g. healthcare then in a non-monopoly situation that culpability is divided, though not abrogated (and beware the bystander effect!). From the employee's perspective, the spectre of physical harm is a bit further off, but it will still colour negotiations.
It's especially insufficient to generalize the working of the entire system from an example of a market in which employees currently have enough power to not really have to worry about the prospect of physical harm because it would be disadvantageous to the employers to cause it. Even if we take the current state of the SaaS startup market as reliable (which it isn't) the original argument was not limited to SaaS startup employees, and in other industries (including ones that are a bit down the pyramid from the SaaS companies) things are a lot less rosy for employees.
A sole consumer of labor is a monopsony, not a monopoly (that would be a union). At any rate, the point is that there are many many employment negotiations that no reasonable person would agree to amount to duress. This is a counterexample to the idea that any negotiation of employment involves duress. I don't need to disprove the existence of any coercive employment. But SaaS companies are especially relevant since pg specializes in showing people how to become billionaires through SaaS. If earning a billion dollars implies some measure of coercion we should be able to find that in a SaaS startup.
Sorry, misread you — but you can substitute ‘monopsony’ into my comment and I think it still holds.
This is a ‘no true Scotsman’ so I don't think I can really respond to it directly. But I'll point out that my claim is not that some contracts bargaining for safety of life and limb are a form of duress but that all inherently are (to some extent). Especially when the other party's BATNA is ‘no guarantee of safety’.
That's not quite true, the police hold me down and lock me in a cell if I don't hand money to a landlord.
They don't hold me down and force me to hand money to a landlord, mind, they just lock me in a cell if I don't, so maybe it doesn't meet your standard of proof.
You would get evicted at most. Even if you're in debt, bankruptcy does not come with any criminal liability. You'd have to do something much worse to receive jail time. And I'm not really sure what this has to do with startups, is the claim that the founders are in cahoots with landlords in the Bay Area to hold employees captive?
Being homeless is illegal in many countries, including the one where I live. If I am evicted because I can't pay rent, then obviously I also can't get a hotel room, and my existence is illegal
I don't think pg would disagree that the politicians that discuss this want to allocate more resources to labor. But what he takes exception with specifically is the rhetoric used to justify this "reallocation." AOC's claim:
> “There’s a certain level of wealth and accumulation that is unearned,” she said. “You can’t earn a billion dollars. You just can’t earn that. You can get market power, you can break rules, you can abuse labor laws, you can pay people less than what they’re worth, but you can’t earn that.”
You can produce a motte-and-bailey-type argument where the "get market power" and "pay people less than their worth" are doing all the heavy lifting in that statement. But I think we can agree that she is very much tying the accrual of wealth to various kinds of villainy. That is what pg is taking on. And that matters because the common person would agree with the statement that you should be rewarded for what you create - if wealth accrual is all theft, that perception would make a much stronger argument for the reallocation of resources.
One can agree that they would rather see wealth more equitably distributed while also admitting that the current system of private property and capitalism is the most effective at broadly generating wealth.
You could say that, but you also don't have to concede it.
In fact, my argument would be that the more regulated, industrial-policy-driven economies of the recent past were better at generating wealth and improving society.
For the most part, the real conflict that we're having around these topics is about the reorganization of the economy that happened starting in the mid-1970s.
This change shifted the focus of the US economy to financial extraction and away from industrial policies, a role that we sold out to China for the benefit of our elite classes and the severe detriment of our working class.
It's an interesting assumption. The idea behind this with nukes was that we'd like to nuke Germany before they could nuke us. Even after we defeated Germany, we nuked Japan even though they had no possibility of getting their own nukes.
The nuclear 'race' was based on the premise that the winner could use it to destroy all other racers (a faulty assumption, see the USSR among others). I will charitably assume Anthropic does not intend to literally destroy anyone and merely wants to become an AGI monopoly. But if AGI is so powerful, any monopoly would not be stable since the incentives for entry into the market are massive. Why would China stop developing AGI just because Anthropic has it?
Do you believe the current situation is more akin to the race to the first nukes, where no one could know for sure the other competitors were even racing...
or is it more similar to the Cold War, where there were obviously competitors engaged in the race?
And yes, agreed the equilibrium dynamics for AGI are very different (and far harder to predict) than nukes. That sounds like a good reason to be sure we get there first since presumably any potential advantage wouldn't go to the second or third runner-ups
I can't really say I see a similarity to either the Manhattan Project or the Cold War. I don't see how one could apply either massive retaliation or MAD. These are private companies, they are not vested with the necessary authority to destroy anything. Even if they had it, they couldn't. You can't destroy China, they have 1.4B people, nukes, and a large part of the world's manufacturing. So multiple organizations want to do something first, that could be anything from nukes to railroads to lining up for communion wafers.
Well it's difficult to argue against something that was never specifically stated. If someone is able to state specifically how this is an arms race in any other way than that it's a race at all then I'm happy to have that conversation.
"Arms race" is the term used colloquially to describe the dynamic that emerges in "winner-take-all" markets.
It seems that the frontier labs believe they're participants in a winner-take-all market. Therefore they're in "an arms race."
Winner-take-all markets do not require that the winner literally destroys the losers, but only that the winner enjoys disproportionate returns compared to their actual superiority.
Whether or not this is actually true is TBD, but I think you're naive to think the frontier labs do not believe this to be true.
I don't know why you think I'm taking anything literally, cf. my first comment. I understand what a metaphorical arms race is. I don't think that Anthropic can forestall others' AI development by getting there first. It can't be literal destruction. It can't be economic destruction (some actors interested in it aren't motivated by money). What's left? I'm all ears.
As far as naivete, wouldn't it be more naive to take their EA claims at face value, rather than the more realistic assumption that they like money?
Creative destruction is absolutely a thing in the market, but the way things are going it seems more likely that open source models will just destroy everything else as far as most users are concerned. The big proprietary labs will be effectively left with Fable, GPT-Pro and Gemini Deep Research - stuff that by all indications needs very large scale compute to even feasibly run. We'll probably find out that each has its own strengths, weaknesses and viable niches, so there's no reason to expect any of those models to utterly destroy the others. They can all survive as specialty services.
reply