Could we not get the same with EAFT? Maybe that’s what it’s doing but definitely not the first to think “let’s lock in high probability solutions”
In nemotron the high perplexity solutions are selected for RL, in VLM training a few people are looking at the entropy distributions of the training set, etc
These are completely orthogonal. That’s cool if you want to appeal to an in group, but I think you’ll find that a huge portion of the country thinks that we should have rules around immigration. So do most other countries.
You should probably argue your actual position instead of “your guy bad my guy good”. This comment is more Reddit than HN.
No, it doesn't solve any. You know this because of all the lying and child rape and murder. That's not how people who genuinely want to solve problems and improve society act.
It's basically just a power grab by a tiny fraction (< 1%) of the population against everyone else (including the rubes). And it doesn't make the rich happy either because we are social animals. It's better to be poor in Finland than super rich in Somalia.
Training for tasks still works petty well, but “vision” is a super broad domain and most seem optimized for OCR and screen processing (which have verifiable outputs and relatively straightforward data generation)
A weird claim when science is littered with a history of poor, insane explanations for phenomena.
People play back the “Greatest Hits” without really going into the historical misses. The reality is that the quality and predictive power of science is covariate with culture.
There are a lot of good reasons to think that academic culture right now has a groupthink problem, mostly because the group is so much larger. Alternative theories typically have to wait for an incumbent group of thinkers to die. But if the gradient of thought is more continuous then do bad ideas become more sticky?
what types of problems are you seeing this for? i'm wondering if it's just a difference in domains, but this is not our experience at all. llms will write something, but the odds that it is efficient for any problem with deep complexity has been basically zero.
if it really were fuzzing and finding different candidate spaces, then I'd expect it to be good at things like dynamic programming, where I've only seen it fail. usually i see it get stuck in a bad solution and just thrashes around in that minima. these are problems where we can construct a verifiable test space, and it will eventually wind up with a solution, but one that is thousands of lines long and uses no structure of the problem space
Is there a reason to prune individually instead of introducing a tombstoning approach like kimi?
Most of my harnesses around agentic retrieval wind up implementing the poor man’s version of this via isolated context windows and recursion. But it seems like an entire trajectory is more likely to be erroneous than its docs, and you could just rewrite true positives in poor trajectories as the summary?
It’s probably the single worst decision of the entire bill and one of the largest wealth transfers in history.
If you tell me you’re going to light your house on fire and then ask me for fire insurance, I should be able to say no.
Instead what we have is not insurance, but the world’s worst socialized health plan. Insurance is for managing tail risk, not for distributing the cost of healthcare. If we’re willing to pay a tax to subsidize the elderly, we should cut out the middleman and let the government fill that function.
ACA enshrined the worst parts of the American healthcare system for years to come. It is a politicized victory that is the best solution for no American citizens. Places I’ve been with fully privatized healthcare or single payer are both significantly better for consumers.
Insurance companies have raised prices to restore profit, were briefly a mandatory expense, and will exist for years to come.
Not really, because whereas before things were bad for people with pre-existing conditions, now they are really bad for everyone.
People are paying exorbitant prices either for insurance, for routine health care stuff, or for both.
There was no free lunch, so we traded some health care for the chronically ill, for slightly less healthcare for everyone else. The insurance companies make sure it's an extractive zero-sum game in terms of actual healthcare provided.
When I was looking for healthcare after my employer went tits up right before the ACA went into affect, I couldn’t buy insurance at any price because of a pre-existing condition. Mine you that I have had my ore-existing condition since birth in 1974, had one surgery my entire life - foot surgery in 1996 - and this was 2011 and now I’m 52 and still haven’t had any complications from it and don’t expect to.
I was also a part time fitness instructor, runner and could past any of the standardized fitness assessment standards for someone who was 5 years younger as far as push ups, sit ups, running etc.
I had a contract so I could have easily paid more based on risk.
Before anyone mentions COBRA, that’s only an option if your former group plan still exists and it didn’t when the company went out of business.
Just looking, even now the ACA Silver for my wife and I would be around $800 a month in my state. Even ignoring medical costs have gone up more than inflation, that would have been $550 a month in 2011 if the ACA had been available.
Before ACA, insurance had a more traditional "insurancey" role by excluding pre-existing conditions (aka managing moral hazard) in order to make money via premiums. In the "guaranteed issue" world post-ACA, insurance companies have pivoted instead to extracting as much money as they can from an increasingly vertically integrated ecosystem (PBMs etc)
You have the two mixed up. Insurance companies - even for group insurance like through your company where they always had to accept everyone - required you to have “continuing coverage” and not have gaps or you had waiting periods.
The ACA also was written to enforce that through mandates and subsidies - a carrot and stick approach. The moral hazard was caused once there weren’t any mandates because of lawsuits by Republicans and the insurance companies still had to accept everyone.
We are talking past each other. When I said moral hazard I'm talking about adverse selection(1), my bad if that was unclear. And I was responding to some comments about denial for pre-existing conditions, which as you point out is irrelevant for continuous coverage group policies. Removing the mandate penalty without also adjusting the pre-existing condition protections did introduce adverse selection in the current regime. None of that disqualifies my argument about the current state of affairs though re the business model of modern insurance companies
If you were just looking to shout "dems good GOP bad" and find others who agree with you, that's cool too.
Before ACA, insurance had a more traditional "dump you when you were in need" role that leveraged pre-existing conditions rules by, if you fell ill in a way which was likely to be sufficiently expensive to make this profitable, looking for any minimally tenable evidence of an undisclosed pre-existing condition (just to have something to cite as a pretext, it didn't need to be convincing), using it to justify cancelling your insurance, and avoiding any legal remedy by refunding your entire lifetime of premium payments.
Of course, whether or not you actually had a pre-existing condition when you had signed up for the original insurance, you now have one that prevents you from getting new insurance,
The problem here is that pre-ACA you didn't have insurance, either.
Yes, they would (maybe--some plans saw huge sticker shock because the original didn't actually pay much of anything) sell you "insurance". They would offer a plan for a few years, then close it, offer something new. The old plan would see patients getting sick, costs would rise. Premiums were based on costs *for that plan*. Soon the healthy jump ship for something else, now the old plan is only the sick and the premiums go into a death spiral.
Thus the reality was that any ongoing problem soon you were uninsured.
> Why is that inherently bad? Should I be able to buy fire insurance on pre-existing embers?
What if someone gets Type 1 diabetes as a child so they can no longer get insurance because of that "pre-existing" condition: if they get cancer for unrelated reasons they should just be saddled with medical debt? Or because of your Type 1 you can't get coverage, and you get t-boned in your car by a drunk driver.
Certainly it sounds 'unfair' that someone who smokes (a personal choice) gets similar cancer coverage for someone who does not smoke. But it also means that if your ((great-)grand-)mother had cancer, and you get it through no fault/choice of your own (i.e. genetics), you can also get coverage. (This latter effects a cousin of mine: her aunt (mom's sister) died of cancer at 37, her mom at 63; so now she's wonder when here number will come up. We're in Canada, so have universal care, but it's still something in her DNA.)
There are many circumstances in which you suffer through no fault of your own, and universal health coverage is present in many societies because it was decided to protect those people—even if it allows some 'free-riding' by others making poor choices.
People make all sorts of crazy decisions to prevent the "wrong" people from getting what they "don't deserve":
Pre-existing conditions also continue to frame healthcare as 'insurance' against a bad thing happening to you, when it should just be a regular service like any other.
You don't need 'insurance' in order to get your vehicle serviced, but that is what the US does with healthcare.
The most it will ever cost me to go from “not having a working car” to “having a working car” is the cost of used car that will reliably get me from point A to point B.
When one of my kids was 4, they had an unexplained seizure. Hospital workup, whole nine yards, never recurred; it was probably a medication reaction. Years later we were denied coverage from all the private insurers over it (more accurately: we were denied any coverage for that child).
Similarly, insurers would as a matter of course exclude from coverage any woman with one of several extremely common conditions, including endometriosis, PCOS, fibroids, and adenomyosis.
Prior to Obamacare, insurers were free to deny coverage wholesale for these conditions. It would have been fucked up to extend coverage but exclude any neurological conditions from my kid, but the actual outcome was worse: they were under the law entitled to withhold any coverage.
If you live long enough, you will have a pre existing condition.
The way it was suppose to work with the original mandate is that everyone had to be insured either through their employee or the exchange. So you couldn’t just buy insurance when you were sick. The Supreme Court struck that down.
If you lost your job, before the ACA, you could not get health insurance outside of working for someone and having group insurance at any cost.
But you do realize that the entire idea of not being able to get insurance because of pre-existing conditions is completely unique to the US?
Costa Rica for instance (where I am right now for a month and half) allows anyone to become a resident as long as you have guaranteed income of around $2000 a month or you deposit $60K into a local bank account and they arrange monthly disbursements and you pay 15% of your stated income to CAJA. Healthcare is both better and more affordable here.
The same is true for Panama. Why can’t the US figure this out?
And if you had an expensive dependent better work for a big company. Saw it happen--wife lost her coverage through work because they had an employee with an expensive kid. Next year insurance wasn't offered.
It interacts badly with insurance being offered as workplace benefit. If you quit or lose your job, you'd lose your health insurance. And any plan you signed up for after that would then treat you as "pre-existing embers" and expect you to pay accordingly. The bundling of health insurance with workplace seems like the healthcare original sin to me.
Obama couldn't change that, so the ACA redesigned the system to work with it. Despite being called insurance, health insurance is no longer really viewed or designed to be any kind of insurance. Instead, it's supposed to be Netflix for healthcare. You pay a flat rate, and then get unlimited healthcare. Obviously, the issue with this is that if you don't need healthcare you can just not sign up for the subscription. So the ACA tried to solve this by requiring everyone to sign up. Once everyone is required to sign up, it's not right to discriminate against preexisting conditions. It may not be an especially good system, but it is coherent.
The US is allergic to taxes. Maybe it's a marketing thing. Benefits paid for by society.
Maybe a department of Return on Investment. See what those taxes pay for. Contrast to buying private versions of the services at the same SLA or better.
It’s more that the US is more like a collection of 50 little countries, and it’s supposed to be hard to accomplish much at a federal level. That separation has eroded a bit in the last 50 years but it’s still very much a part of our political ideology.
it's bad for the person, obviously. The point of society-wide policies is not to maximize economic efficiency; they're supposed to making society a good place to live. Of course if you only look at them under an economic lens they're going to seem bad. Economically the best policy would be to kill all the sick people.
We as a society accept the insurance system as an implementation of "funding healthcare" because market capitalism is supposed to lead to lower prices, fair allocation of scarce resources, and innovation, among other things. That is, the insurance industry is a market solution to a moral problem.
If insurance companies then can wiggle out of covering pre-existing conditions, they're no longer solving the moral problem they were brought into the world to solve, and now we need some other solution to solve the rest of it. Then, whatever that other solution is, it's solving the hard part, so why not extend it to solve the whole thing and cut the insurance middlemen out of the economy entirely? What are they even doing at that point besides extracting a rent?
(This is one answer among many good ones to what is really a bad-faith question—health-insurance is not a lot like fire-insurance at all)
In nemotron the high perplexity solutions are selected for RL, in VLM training a few people are looking at the entropy distributions of the training set, etc
reply