Stop emailing my utexas address
A month ago, UT Austin changed its email policies—banning auto-forwarding from university accounts to Gmail accounts, apparently as a way to force the faculty and other employees to separate their work email from their personal email, and thereby comply with various government regulations. Ever since that change, the email part of my life has been a total, unmitigated disaster. I’ve missed (or been late to see) dozens of important work emails, with the only silver lining being that that’s arguably UT’s problem more than it is mine!
And yes, I’ve already gone to technical support; the only answer I’ve gotten is that (in so many words) there is no answer. Other UT faculty are somehow able to deal with this because they are them; I am unable to deal with it because I am me. As a mere PhD in computer science, I’m utterly unqualified to set up a technical fix for this sort of problem.
So the bottom line is: from now on, if you want me to see an email, send it to scott@scottaaronson.com. Really. If you try sending it to aaronson@cs.utexas.edu, it will land in a separate inbox that I can access only with great inconvenience. And if, God forbid, you try sending it to aaronson@utexas.edu, the email will bounce and I’ll never see it at all. Indeed, a central purpose of this post is just to have a place to point the people who contact me every day, shocked that their emails to me bounced.
This whole episode has given me immense sympathy for Hillary Clinton, and for the factors that led her to set up clintonemail.com from her house. It’s not merely that her private email server was a laughably trivial reason to end the United States’ 240-year run of democratic government. Rather it’s that, even on the narrow question of emails, I now feel certain that Hillary was 100% right. Bureaucracy that impedes communication is a cancer on human civilization.
Update: Thanks so much to commenter Avraham and to my colleague Etienne Vouga, who quickly gave me the crucial information that tech support would not, and thereby let me solve this problem. I can once again easily read emails sent to aaronson@cs.utexas.edu … well, at least for now! I’m now checking about aaronson@utexas.edu. Again, though, scott@scottaaronson.com to be safe.
Follow
Comment #1 February 23rd, 2021 at 3:36 pm
There is a danger that people will email both just to be safe…
Comment #2 February 23rd, 2021 at 3:51 pm
tas #1: Yeah, I’ve already noticed that, even if I tell people “DO NOT USE MY UTEXAS ADDRESS—I CAN NO LONGER READ IT,” they simply don’t believe me and keep using it anyway. (To be fair, I still have trouble believing it.)
For that matter, I have people who still write to my MIT address and email me about joining my group at MIT…
Comment #3 February 23rd, 2021 at 4:11 pm
On the off chance that you haven’t tried this – there’s two different ways to get messages from another account into Gmail. One is forwarding, which requires the cooperation of the other account. The other is importing, which only requires that they enable POP3 access to emails. You can try the latter with Settings -> “Check mail from other accounts:” in Gmail.
Comment #4 February 23rd, 2021 at 5:09 pm
Hello Scott, I understand your frustration and you have my sympathy. I encourage you to explore alternative ways of managing and hosting your mailbox rather than Google. This would not solve your problem that, as you explain, is due to UT Austin’s policies. It will, however, make you and the people around you less dependent on a Parallel Government whose business model and way to enforce power is based on the unethical manipulation of people’s data and siphoning of personal information through any possible means. I know most people don’t care.
Here is some read to better highlight the issue: https://www.eff.org/deeplinks/2020/03/google-says-it-doesnt-sell-your-data-heres-how-company-shares-monetizes-and
Thank you
Comment #5 February 23rd, 2021 at 5:33 pm
Avraham #3: Inspired by your comment, I guess I’ll try that one more time when I get home. But here is my understanding of the situation:
cs.utexas.edu is not, itself, a POP/IMAP server. So in order to do what you propose, I would first need to redirect from cs.utexas.edu to some POP/IMAP server with a utexas domain, and then set up my gmail to pull from that server.
Alas, when we tried that, it did not work. And not only did it not work, it completely destroyed my ability to see anything sent to aaronson@cs.utexas.edu. So I asked them to reverse whatever they did in order to limit the damage.
The tech support people keep telling me that aaronson@utexas.edu does not exist, and in any case, finding a different utexas server to forward my cs.utexas.edu mail to is my problem, not theirs. Meanwhile, I keep telling them that aaronson@utexas.edu worked just fine until the catastrophic policy change of a month ago. So we’re at an unresolvable impasse.
At some point, though, there’s a higher principle involved. If my university would do this to me — and moreover, in the middle of a pandemic, when there’s no ability for anyone to come over to my office and solve the problem in person — then I never again wish to place myself at the mercy of my university’s IT infrastructure. Whenever possible, I’ll simply work around it, as though it never existed.
Comment #6 February 23rd, 2021 at 5:33 pm
There are a couple of things you can do:
1. You can turn off forwarding of your cs.utexas account completely, and instead fetch your cs.utexas email from gmail using POP3. The only downside here is that there is a delay in receiving the email since gmail only seems to poll the cs.utexas server once every 30 minutes or so. You configure this in Settings->Accounts->Check mail from other accounts.
2. You can set up a UTMail account and forward your CS email to UTMail, and then forward UTMail to gmail (Settings->Forwarding and POP/IMAP->Forwarding). You’re not allowed to forward directly from your cs.utexas account, but that doesn’t stop you from forwarding from UTMail.
I’m sure the above violates some number of department/university/state rules—I haven’t checked closely. But FYI.
Comment #7 February 23rd, 2021 at 6:03 pm
Can’t tell exactly how sarcastic you were being about her emails, but what you say I think is literally true and an under-reported aspect of that entire wad of bullshit. [1] She asked the NSA for better communication tools, and was told no. Apparently (but I think this lacks documentary evidence), she was actually given advice by “past Secretaries” to set up something herself. And then she did so. Sorry for jumping on this, but given the NSA repeated failures at securing our communications infrastructures at all levels of society (and indeed, taking specific actions to undermine their security, e.g. undermining crypto) while spending billions upon billions on domestic spying, the lack of reporting here grinds my gears.
As other suggested, if your uni allows pop/imap, you SHOULD be able to configure that to pull from GMail’s side. If you can’t, you might be tempted to setup an email syncing tool like offlineimap on another server to pull from your texas account and push to gmail. Having done this before myself, I do not recommend. It is extremely fickle and slow.
[1] https://www.cbsnews.com/news/emails-show-nsa-rejected-hillary-clinton-request-for-secure-smartphone/
Comment #8 February 23rd, 2021 at 6:38 pm
I am seldomly a person of schadenfreude, but why is it that every time you talked about your personal disaster, it always gave me so much pleasure at my first reaction?
Probably, it is related that your way of vivid pictorial and eloquent complaint in strong contrast with your helplessness.
You are such a lovely piece of art.
Comment #9 February 23rd, 2021 at 6:45 pm
Meow #7:
I am seldomly a person of schadenfreude, but why is it that every time you talked about your personal disaster, it always gave me so much pleasure at my first reaction?
Because you’re an asshole? 🙂
Comment #10 February 23rd, 2021 at 6:59 pm
@ Scott
you mean little bitch, as a professor, how could you do something so rude to a lady, apologise immediately on your blog!
I have never been called so in my 25 years of life, how dare you?
shame on you
Comment #11 February 23rd, 2021 at 8:40 pm
Hmmm… my (old) institution sent me an email a few weeks ago saying they were going to disable all such auto-email forwarding, which I also relied on. That was supposed to happen a week ago now, but email forwarding still seems to work. I’m not feeling inclined to tell them they screwed up!
Comment #12 February 24th, 2021 at 1:54 am
A good illustration of the amount of inefficiency and frustration in our lives caused by mostly useless regulation and the imagined fear of litigation.
Comment #13 February 24th, 2021 at 2:22 am
Does anyone know which Government regulation is it that the Iat folks are afraid of if they allow forwarding?
I mean how is forwarding any different from say accessing both work and personal emails from a pop or IMAP client locally?
Never made much sense to me!
Comment #14 February 24th, 2021 at 7:53 am
Your university has an IT department? My MO has always been that the moment I catch some poor grad student using Linux, I designate them as my oracle for tech support 🙂
I think that’s why Tselil Schramm was smart enough not to let me know she is somewhat of a hacker until she ended her postdoc..
Comment #15 February 24th, 2021 at 8:39 am
https://www.zdnet.com/article/fbi-warns-of-email-forwarding-rules-being-abused-in-recent-hacks/
Comment #16 February 24th, 2021 at 8:54 am
Meow #9: In retrospect, I misinterpreted your comment as taking joy in my suffering, when in reality you meant to express some more complicated and harder to describe emotion. My apologies.
Comment #17 February 24th, 2021 at 9:45 am
The reason behind this kind of change in policy may have little to do with government regulation, but rather technical problems and Google’s policies.
One problem is that the DMARC/SPF mechanisms that are supposed to stop spam/phishing by assuring that the sender really is who they say they are generally don’t survive automatic forwarding. So, when organizations start trying to enforce these, automatic forwarding stops working, and IT people have to contend with lots of “why didn’t an important email get forwarded?” complaints.
Google has its own way of analyzing incoming mail messages and deciding if they should be rejected because of a spammy source. Their algorithm is completely opaque, changing all the time, and in seemingly random circumstances it doesn’t like automatically forwarded messages. If you run a mail server that is automatically forwarding to gmail, again you may have a lot of “why didn’t an important email get forwarded?” complaints to deal with because of this.
At least at one place I know about, the argument that “automatic email forwarding is now too unreliable to support” is what led to change in policy.
Comment #18 February 24th, 2021 at 9:45 am
My department got out of the e-mail business entirely about 20 years ago, passing the handling to the university. My fantastic Unix filters were obsolete, and the mandated web interface was some absolutely awful crap. Five years later, they switched to some Microsoft plan, which was even worse. Five years after that, they switched to a Google/Gmail ad-free interface, which is mostly pretty good.
There has been no mandate to use official university accounts. The result is that e-mail from university people is not automatically identifiable. Very annoying.
Comment #19 February 24th, 2021 at 9:56 am
Rahul #12 and Peter #16: The university says that the reason for all this is FERPA, the law to protect the privacy of student records, which they can’t “formally guarantee” that Gmail complies with—even if, in actual reality, university emails discussing students’ grades and so forth go through Google servers too, and are subject to exactly as much privacy risk! A second reason is that, if someone sued the university demanding a professor’s emails (on the argument that it’s a state university and the emails are public records), they want to be able to separate work emails from personal emails automatically.
Alas, none of this is remotely compatible with the reality of my life, which for better and worse, has barely any firewalls between any part of it and any other part, because I lack the mental discipline to maintain such firewalls.
Comment #20 February 24th, 2021 at 1:01 pm
Scott:
Interesting. But both reasons seem dicey.
Eg say the email stays on the University mail server and now I access it from my laptop with IMAP and say Thunderbird would the University be able to ensure FERPA privacy compliance ( whatever that means) of the stuff downloaded to your local machine?
I would bet my local machine is leakier than Google’s infrastructure.
Let’s take the 2nd excuse regarding an FOIA request. So long as you are only forwarding emails the primary email must at some point pass through the university server at which point they already had a chance to keep a pristine backup. And I don’t see how this set of mails would be “polluted” with the personal emails.
All in all, both reasons sound pretty weak arguments to me. Maybe someone else can correct me.
Sounds more like security theatre than security. Maybe it just makes IT feel more safe.
Comment #21 February 24th, 2021 at 4:21 pm
You are not alone: https://www.thebigquestions.com/2015/03/10/those-clinton-emails/
Comment #22 February 24th, 2021 at 5:15 pm
steven e landsburg #20: Hear hear!
Comment #23 February 24th, 2021 at 8:18 pm
If NP complete languages are subset of Indexed languages which are subset of Context sensitive languages which are subset of Turing complete languages which humans are capable of why is P vs NP important?
Comment #24 February 24th, 2021 at 10:12 pm
Ars Technica recap on D-Wave’s recent “QC Breakthrough. Or not?”: https://arstechnica.com/science/2021/02/d-waves-hardware-outperforms-a-classic-computer/
Comment #25 February 25th, 2021 at 8:08 am
Why is P vs NP so important ? #22
P vs NP isn’t about whether problems can be solved but about how long it takes to solve them. It’s true that any NP problem can in principal be solved by some Turing Machine (or by a human with pencil and paper) but it’s not difficult to construct a problem that would take longer than the age of the universe to solve even if it were parallelized across as many 1 GHz computers as there are atoms in the observable universe.
That’s using the fastest currently known algorithm. P vs NP is about whether much more efficient algorithms might exist for such problems, perhaps allowing them to be solved in a few seconds on an ordinary laptop.
It’s basically the difference between whether we are doomed to be monkeys forever or might instead hope to become gods.
Comment #26 February 25th, 2021 at 6:58 pm
I disagree. If we are capable of tm complete languages either there is a faster algorithm and we figure it out or there is no algorithm so who cares? Either way we are capable of nspace(omega (n)) languages.
Comment #27 February 26th, 2021 at 3:04 am
From experience, hosting a personal email server is actually a fair amount of work.
Especially if you’re hosting it on a physical machine rather than on the cloud.
You’d want a static IP, which typically involves reaching out to your ISP for a business account.
Then there’s the DNS setup, plus the mail server itself which typically has additional dependencies.
Plus, it’s incredibly unreliable, a maintenance burden, and actually less secure (in so many ways).
Comment #28 February 26th, 2021 at 6:48 am
Why is P vs NP so important? #25
For those of us who with material bodies of finite lifespan it matters quite a lot whether the time to solve a problem is best measured in seconds or in multiples of the age of the universe.
Comment #29 February 26th, 2021 at 8:08 am
Comment #15: I am sorry too… very very…for taking joy from your suffering and prove your words that I AM an…????
Comment #30 February 26th, 2021 at 9:36 am
In regards to the P != NP topic;
It occurred to me during the past few days how diagonalization and self reference is truly at the heart of so many problems in mathematics and logic. From Gödel’s Incompleteness, to the Halting problem, Russel’s antinomy, “This sentence is false”, no bijection between Natural Numbers and Reals,… You always seem the same patterns of self referential/ diagonalization type proofs involved. There is a book from Raymund Smullyan who formulated the Incompleteness Theorems in their most generalized form with the exact same title “Diagonalization and Self Reference”, I should really start working through it. I would not be surprised if “P != NP” would eventually be added to the list of statements undecidable in ZFC, I am somewhat rooting for it.
Regardless, the statement itself seems phenomenologically important, in the end, Turing Machines were introduced by Turing according to the supposed thinking capabilities of a human mind thus (assuming decidability within ZFC) P != NP XOR P = NP would give us a bit more insight into the mental capacities of our minds.
Comment #31 February 26th, 2021 at 10:28 am
The answers provided are meaningless. As a human species why should we be indulging in P vs NP if we our capability is large?
Comment #32 February 26th, 2021 at 1:02 pm
Why is P vs NP important #30,
I am not quite sure I understand you correctly. Do you ask why we are caring about NP complex problems at all? Because proofing that they could be reduced to P would allow us to replace algorithms with faster algorithms for specific tasks which means real life computing improvements.
On a philosophical note I also care because the problem has implications for some aspects of the human mind.
Both considerations cause me to experience joy, thus I care.
Comment #33 February 26th, 2021 at 3:15 pm
Job #26, yeah, self-hosting email is a pain, but changing the MX of your personal domain to your favorite email provider is easy enough. I’m quite happy with fastmail.com. It’s on the expensive side but the upside is that it means you are the customer rather than the product. I would never use gmail to receive email, and I don’t even like sending to it.
Comment #34 February 26th, 2021 at 3:36 pm
Why should we be indulging in the question of whether P vs NP is important?
Either it is important and we prioritize it, or it’s not and we don’t so who cares?
Comment #35 February 26th, 2021 at 4:18 pm
@Why is P vs NP important?: I for one am interested in constructing thousands of gravity inferometers to determine the relative locations of the nearest googol of detectable bodies and through the best n body formula available, find out where everything is or was for a billion years before and after, along with any events that would have disturbed those bodies (gravitational wave events, hypernovas, etc).
At which point there could really be nothing more to accomplish without seeming frivolous.
Comment #36 February 27th, 2021 at 10:32 am
I’m not at all surprised that a university would do this. Iowa State did this a few years ago when I was teaching there. What I find interesting is that someone thought this was a good idea to do *during COVID* where people might have more trouble getting email access or having tech deal with email problems.
Comment #37 February 27th, 2021 at 6:14 pm
I still do not see the point. If at all it indicates against Scott’s wisdom P is NP and human beings are capable of proving it. It is way below our potential and AI is the correct Computer Science field theorists should endeavor as the potential of AI is way above context sensitivity.
Comment #38 February 27th, 2021 at 7:42 pm
@Why is P vs NP important?:
I wouldn’t go that far. Depending on the circumstance and perspective, if anything can be arbitrarily computed, perhaps in polynomial time, than everything can be arbitrarily manipulated.
The logic of the universe runs on math afterall.
Comment #39 February 27th, 2021 at 8:10 pm
Why is P vs NP important? #37
You seem to be willfully ignoring a fact that all humans intuitively understand practically from birth, that it doesn’t just matter what you can in theory do but also how long it takes you. You can, in theory, walk from New York to San Francisco or dig the Suez Canal with a hand shovel, but we all know that those aren’t practical methods for accomplishing those tasks.
As long as you refuse to accept that fact there just isn’t any basis for further discussion.
Comment #40 February 27th, 2021 at 9:44 pm
Does anyone else suspect some of those latest comments were produced by GPT-3? (/semi-sarc)
Comment #41 February 28th, 2021 at 12:22 am
This is a comment in three parts. (Why so many? Because when a system like this heavily incentivizes doing stuff in bulk, of course people are going to do stuff in bulk.)
A similar email story: I updated my pseudonymous email today to duckmaster0@protonmail.com*. (My old email is now for real-name purposes only.) Signing up for the email wasn’t the annoying part – it only took a few minutes, as advertised. The annoying part was switching my email for all the websites I use (which turns out to be nine in total, including this one, because I am Extremely Online(tm)). Needless to
saywrite, I finished it.*Note: Jeffrey Paul, aka sneak, has a pretty good guide on how to set up a personal email on a custom domain. However, since following this recipe would result in the creation of two ProtonMail accounts as well as one custom one, I thought that it would better if I skipped the trouble and just create a single ProtonMail account.
——
A somewhat less-similar calendar story:
I use MacOS (which I’ve done since ~2017; big fan here). In particular, I use the builtin Calendar app (more recently; less of a big fan), which I set to sync with my Google Calendar (so I can get reminders about stuff like classes). Last Monday (22 February 2021), the Calendar app suddenly stopped working, giving a mysterious error, and refused to sync with my online calendar (the last of which I verified by creating a test event). After spending several hours deleting the cache, logging out and in to my Google account, and restarting (not necessarily in that order, and later mostly following this guide), the error went away, but Mail had to re-download everything à nouveau** and Calendar was blank. Then I went to sleep because it had been several hours. On Tuesday morning everything was fine again and I was happy.
On Wednesday and Friday***, Calendar broke in the exact same manner again, then fixed itself the day after. It’s currently perfectly fine, but I’m expecting it to break any time now.
**Yes, I understand French.
***Not very certain of the exact days now, but I know that this happened twice.
——
Some comments on the \(P \stackrel{?}{=} NP\) discussion**** that has derailed***** this comments section. (Wow, this is an enormous comment.)
Why is P vs NP so important? #22, Gerard #25, Why is P vs NP so important? #26, Gerard #28, Why is P vs NP important? #31, 1Zer0 #32: \(P \stackrel{?}{=} NP\) is important because it is a qualitative question on how long it takes to solve NP-complete problems in the worst case (and, by extension, a question on how long it takes to do other NP-ish tasks like attacking a cryptosystem). Yes, we have many resources by our own standards; however, it is one thing to be able to do \(10^9\) computations and a qualitatively different thing to be able to do \(2^{10^9}\) computations (which is how much compute I expect you to need to solve an adversarially-designed SAT instance with a billion variables, per the Exponential Time Hypothesis, and also far beyond the limits of available computational power set by our current understanding of the laws of physics). In fact, the physical hardness of NP-like problems is the basis for modern cryptography, which protects many people every day.
1Zer0 #30: It is actually known (c.f. Baker and Gill and Solovay 1975) that many diagonalization-type arguments cannot possibly resolve \(P \stackrel{?}{=} NP\) because they would also relativize, and \(P \stackrel{?}{=} NP\) admits contradictory relativizations. However, it’s still unknown whether more subtler uses of diagonalization could work (and, in fact, Williams 2011, in an impressive breakthrough, used diagonalization to show that \(NEXP\not\subset ACC\)). Personally, I’m expecting diagonalization to show up only rarely in a final proof, and only if the truth turns out to be \(P\neq NP\). Also, if \(P = NP\) is Independent of \(PA + \Pi_1\) (which I’m expecting it to be, if it’s independent of ZFC), then we can solve NP-complete problems in about \(n^{\alpha^{-1}(n)}\) time, where \(n\) is the instance size and \(\alpha\) is the inverse Ackermann function (or, in fact, any function on the Wainer hierarchy; c.f. Ben-David and Halevi 1992). This is technically superpolynomial, but slowly-growing enough that we can treat it as basically polynomial-time for practical purposes.
Job #34: Our resources, particularly those allocated for research into great scientific mysteries, are very limited in the grand scheme of things, so it’s important that we prioritize the things that are most likely to impact us in the future.
botnet-client #35: Please tell me this is a joke, because I will be extremely concerned about the future if it isn’t.
Why is P vs NP important? #37: Like you, I’m more skeptical about \(P\neq NP\), which is basically received knowledge in computer-scientific communities. Unlike you, I am more skeptical about \(P = NP\) than the numerous people who claim to have designed polynomial-time SAT solvers******. My reason for the first skepticism is that \(P = NP\) is intuitively plausible (we already have fast heuristic SAT solvers, what if we optimized our heuristics so much that the runtime shrank to polynomial time? we already have polynomial-time algorithms for very similar lookalike P problems, what if we adapted them to work on our hard NP problems?). My reason for the second skepticism is that this is an open problem.
****Initially, I treated P vs. NP as a Big Fucking******* Infohazard(tm) and kept my investigations of it strictly under wraps, because a sufficiently \(P = NP\)-like discovery would destroy all the cryptography in the world, which would endanger the security of an enormous number of people (and similarly, a sufficiently \(P \neq NP\)-like discovery might encourage terrorists to plan attacks in perfect secrecy, which in turn would endanger the security of an enormous number of people*******). However, since my parents pushed really hard for me to get a mentor for this, and people like the OP are apparently not-scared enough about this to write publicly (and very technically!) about it, I finally, reluctantly, decided that it was safe to go public on this topic for now.
*****I’m actually currently working on a project to review each of the 116 claimants listed therein. (If, as I half-expect, it turns out that each of the arguments presented are utterly fallacious and we have not even a hint of how to proceed in either direction then this project is not very important; however, I’m still holding out hope.) Since this project is very amenable to being done polymathematically (as each review can be done independently of the others), I’m hereby extending an invitation to the public: If you want to work with me on this, please email me. (To find my email, scroll upcomment.)
******Although maybe I should view this as a positive thing! (Like it’s as a sign of hope that, even in 2021, dominated by the SARS-CoV-2/COVID-19********* pandemic and the now-extreme unending American culture war, we still have enough hope left to discuss one of the world’s greatest scientific mysteries, rather than, panicked, retreat to our personal bunkers.)
*******I hope (rare, usually mild) profanity is allowed here.
********I think I might have found a political litmus test! Fear of a \(P = NP\)-like situation is left-coded, and fear of a \(P \neq NP\)-like situation is right-coded. (By this test, I’m leftist.)
*********I always use the scientific name of the coronavirus, because I like promoting scientific literacy.
Comment #42 February 28th, 2021 at 9:11 am
I am not understanding why a linear time algorithm has anything to do with the modest task of even NLP through AI. Anything which explains things to me better be capable of explaining at the least Grothendieck’s motives to a toddler (however you may interpret the meaning). If an AI cannot explain geometry to me it is not proper AI formulation of NLP and the goals appear nothing to do whether P=NP and SAT has a linear complexity and runs in logarithmic space simply because to explain NLP it has to capture at the least context sensitivity and I do not think I understand grammar necessary for geometry but it should be at least recognizing these grammars if not producing theorems which are not going to be in Indexed languages in the first place to formulate (the production of constraints are not there in indexed languages and SAT can solve iff the constraints are produced and I DO NOT SEE THE POINT OF P=NP and proofs).
Comment #43 February 28th, 2021 at 1:07 pm
duck_master #41
If you listed the consequences of having a practical algorithm for solving NP complete problems (which would imply P = NP, but the converse is not necessarily true) breaking all cryptography would be one of the least significant, even in the unlikely worst case scenario where that led directly to another World War and 6 billion deaths.
The most important consequence is that it would give you a perfect optimizer capable of finding the global optimum of any effectively computable function.
In the history of the universe there have been two main paradigms for intelligence that we know of: evolution and neural networks. The first gave rise to the second after billions of years but the second seems to be much more powerful than the first. Both of these paradigms are built on some kind of optimization algorithm. Evolution optimizes for quantity and is based on genetic algorithms. Biological neural networks are less well understood in terms of their algorithmic basis but it’s clear that they function through the optimization of individual utility functions.
As far as we can tell both of these paradigms use heuristic optimizers which can at best hope to find only rough approximations to global optima. I believe this is why we find ourselves condemned to this hellish ape world. In short these heuristics are able to create more complexity (ie. problems) than they are able to solve. As time progresses things will only get worse until our species destroys itself, devolves back to the pure animal level or otherwise goes insane.
The only hope I see out of this situation is to find a truly optimal optimization algorithm, which can only exist if P = NP.
Of course such a scenario would have it’s downsides, not the least of which would be that humans would become functionally obsolete. I don’t know how we would deal with living in such a world but I find the alternative of collapse into insanity far more distasteful.
Comment #44 March 1st, 2021 at 3:58 am
My university instituted a similar technological measure recently, where email to my official university account goes to an MS Outlook mail system where I’m supposed to access it using a web interface.
But that’s not going to happen.
So I figured out how to ask GMAIL to scrape the mail off it.
Then the university turned off the interface they use, and restricted non-web access to the server to “internal” IP addresses.
So I figured out how to run a fetchmail process on my desktop-at-work which scrapes the email off the university server and delivers it to where I want it delivered. As far as the university can see, I’m at my desktop using Outlook to read my email. That’s not something they can really forbid.
If this is anything like your situation, I’d be happy to share details, like my .fetchmailrc.
Comment #45 March 1st, 2021 at 2:18 pm
duck_master #41, thank you for posting these papers – the question whether the P=NP conjuncture is independent of ZFC or similar axiomatizations of mathematics occasionally crossed my mind without bothering to look into it deeper – I have to admit my ignorance in regards to existing proof theoretic work on this question but will gladly update my understanding.
Comment #46 March 2nd, 2021 at 8:42 am
There are types of “computers” that don’t seem to be limited like the traditional Von Neumann model.
For example, distributed systems where the computing power can grow on demand.
Cloud computing is a bit like this, but nature has come up with very unique types of problem solving, like that famous slime mold:
https://www.youtube.com/watch?v=HyzT5b0tNtk
and
https://www.sciencedaily.com/releases/2021/02/210223121643.htm
When the computing elements can grow from local resources and spread independently, it’s often possible to “flood” the entire solution space very fast.
Life itself often relies on exponential growth (only limited by resources), e.g. the way an embryo grows (by cellular division) or the way covid 19 spreads.
Comment #47 March 2nd, 2021 at 7:42 pm
Fred #46,
If your model of computing allows you to split into new computers at each stage, and they can communicate with each other instantly then yes, you can do more things, including solving NP-complete problems in polynomial time. But a cloud computing or similar system isn’t going to actually be able to do that, for a simple reason: there are only three spatial dimensions. That means your computer can only grow at a polynomial rate. If you are operating in a classical framework with a fixed number of spatial dimensions, then no matter what tricks you try, you really are only going to get a polynomial speedup, and so your big classes will remain the same. None of this depends on anything about Von Neumann architecture.
Comment #48 March 2nd, 2021 at 9:29 pm
Joshua Zelinski #47 and fred #46
It seems like your model makes assumptions beyond just the number of spatial dimensions, such as the propagation speed of your “computronium sphere” being limited to the speed of light and something about the density of useable states also being bounded. But for anyone interested in practical matters the fact that an exponentially expanding computation would max out the amount of matter in the observable universe for quite modest values of n is probably a sufficient argument against that idea.
Comment #49 March 3rd, 2021 at 2:09 pm
@Joshua ,Gerard
right, I forgot to underline that this type of computation is obviously constrained by many resources, whether it’s physical space, speed of light, or material resources – and that’s true about any physical system that’s trying to accomplish something.
We often need to consider what’s the best way to accomplish something given a fixed set of resources. E.g. the human brain is limited in terms of teraflops (raw computation), but it can still achieve quite a lot from just ~3lbs of matter (compared to a laptop of the same weight). Right now simulating an entire 3lb human brain with classical Von Neumann architecture is quite out of reach. So it’s important to also look at the O(1) constants, not just theoretical asymptotic behavior that may never be achieved using a given set of resources.
It also depends on what we’re trying to achieve. E.g. If you consider the so-called “von neumann probe”, then running out of resources (possibly as fast as possible) can be part of the goal:
“Von Neumann proved that the most effective way of performing large-scale mining operations such as mining an entire moon or asteroid belt would be by self-replicating spacecraft, taking advantage of their exponential growth”.
Similarly it’s still easier to build a biological organism by using an exponential growth process consisting of individual cells splitting in 2 and having local interactions (as opposed to having a central unit creating and assembling the cells one by one).
It’s also why the internet has been so successful (i.e. small machines that can be build fast and then networked together), vs building one giant super-computer that would fulfill all the computational needs of the world.
Comment #50 March 3rd, 2021 at 6:54 pm
Scott:
So who do you think will break RSA first, Shor or Schnorr ?
Comment #51 March 3rd, 2021 at 8:29 pm
Gerard #50: I hate predictions. All I know is that, while I’m not competent to judge the Schnorr paper, people I’ve talked to who are more expert than me are not particularly excited…
Comment #52 March 4th, 2021 at 6:48 am
Gerard: And, I’m told by the experts that an error has now been found in the determinant calculation in Schnorr’s paper.
Comment #53 March 4th, 2021 at 8:08 am
It must be nice to be a determinant calculation expert.
Comment #54 March 4th, 2021 at 9:53 am
Schnorr said via mail that he uploaded an old version (2019 paper?) on eprint. The new revision has even more unlikely results I think, based on what people think on StackExchange and the Twitter thread.
Comment #55 March 4th, 2021 at 11:57 am
@Comment#50 Scott likes snore.
Comment #56 March 4th, 2021 at 12:20 pm
Yeah, it doesn’t seem to be holding up too well. Someone has now implemented at least one version of the paper and the results appear to be not as claimed: https://github.com/lducas/SchnorrGate.
Comment #57 March 4th, 2021 at 2:11 pm
Scott has a very good reason. Unless P=NP these problems cannot be solved. He believes in Quantum computing because precisely it is powerless against non-determinism and he *thinks* to tame the clause structure you need an extra ingredient and looking at the way sieves and schnorr’s svp based approaches work and indeed his own hierarchy theorems work in quantum complexity his opinion in the consensus of worldviews is correct.
Comment #58 March 5th, 2021 at 9:28 am
Ditch gmail anyway. Degoogle your life. Google is a monstrous evil multinational corporation that will spy on you and abuse your information. Switch to Proton mail.