logisticseh

> Naming a government contractor Palantir is a joke on whoever did it of course, since the palantirs destroyed everyone who tried to use them.

Sometimes trolls are all in on the same joke.

> I think this is missing the forest for the trees.

Not at all. People have a sense of right and wrong, and this fact offends them far more than the nebulous task of assess the effectiveness of research.

I have a PhD and spent a lot of time within and interfacing with academia. I think even the average science-skeptic congressperson massively over-estimates both the short-term and long-term value of the sort of research that happens in most of academia. But we're probably not going to agree on that.

So, see, we can go back and forth all day about the relative merits of academic research. And there's always cover because most congressional people -- let alone voters -- don't know enough to have a really informed opinion.

If I go to a congressional staffer and complain about the NSF or NIH, not much is going to change unless the candidate already agrees with me and is passionate about the issue. They know voters don't care.

But I can go to a congressional staffer and ask: why does your candidate support bailing out student loan debt while simultaneously funding multiple annual week-long conferences in Hawaii and Europe for professors who scoff at their teaching assignments?

And THAT question is going to be viewed very differently, because it's something that could anger voters who don't have the time and attention to vote on the basis of whatever useless crud the NSF is funding this year.

(It's not, especially for disciplines like CS and math.)

In wet lab disciplines the story is more complicated. But there are many possible configurations that are more efficient than the current university-as-gatekeeper system.

2. He raised $20MM for an AI startup, which is fine and well but also makes him not entirely disincentivized from hype.

3. I wouldn't characterize him as someone in the trenches of deep learning.

More of a meta point: technical depth in more than a few things is impossible in a human lifespan, and just a bit harder once you become a "somebody" since a portion of your life becomes consumed by the fact that you're a "somebody". You end up doing things like raising VC money and starting companies with bold ambitions. Its own time sink.

I had this realization when I had a conversation with Lamport about a niche topic in distributed systems and he expressed a position that was just wrong. It was a minor point that didn't really matter much at all, but he was pretty confident in a conjecture I knew was wrong. To be clear, the fact that no one can be an expert on everything -- even everything within a subfield of CS -- doesn't detract from the fact that geniuses exist. Someone can "forget more than you know" and also not know something that you know. Life is just sadly very short.

IMO, the US needs to clean house. We should restructure our research funding so that it goes to motivated and curious people instead of politically connected PIs. This means totally disconnecting NSF/NIH/DARPA funding from the university apparatus. Treat academic affiliation -- and the associated albatross of "overhead" -- as a hindrance to grant applications as opposed to a requirement.

Imagine how much better the caliber of our research will be if we pay grad students $100K instead of paying them $35K and their university $65K. We could 10x our research spend by giving money to the right people instead of giving it to people who can stomach and afford giving 2/3rds of their funding to sad excuses for institutions that American universities have become.

At the very least, tax payers need to stop bailing out student loan debt while simultaneously funding multiple annual week-long conferences in Hawaii and Europe for professors who scoff at their teaching assignments.

For Computer Science:

1. replace the current academic funding model with pure fellowships. Each individual, from most junior to most senior, gets their own N year funding.

2. Each has a legal entity under which their IP lives and in which the government takes a small, fair, non-voting share.

3. Completely divorce this funding infrastructure from universities -- if someone wants to use part of their grant to pay for PhD courses/advising, great, but make it so that funding science is not contingent on that institutional apparatus.

For lab sciences things are more complicated.

1. Actual deep learning researchers, actively doing research. I rarely hear nonsense about impending AGI or mass technological unemployment from these folks. They might express tepid concern about the ethics of various applications of deep learning and tepidly point out that some technological disruption in employment is possible, but in both cases with an emphasis on tepid. And who, if anything, tend to under-estimate progress. (Self-driving is still very far off and also much further along than I thought it would be 9 years ago. I turned down jobs at self-driving cos because I didn't believe they would make enough progress to even have an ADAS product, but I was clearly wrong and if I could do it over I'd work on self-driving from 2013 onwards even though L5 is still a long way out.)

2. Deep learning fan-boys, for lack of a better term. The rationalist community in particular has a sub-community of tech-adjacent folks who aren't publishing in major conferences every cycle or running research labs but do talk a lot about AGI/UBI.

IMO it's not that dissimilar from climate science or even in the extreme the existence of aliens. Scientists with a lot of real expertise will sort of tepidly talk you through the full complexity. And then some "true believers" who aren't actually expert will sort of run to the extremes of anti-natalism or aliens among us. If that makes sense.

The servicers are paid to service; they don't own the instrument. You own the actual debt.

Think of it as a particularly regressive tax -- levied against everyone (including the SEVENTY PERCENT of people who couldn't access higher education) to fund the relatively small and precious set who got a four year degree and couldn't even manage to pay off a used Toyota Corolla with the skills they learned.

If this angers you, write your congressperson and urge them to completely defund the research agencies -- NSF, NIH, DARPA, AFRL, the National Labs, and so on. These institutions are rotten to the core and are today completely controlled by twitter personalities. We can get a working man's jubilee by killing off funding for useless junk.

I have a college degree and a PhD. I have faculty in my immediate family. We feel the same way. Hit academics -- and academia -- where it hurts. The research these agencies fund is NOT an investment in research. It's just BS soft money that allows professors to be a jet setting group.

I promise you that these faculty literally laugh about their obligation to the US taxpayer, right before flying off to Europe for skiing or a nice summer vacation after their conference. Pick a faculty member and go read the names of countries where they present their papers.

I also promise you that the research is useless. Go read the proceedings of any CS conference.

To pay working class folks back for student loan jubilee, we should start by cancelling Twitter-personality-cum-faculty's summer jet setting vacay by defunding the NSF's Division of Computer and Network Systems.

I am highly confident in my assessment that the personalities found on the typical R1 tenure track are exactly the sort of personalities I avoid hiring or working with at all costs. There are exceptions, but they prove the rule (and I can often poach them anyways).

I don't think I said anything about industry other than that it pays 3x-5x better than the TT, and I'm pretty darn confident that's true. I am clear-eyed about the issues in industry, but the personalities are much better.

I really do believe that the massive pay disparity between CS industry and CS academia is, in part, a "toxic personality that can't play well with others" tax. And I really do believe that you'd get more mentally/emotionally healthy people on the TT if it paid better.

Anyways, we can agree to disagree, because we agree on the solution in any case.

I'm generally nauseated when I interact with American CS academics. Every time I attend a conference, PC, or NSF panel, I am so glad I chose industry. It's like IRL twitter.

(Europe seems to be better for some reason.)

> If you're smart enough to rise, you'll get an offer from the private sector you simply cannot refuse. It doesn't matter if your passion is Academia, they can and will buy you out and own whatever you're working on.

IME it's less about "offer you can't refuse" on the industry side and more about "offer you can't take" on the academic side.

After 6 years of deferred income I simply could not take a job that paid $80K-$100K in an HCoL area or $65K-$80K in an LCoL area. I had loans to pay back, no 401K, and not enough savings for a down payment.

If you want good people to stay in CS academia, I think a few things need to change:

1. First, and most importantly, the faculty culture. I don't really know how to describe the problem, but "the old folks are checked out and the young folks are Twitter personalities" is probably close. What's the point of being in academia if you have to be surrounded by the intellectual equivalent of used car salesmen, especially when you can go to industry and do interesting work without the BS?

2. Double the income of PhD students so that they aren't financially ruined by choosing the academic path. This isn't a super unreasonable request -- they'd still be paid less than their peers in industry while doing what's effectively a full time job.

3. Pay faculty more. Not a lot more... just, like, "at least what my undergrad students make at their first job after graduating".

I think if you solve items 2 and 3, then item 1 will take care of itself.

1. billed hours in the trades don't include travel to or from the worksite; if the plumber is 30 minutes from you and the job takes an hour, they're already down to $75/hr.

2. software contracting can be done from anywhere. You can live in rural Kentucky and charge a high rate. But plumbers and electricians in rural Kentucky probably aren't charging $150/hr for house calls.

Specifically, the 130 or so institutions listed here: https://en.wikipedia.org/wiki/List_of_research_universities_...

Heuristically, think "major private universities and flagship public universities".

I've reviewed for some (high quality) Mathematics journals. Papers tend to be more complete, for sure, but the reviewing is much less rejectionist. I'm not aware of any Mathematics journal with a 10% acceptance rate, and even 20% is probably on the low end.

> It's an unfortunate reality of academia that there are fewer resources (jobs, grant funding, etc.) available, than there are researchers who are prepared to put them to good use.

I don't think this is true in CS. Universities outside of an elite set really struggle to hire and retain high quality faculty. It's at a crisis level outside of R1. Teaching-oriented institutions have mostly have stopped trying to hire traditional academics; a masters degree with some teaching experience is sufficient.

Some of this is due to industry -- high-quality faculty candidates tend to also have 3x-5x offers in industry, and it's hard to turn down a guaranteed early retirement for the grind and uncertainty of the tenure track. But I think some of it is also that students who would make good teachers and mentors lose confidence due to a series of unnecessary paper rejections and decide to nope out of academia.

Again, I spend a lot of time around academic mathematics. The rejectionist culture in CS is real. And not just conferences, btw. An NSF program manager started my last review panel by telling us that scores are consistently way lower in CS than in any other field and to please chill out.

2. With respect to R&D, one big difference is that the government doesn't provide seed funding. They provide grants. If the government wanted equity in research labs, they'd have to pay a lot more. You'll see this in practice if you ever have the extreme displeasure of doing non-useless research in academia. Companies that insist on IP ownership/sharing end up paying much higher premiums for university research contracts. Repealing Bayh-Dole would have no effect on the accessibility of actually useful research; universities and companies would privately fund the useful stuff and leave the government to fund the labs of politically-connected/twitter-famous but otherwise totally useless academics.

(To be clear: we're on the same side here with respect to open access publications.)

The USG spent $6B on cloud computing in 2020. That number is increasing quickly. To say nothing of the massive quantities of non-OSS software that the government buys and incorporates into is own business-critical processes. And it's not just government licenses, but also anyone who interacts with the government. E.g., try interacting with any government agency without an Office 365 license.

You get really funny looks if you say that MSFT should have to give away Office 365 for free if the government is going to use it for anything.

But total USG spend on closed-source software has to be well into the 30B-50B range conservatively. For reference, the entire NSF budget is $10B.

The main reason for this is that there are many monied and powerful stakeholders who benefit from selling closed software to USG, whereas the academic publishers a tiny, often not even American-owned, and got super greedy and screwed their natural contingency (academics hate them as much as or more than anyone else).

UI - Logging mouse movement, keyboard events, and other types of attention proxies.

Webhooks and/or Bottlenecks - calling out to either internal or third party classifiers, or more recently even generative models, for personalization based on user tracking data.

And I don't think this is just my unlucky experience. The fine article includes user tracking as one of only three explicitly enumerated reasons for queuing:

> Why? Because users increasingly expect a real-time experience. In use cases like order flows, webhooks, user tracking, etc. users expect to be able to see the new data in the user interface instantly, instead of having to wait for some background batch processing to periodically reload.

Of the explicitly enumerated motivations in the article:

1. "Order flows" - tracking/modification is often one of the higher latency items in order flows ("you might also like" / "what to order next" features).

2. "Webhooks" - often used for tracking/personalization

3. "User tracking" - ...this one is easy :)

When I try to estimate ROI on those features it's usually miniscule. A/B test are unconvincing and the aggregate reports synthesized from the data look more like numerology than rigorous science. More-over, senior leadership usually doesn't realize how much additional liability they take on by collecting and storing so much customer data.

It's not 2010. The large tech firms are your direct competitors and regulators across the world have caught up, including in many US states. If you have a primary product for which people will pay money, then the "surveillance agency as a side-business" model of the 2000s-2010s is probably a bad idea. It's a net expense that exposes you to liability and distracts your team.

Additional benefit: without real-time analytics and personalization, you can go even further without a message queue.

Conjecture: most of the people with $15K home rigs and the know-how to do this also have a lot of earning power and professional status to lose by publicizing something like this. I'd bet it's been done hundreds or even thousands of times. Just not published.

LOL you started by asking me if I know what BFT even is and now you're going on a rant about totally unrelated things.

BFT is a mathematical property about models of distributed computer systems. A property I'm now willing to bet you don't understand. and which you are now also invoking, vaguely, as part of a sweeping social critiques. This is such a silly pitch. I genuinely can't believe people fall for it.

Why do you drag no less than half a dozen vaguely related hot-button emotional topics with your sales pitch? The only thing they have in common is the constituency you're trying to exploit. People reading this for whom the parent post resonated: you're the fucking mark.

What's the point of this rhetoric? What are you trying to sell? To whom?

The solution to the Social Media Problem is really very simple:

Step 1: log off.

Step 2: Lobby for better privacy laws.

Step 3: Avoid repeating the original mistake; the crypto bros make the tech bros look like saints.

The problem with social media not primarily that it's centralized, so the fact that anything other than "log off and pass privacy laws" is pitched as a solution to FB/Instagram is probably the most astounding "Emporer Has No Clothes" situation I've ever seen.

LOL. I've been working on distributed systems for... a while. I've even had a few conversations with Lamport, pre-Turing award.

Apparently today BFT is a buzzword used by alt-suits to hawk unnecessary solutions to fake problems.

> With redundancies built in, so your data is guaranteed to be online regardless of any third party, and so on?

1. That's... not what byzantine fault tolerance gives you.

2. Fault tolerance was a huge problem in the 90s and 00s but today is in most practical cases a solved problem (and without any use of "crypto").

> For example, everything I've built and given away over the last 10 years has been building towards an alternative to Big Tech: https://intercoin.org/overview.pdf

Lots of nudity AFAICT.