1,226 karma · joined March 23, 2014
An elementary resource that goes through basic steps for a computer scientist (non expert in QFT) would be a great. A simpler particle than electron is also ok, but I'd love to understand how you mess with that equation.
In the context of the OP paper, approximately solving the t-bounded Kolmogorov complexity (in a precise technical sense described in that paper) is akin to breaking one-way functions.
A method to breaking one-way functions would in fact break all of the cryptographic schemes (enc, signatures, prgs, hashing, zero-knowledge, mpc, bitcoin...) that rely on computational assumptions that we know. There is then no hope for doing things like we do on the internet today.
A secondary interpretation relates to breaking a specific widespread cryptosystem like ECDSA or Ed25519 (which can both be broken with suitably large generic circuit quantum computers). In this context, maybe some important things break, but in principle, we can rebuild them using lattice-based schemes or something else.
If you spend the time understanding just the statement of their result in that paper, you get to an eerily small gap between "heuristic algorithms" for a version of the time-bounded K-complexity, and "errorless" versions of it.
Their line of research has been even more inspiring than this very well-written quanta article suggests.
But here are some observations to help interpret the interesting story:
1. When you go for tenure, your tenure committee solicits letters from 8-10 external reviewers who you have not written papers with, have grants with, etc; i.e., "arms-length" letters from experts in your field. These letter writers generally need to be "full professors" (i.e., tenured, which means "Associate Professor", and then promoted post-tenure to "Full Professor"). These letters are a crucial aspect of a tenure case.
For tenure at a place like MIT, these letters basically need to say that "you are the top person in your field" and who "is driving a significant line of work in the field..." etc.
Even if you have "2 books, 13k citations, and a few good papers at top venues" as the OP author does, this is a very high bar to clear. If there is another researcher at a less prestigious place who is doing as well or better, this is going to be a structural problem for your tenure at a top-5 place.
There are several people who I know in CS who have amazing groundbreaking results in CS who did not get tenure at MIT, and I've guessed this to be the reason. They go on to great amazing places and continue their work (like this author has seemed to do).
2. Above is the process at the top 5 CS schools; it is different at schools in the 5-30 ranked places. I have never seen the type of politics this author talks about. For the most part, pre-tenure professors (called "Assistant Professors") get annual feedback on how they are doing. They get a 3-year review which is a "mini-tenure" case to pinpoint any issues with their record so they have a chance to improve it. And when they prepare their cases, they get a lot of help from mentors in the department. Tenure committees are supportive and try to present the strongest case possible.
I am sure there are political places, in which this "who benefits from your tenure" argument makes sense. Maybe the media lab is one such place.
My point is that academia is still, for the most part, a model of a great working environment which allows people to thrive and reach their potential.
3. It is a huge privilege to hold an academic position. We produce more research scientists than there are academic spots in the US. I can think of top researchers in the US who could have easily cleared the MIT tenure bar, but who are at "lower ranked" places for various reasons. This author is still doing academic work, and thus has a very fulfilling career in which they drive their own agenda.
I for one feel extremely lucky to be an academic; I know there are other people in this world who are more capable than me and would have done better than me had they been given the same resources and opportunities.
Many IRBs are unaware that these kinds of "public surveys" unduly burden respondents and cause them unnecessary stress.
The little benefit will be a series of graphs indicating how site operators respond, which could be interesting, but does not justify the burden.
Share buybacks technically accomplish the same; the proportional increase in shareholder value should be the same. However, investors who do not want to recognize income that year do not have to; whereas investors who need the dividend income can sell a small number of appreciated shares.
They seem to learn a lot by centralizing a lot of the internet's traffic through their core. Perhaps it makes their ddos-protection more robust when they can train on "what normal usage" looks like "for the longer tail".
Perhaps they are building moats around that business and expecting a future when Fortune1000 requires their cdn.
The main question: how can Cloudflare make this into a sustainable business?
* cost/gb is cheaper or same as s3, gcp, azure
* no egress charges to customers, but they still have to pay for transit when they cross an AS!
what is the hidden angle Cloudfare is using here?
It is impossible for me to explain how/why it works so well.
I've tried it in person and it was truly amazing. They used some very fancy tech when I saw the demo, so I'm thrilled it is finally being announced and possibly shared with a larger audience.
Explanation of why it is amazing: It totally fools your perception. No glasses or goggles--but rather an 8k display with special glass that allows your different eyes to see different pixels (a light field).
They also optimize the sound, and the rest, so as all the testimonials point out, you actually feel like the person is in front of you.
It also works for the "cube" around them, so if they hold up some object, it also feels like that is in front of you.
Amazing...
Emily Barnet, 2020, Yale (2015)
Diana Li Kim, 2020, Yale (2017)
Arjun Ramamurti, 2020, Yale (2018)
Daniel Richardson, 2020, Virginia (2018)
Brittany Jones-Record, 2020, Stanford (2016)
David Scott Louk, 2020, Yale (2015)
Elizabeth B. Deutsch, 2021, Yale (2016)
Joel F. Wacks, 2021, Chicago (2018)
[1] https://en.wikipedia.org/wiki/List_of_law_clerks_of_the_Supr...
"Google’s limited copying of the API is a transformative use. Google copied only what was needed to allow programmers to work in a different computing environment without discarding a portion of a familiar programming language. Google’s purpose was to create a different task-related system for a different computing environment (smartphones) and to create a platform—the Android platform—that would help achieve and popularize that objective. "
...
"Here the record showed that Google’s new smartphone platform is not a market substitute for Java SE."
...
"Google copied these lines not because of their creativity or beauty but because they would allow programmers to bring their skills to a new smartphone computing environment. "
...
"the Court concludes that Google’s copying of the API to reimplement a user interface, taking only what was needed to allow users to put their accrued talents to work in a new and transformative program, constituted a fair use of that material as a matter of law. "
Does the AI processing happen on the phone or on the cloud (Pixel 3 user) ?
Density of air at sea level: 1225 g/m^3
C02: 0.0383% by volume (383 ppmv) corresponds to 0.0582% by weight.
Ergo, 1m^3 of air has 0.713 g of C02 in it.
Ergo pulling 1 metric ton (10^6 grams) of C02 per day requires processing AT LEAST : 10^6/ .713 = 1.4m m^3 of air per day or 16 m^3 of air per second!
(That would assume 100% capture)
I was skeptical that 1 cooling tower generates this much flow, but the example in [1] suggests 17*10^6 ft^3/minute, or roughly 8000 m^3/sec.
Thus, as long as your capture chemical has 2% efficiency, it seems reasonable.
[1] https://www.power-eng.com/emissions/cooling-tower-heat-trans...
[EDITED after I detected an error in my math]
The amazing observation for me is that evaporative cooling towers process A LOT OF AIR per second.
This is an extremely well written paper. A high-school student with motivation could understand all of it in about 1 week of work.
The key idea here is in Lemma 3.3 which is a simplified presentation of Lehman's original idea. All of the "good" algorithms in this class are exploiting this observation.
Hittmeir adds a new ingredient, namely applying "baby-step-giant-step" searching. And Harvey makes a slightly better parameter choice and strategy than Hittmeir to get his result.
The new idea and value to me would be the concise explanation in Lemma 3.3.
A lot of the feedback is of the form, "who cares, since it doesn't push the state of the art in practical factoring", but for me, there is art in ideas like this. Lehman, Hittmeir and Harvey have put forth a really elegant and beautiful line of thought here---that is accessible!---and I hope you can enjoy it without too much worry about how much it is worth.
VCs like to brag about the "deals they missed" just to remind you that they were in the room where it happened, but then left the room.
Based on the "furnace" picture and language about melting down the batteries, it seems like their approach is to just treat the incoming batteries as "enriched ore" and proceed with an energy-intensive, standard, metal extraction process.
So what they save is the huge amount of energy and dirtiness required to dig in the earth (in the few suitable places on the earth) for rocks with <.5% metal content and crush those rocks into dust, which is substantial, but not revolutionary?
Maybe this OP is also looking for a media deal.
Although you might be smart enough to know that and misdirect.
Even at the best universities, best tech companies, best startups, you will find pools of miserable and unmotivated people.
Don't get stuck.
There are brilliant people everywhere and certainly at faang. Find them, learn, and seize your opportunity to have an interesting life.
For years, my academic niche has tried to break free from the likes of Springer/Elsevier. Here are the bottlenecks:
* There are wonderful "pre-print" servers like arxiv and eprint.iacr.org. However, these do not maintain the "archival quality" document storage that is needed for academic scientific literature. In day-to-day, all researchers use these to stay informed on recent results. But how to guarantee that nobody hacks in and figures out how to change a few bytes in one paper that is 10years old? How to guarantee that these documents are available 75 years from now? I'm sure that many of you can devise solutions to this, but they will be costly, and they will need constant labor to implement. How do you pay for this? It is OK when 20,000 researchers in a field are downloading papers every once in a while, but what happens when every student in the world wants to read these? The bandwidth charge becomes non-trivial. It seems like it needs to be outsourced, and some commercial entity with experience handles it.
* The tenure process is slow to change. Many academics need publications in prestigious journals with "high impact factors" in order to get tenure because the upper-level tenure committees in older institutions use these metrics to evaluate cases. These people are not stupid: it is just hard to evaluate cases across a university when you are not an expert. Instead, you assume that certain journals represent "the highest quality work" and thus use the presence of those publications to judge researchers. This means that the top papers still end up in Elsevier/Springer journals.
When I was a grad student at MIT, it was easy to read papers; if your IP was from MIT, every paper was 1 click away. I wonder how it is going to work now that Elsevier's catalog won't work this way...
news.google.com doesn't seem to show ads.
* If it is "user search on foo" --show ads on results page-> "user clicks on ad and also goes to nyt article on foo"
that seems fair.
(2) If Mike Hamburg is behind it (no author information is available on the page), then it has high credibility.
I would take objection to the following statement: "However, modern elliptic curve implementations with fast, simple formulas don't provide a prime-order group."
The Broker-Stevanhagen (and prior) methods can find good curves with prime order, and complete formulas for these curves that are reasonably efficient are given by Renes, Costello, and Batina in https://eprint.iacr.org/2015/1060
So this performance knife-fight in the ellpitic curve streets may require some benchmarks to resolve.
The institutional investors seem to recognize the ponzi nature of this; first money in, first money out at several X. The veil of "crypto-economics" around this gives them plausible deniability in engaging in this wealth transfer mechanism.
In my own circumstance and what I know from my colleagues, CS researchers rarely write their best ideas in grants---not because they are afraid the ideas are too bold---but rather because those ideas are often not fully worked out, and nobody wants to just give those away to a review panel full of top-rank scientists who might make connections faster than you!
The problem with "blind review" is that project proposals can rarely be anonymous because just explaining the work and citing the relevant prior work leaks a lot about the possible author. So making review blind can often be an advantage to the higher profile researcher and a disadvantage to the capable but slightly less well known one.
That said, as my own experiment, during my next NSF review, I am going to tear away the first pages of all the proposals, and make my first pass without knowing the authors to see if it makes a difference.
(a) Spotify does not hold copyright on the open-source modifications that are posted in that repo. Those instructions are work-product of linuxctl and he holds copyright on them.
(b) The posted instructions are not a copyright protection circumvention mechanism. Spotify does not need to provide an ad-supported freemium plan; they choose to do so to prime their funnel. Removing the ads does not circumvent any copyright prevention mechanism.
* How Amazon is evil: They lure people into buying from them, soon you will only be able to buy from them and the damage will be felt; they create addicting home surveillance devices that pretend to be helpful ?
* How Apple is evil: They engineer a basic necessity to break down every 18 months, they introduce progressively more useless "features" into their new products, they rapidly change standards in order to force people to buy stupid connectors, they no longer create amazing products? [Really is this evil?]
* Exactly HOW is Google doing evil now? The only argument I can see is that they collect a lot of information about you which they plan to use later. They still provide a quintessential information service. You can avoid their ads very easily.
(1) The conference hall right next to the FOCS conference room was hosting the international lingerie trade conference ! Still, the talks were more interesting and well attended.
(2) I, like many others, didnt gamble, but I watched my office mate from MIT (now a brilliant scientist who recently won the Godel Prize) lose his 40$ (a big amount for grad students back then) at the blackjack table in literally 3 hands (35 seconds).
He said, "I cant leave vegas without at least trying" and he proceeded to show exactly how LV works.
In contrast, the IOTA team members Sergey and David have no idea what they are talking about. The IOTA constructions were broken, and instead of understanding that, they made bogus points and tried to attack the four DCI researchers in various ways.
In that context, I think Green's tweet represented a bit of frustration at the way public rhetoric is being misused.
[1] http://www.tangleblog.com/wp-content/uploads/2018/02/letters...
However, when i run ./bundle/Sekey.App/.../sekey, i keep getting a "Killed 9" message.
When i run the unsigned version, the binary at least runs (shows the -h messages). Any hints on how to fix?