Most of the time it’s the other way around, at least in the US: because cash and credit card prices are almost always the same it is the cash users who are overpaying, to the tune of 2-3% of the purchase price. It makes no sense to use cash.
Following your analogy, what equity efforts turn in practice is to not only accommodate for track length for those that start behind, but also to cut one leg off of those perceived to be ahead.
My point wasn't that every existing equity effort is justified and flawless, but that there is a clear reason why some kind of levelling is required if you want to live in a fair society - and I do believe most of us want that.
That all sounds great in theory but in practice it devolves not into only giving extra help to those in need, but also to _take away_ from those perceived to have some sort of advantage. See for example NYC's idiotic plan to close gifted and talended kindergarten programs in public schools.
The truth is that it is a hell of a lot easier to lower the bar for everyone than to raise it. I.e. it's a lot easier to make dumb kids than to make smart ones, so in the name of equity we shall have dumber ones.
The ashtrays are there, even today, because it is suspected that this flight [0] went down when someone disposed a cigarette butt in the lavatory trash, causing a fire.
A reminder that aviation regulations are written in blood.
> A reminder that aviation regulations are written in blood.
It's enormously expensive for an airframe manufacturer to deal with the fallout of a crash.
There aren't any engineers in an airframe manufacturer willing to sign off on a faulty design. Some good engineers are so worried about that they get shifted to working on conceptual projects.
I took a loooong time for Boeing to convince the FAA that a twin engine jet was safer than a 4 engine for ocean crossings.
> took a loooong time for Boeing to convince the FAA that a twin engine jet was safer than a 4 engine for ocean crossings
I don't believe they convinced the FAA twin is safer, just that it meets the necessary safety margins. Airlines want them to meet that regulation for fuel efficiency, but I'd want a source that they're actually safe-er, instead of simply safe enough
Now see, the worst part is, I believe you. Your username pops up frequently enough, and is recognizable enough, that I consider you a reasonable, thoughtful person. And the rationale makes sense - juggling multiple engines is extremely complex
But now way in hell can I, in good conscience, repeat that without a source
this plane did not crash, it made an emergency landing 2 miles from the airport in an onion field. Only 10 crew and 1 passenger survived. The other 123 souls aboard died of smoke/CO inhalation from the fire.
the sole surviving passenger, 21-year-old Ricardo Trajano, disobeyed the instructions to remain in his seat.
You don't want to use a standard outlet, since it's not designed to handle full current(?) for hours. There are special outlets for EV charging, and they work outdoors. Just be very sure to have a GFCI breaker behind it.
Can I ask why don't you want a Tesla charger talking to Tesla? Seems a bit odd if you already own a Tesla vehicle that is just piping the data to Tesla all the time.
Fund them to do what exactly? Come up with their own research ideas?
You got the pipeline backwards. The government picks the research areas/priorities then allocates funding for those, and universities apply and compete to get grants. _Then_, once a grant is given to a school, is funding for labs and graduate students allocated.
If the government has no interest in doing research and provides no funding then schools don’t have projects to work on and no money to hire graduate students.
A sizable chunk of the endowment likely has legal restrictions that limit how funds can be spent. E.g., they could be earmarked for undergraduate scholarships or a specific lab at a specific department. The endowment isn't a slush fund.
It's also worth noting that the structural costs of research are far larger than what any single institution would be able to shoulder. For instance, MIT has extremely limited supercomputing resources under their own maintenance. Researchers would typically use such resources from centralized places funded by the NSF or DOE, where larger pools of money can be assembled.
And of course this doesn't even get into the reality that the annual operating costs of somewhere like MIT likely far exceeds the investment returns generated by the endowment.
You might as well argue that companies should never take venture capital - e.g. if they can't finance their growth through profits alone then they shouldn't raise any money. The whole point of grants or investment is to subsidize and incentive work which has payoffs on much longer timescales than what market dynamics can sustain alone.
> A sizable chunk of the endowment likely has legal restrictions that limit how funds can be spent. E.g., they could be earmarked for undergraduate scholarships or a specific lab at a specific department. The endowment isn't a slush fund.
Some of it has some restrictions, but money is fungible. I do not believe that MIT is actually limited (in practice) from writing their own grants because of donor restrictions (if they wanted to).
> And of course this doesn't even get into the reality that the annual operating costs of somewhere like MIT likely far exceeds the investment returns generated by the endowment.
Somehow they spend $1.2B/year on administration, so, yeah. Don't do that. But they easily have enough principal to cover grant funding for the remaining years of this administration. Especially if they can play on their lib donor heart-strings about how mean the current administration is being to them.
The vast majority of the endowment isn't money (dollars in bank accounts). University endowments work like private equity funds, most of the funds will be invested in assets, most of which hardly liquid enough to reasonably convert them into cash on short notice. They could try to borrow money against the valuations of those assets, but it's not sane to take on debt in order to sustain a level of expenditures that was adjusted to a much higher level of income (true more generally). Especially when the alternative of temporarily scaling back expenses is relatively easy.
Money is not fungible when you are a large organization. Many things that should be possible in principle are impossible in practice due to rules, politics, and institutional inertia.
MIT's endowment is ~80% earmarked to whatever purposes the donors considered important. The remaining ~20% is unrestricted, but unrestricted does not mean unallocated. Everything has already been allocated to some purpose, at least implictly. If you want to allocate more money towards something, you need to take that money from somewhere else. And then you get politics.
>practice due to rules, politics, and institutional inertia.
Isn't that the responsibility of the dean to fix? I think a lot of us have no idea how this actually works, but do understand the difference between impossible and hard. This seems more like it's on the hard side than impossible.
A dean is a mid-level manager in an organization, where effective power is widely diffused.
Many things are possible in the same sense as rewriting the US constitution. The mechanism for it exists, but using it in practice would require widespread agreement on the specifics. When there are many people making independent decisions, it's best to see the situation in statistical terms. Outcomes that are too many standard deviations away from the expected are effectively impossible.
> Especially if they can play on their lib donor heart-strings about how mean the current administration is being to them.
Yes, like those famous liberals the Koch family who paid for prime real estate across from Stata.
It's just not as simple as you lay it out to be. Do you _seriously_ think that if hunkering down and paying out of the endowment to sustain nominal operations for a few short years was a viable strategy that they wouldn't be doing just that?
I think this is a valid point, but if the talent pool shrinkage was truly a threat to your academic institution are you really going to just watch?
And the argument is that research funding is coming back but just not to MIT. So I think it is a serious long term issue that they have to consider going forward, and not something that they can just hope goes away.
Government is not great at picking up or creating ideas. Academia has to lead in that and then show government why it would be best for the nation to fund those. The government is good at long term funding for ideas that may not be the best for private sector right away but it should not be creating ideas themselves otherwise you would get things like Lysenkoism.
Allow me to disagree while I look back at the last 80 years of US government-funded research.
The government isn’t “picking” the research topics. It’s the scientists in places like the NSF that are. No system is perfect but some system is better than nothing, which is where we are going.
Local models will never compete with large SOTA models, in the same way an iPhone doesn't compete with supercomputers doing nuclear simulations.
They paths will differentiate and split. Probably SOTA models will eventually be locked down and only accessible to state actors because of how expensive they will be to run (already started with Mythos).
that might be true for us based providers but i dont see china turning closed source anytime soon.
a lot of chinese labs come from big non ai focused cloud services (alibaba, tencent, huawei) who want new models with higher benchmark scores and lower inference cost. they dont care if the competition gets better because its all open so they can build off each others tech, and if anything happens they got other profitable services to fall back on instead of depend on llms only like anthropic.
also the business culture is way different, in vc backed america you would get laughed out the room for saying "there is no moat we just do the same thing as everyone but better". you need to show infinite potential growth and lock everything down to prevent competition but you can get millions to start with no customers and no profits. in china its all about the real money they dont care if your margin is 10 or 90 percent as long as you stay profitable. the llm providers are profitable so they keep their business model.
A Qwen3.6-35B-A3B or whatever it's full name is, when on a 3090, can at the very least, with very little fine tuning, compete with Haiku and blows away GPT4.1 (aka, the cheap models).
It might keep up with Sonnet 4.5 with some tinkering.
But long story short: it seems to have better performance and similar quality for a payoff of a year or so compared to cloud models. In the same way you can self host faster/easier/cheaper than cloud hosting, if you are okay with the negatives.
I'm returning my 3090 soon for a R9700 after some more basic benchmarking, since the higher RAM should improve my observations more.
> It might keep up with Sonnet 4.5 with some tinkering.
I would love to see that. I've been using Qwen3.6 35B and the dense 27B, and they are both too slow with not such great results for agentic coding tasks. It's ok, but not impressive. I had better luck with the BF16 and Q8 than the Q4 from unsloth (really love what unsloth is doing in this space).
Another problem I had with Qwen, which I did not ever encounter with Sonnet - even the BF16 gets stuck and needs a "continue task" prompt from time to time, the lower quants are even worse in that regard.
If you get some interesting results, I would love to read about it!
You don't mention runtime, hardware and harness which are critical. The 35B A3B model should be pretty fast, you do need a decent setup but nothing too fancy. I'm using Q8_XL from unslouth with llama.cpp and opencode and it's pretty awesome. I find that opencode drives the model best, it very rarely gets stuck even with a ton of tool calls. I agree it's comparable to Sonnet 4.5 for most tasks. You may also try the Gamma 4 models which are faster but not as good for coding.
You don't need a (one huge) model to do everything. You need specialized & smaller models that are very good at specific tasks. Collaborating among themselves.
The fact that we see stagnation in terms of billions of parameters shows that efficiency does not scale linearly with the model size. More of an S shaped chart. The middle was Claude 3.5. Since then, it is more about integrating and collaborating with different systems.
its a big assumption that larger models bring any measurable benefit in the long term. there's a point where its not worth paying the expense of a bigger model and we dont know where that will be as both, models and hardware improves.
we do know however where evolution is at right now with our brains, but thats probably not comparable - yet the only thing I can see to make any kind of prediction at all
In the sense that its capabilities were a little overhyped as a quantum leap / unique security threat, and it's so expensive that it's impossible for Anthropic at scale. That hype looked a bit silly when GPT-5.5 matched or even exceeded its capabilities while actually being a model ordinary people could access. But its level of capabilities is only currently matched by OpenAI at this time - the other AI providers, even Google (at least publicly) are quite a bit behind.
You are missing the point. Parents says the market to win need economical models more than SOTA models. Whoever is running those nuclear simulations is not making as much as Apple.
reply