Oddbean

The value of limiting chatter with others - Tim May https://groups.google.com/g/alt.cypherpunks/c/8fU395P7Orw/m/UCtlFOsoUfcJ (Note: A long message, especially in this time of short, repartee, chattering. I composed this here in my newsreader, not bothering with using an outliner or even with running the text through any kind of spell checker. Some things just need to be dashed-off, without too much introspection and reworking.) Recently I've gotten a few e-mails, anonymously sent, either urging me to return to the CP list or thanking me for staying away and not giving my "racist" opinions. I sometimes look at the archived messages. I'm glad I never moved over to the "Al Qaeda" list when the last of the earlier nodes went dark. What I see is a very small base of posters, many of the same Eurotrash lefties who had dominated the list in the last months of 2003. And mostly just chatter: news snippets on minor security issues, the 2004 election, etc. In other words, a very low-volume list with several opinionated lefties and geeks. The list was exciting in 1992-94, got more oriented toward startups and business opportunities in 1995-97, and then entered a period of recycling the obvious after that (and after a bunch of formerly active members were off trying to either start companies or save their existing companies in the face of the dot com meltdown). Only one or two significant contributors joined the list in the final few years (one of them is now a grad student at Berkeley, the other I haven't from in a few years). A few of the last to join the list were especially clueless about even the basics. There are several other lists I monitor. The capabilities lists, including cap-talk and the E language lists, and the low-volume BitC list (where they appear to be migrating to an ML-like language base, which could be interesting, given that I've been using Haskell for the past year or so and am positively encouraged about the connection between security/containment issues and monads). I also follow the Lambda the Ultimate blog site on a daily basis. And a few other groups. Mostly, though, it is good to avoid joining one of these "communities." These communities come in several different flavors. The first flavor is the Slashdot flavor: lots of repetitious chatter and repartee. Sort of a CP list with 2000 active posters instead of fewer than a dozen active posters. The "one line repartee" is utterly useless. I usually avoid these lists or sites, but my impression (from when I dip into them) is that the very high volume and repartee-oriented nature almost ensures a "current events" focus, so the lists and sites are really just about people chattering about what they read in the newspaper (of whatever kind or focus). The CP list is mostly like this these last few years, albeit a very low daily volume. The second flavor of community is the "ideological mission" kind, whether formally adopting a mission statement (as with the Extropians or Cryonicists) or just more-or-less adopting similar beliefs (as was mostly the case with the Cypherpunk subscribers and attendees, circa 1992-96). Some of the language and developer lists are like this. The third flavor of community is the "interest group" or "working group," such as various language lists (such as the Haskell list and Haskell Cafe list, which I follow via the archives). These are often high volume lists, mainly made up of academics, researchers, the languages principal developers, etc. Reports of new features, bugs, pending releases, and "how do I do X?" tend to dominate. (Usenet was much like this, especially in the late 1980s. Then the obvious things happened. No need to get into this here. A few newsgroups are still useful, such as comp.language.functional (for me, with my current interests). I still post to a few other groups, but mostly to vent or to attack leftie/PC/innumerate nonsense. Not much effort is needed to make these posts, and they are entertaining to me.) These types of communities change, and there are trajectories (maps) from one type to another. The change in the CP list is a good example. Anyway, I could digress on this for a while, but to what end? The point is this: communities are islands which cause their members to adapt their own interests to the dominant interests, or to change those interests through argumentation. "What have you done for the war effort today?" is the key meme. (Yes, the Usual Suspect will probably write one of his faux-literate "er, um, Tim, um, you _do_ realize, um, that the original was in German?") A classic example from the CP list was the "how many lines of C have you added?" put-down of anyone not "writing code." And yet, of 1000 or so subscribers at one point, and of perhaps 100 who contributed most of the postings, how many were actually writing any code at all? How much of their code ever appeared in products, programs, tools, etc.? Obviously not much, given the current state of things in crypto and financial cryptography. I have resisted the temptation to join the Cap-talk list because I know that the adaptation meme would strongly pull me toward "helping with the effort." This is natural, as these are the people in the community. It's very similar to what happens when people join a company or a sports team: the goals of the organization become one's own goals. (Which is sort of the point of being hired by a company or team!) Sometimes people join a list or a community but don't "support" the goals fully. If they are polite and academic about it, they may fall into the role of the "resident skeptic," whether about the nature of capabilities, or the need for a new language, or the value of anonymous transactions. If they are less polite, they may vocally defend their non-support. Sometimes, as with Detweiler and Vulis and a few others, this is pretty disruptive. (Which sometimes triggers calls for "moderators" to approve or disapprove posts, etc.) And becoming a team player, a supporter, may not be best for a contributor. Instead of going his own way, he spends valuable time defending the approach of the group. Further, the approach already favored by the group constrains certain approaches. (For example, someone with some new ideas about domain naming and pricing may not find the ICANN working groups or lists very hospitable to useful to be in. Or someone with developing ideas about an approach to AI may not find a neural net or machine learning list a good place to be, both because of hostile reception, the "help the war effort" pressures, and the time wasted in reading other opinions and in defenses of opinions. The fact is, "communication" with others is often over-rated. Personally, when I think of the most significant things I ever did, they had little to do with communicating with others. The communication tended to come after the key ideas, even the key experiments, were done. Most casual communication in the hallways of Intel, where I once worked and did some of my best work, were of the "gossip" sort. Lunch, vending machine chatter. Sure, sometimes people wrote on blackboards and had ad hoc seminars in the hallways, but this happened a lot less than might be thought. The Thomas Watson Research Center at IBM, in Yorktown Heights, was famous for building its office and hallways so that whiteboards and ad hoc conference areas were ubiquitous. And yet most of the best work that came out of that place was the work of individuals deep in thought (from my visits to the place, from talking to people, and from what I know of the work habits of some of the key people there). Some types of communication _are_ fruitful. One example of relevance to this list is the story of how Eric Hughes visited my place in 1972, ostensibly to look for a place to rent. He had just returned from several months working at David Chaum's operation in Amsterdam, and I had spent several years working on the implications of strong cryptography (anonymous remailers, Blacknet, data havens, digital cash, etc.). He stayed in a spare bedroom at my place for three days...and ended up doing no apartment or home rental searching. Instead, we talked about logic, number theory, the long-term implications of strong cryptography, the "essence" of why digital money works, and areas likely to become important (this is when I first had the vague glimmerings about category theory, for example, but I didn't pursue it at the time). I had made hundreds of pages of notes about "crypto anarchy," developing some of the consequences of strong crypto. Many of them came to pass...I even made a few bucks investing in some of them. The point is that this communication was very fruitful. It set both Eric and myself on paths which lasted for a bunch of years. He suggested we have a seminar and communicate some of our conclusions to a group of our friends (mostly that we knew through the Hackers Conference and various Bay Area circles). We put together a list of about 30 people, with a little more than 20 showing up for our August/September gathering in Oakland, at Eric's newly-rented house. The morning was spent on tutorials about public key encryption, the basics of RSA, and some of the notions like "bit comittment" (tossing a coin and covering it, but doing it digitally and irrevocably, using crypto protocols), "blinding" (the essence of David Chaum's system), and several other "building blocks" for a future where bits and bytes were more important than bricks and mortar. The afternoon was spent doing a simulation, a kind of war-game, which I called "the crypto anarchy game." Using simulations (paper cards) representing digital money, buyers and sellers set up markets in things like "B-2 bomber blueprints for sale" and "Merck's newest drug release date" information. (If the motivation for this is unclear, much has been written about this stuff. Consult Google on Blacknet, information markets, untraceable assassination markets, etc., or consult books like "Building in Big Brother," Levy's "Crypto," Kelly's "Out of Control," etc. While flawed and unrealistic in obvious ways, this "game" put a picture in the minds of participants about how, with some work on the building blocks, there was really no reason this world could not be made real. (In category theory terms, one would say that the parts of the game were categories (objects with arrows or morphisms), the playing or behavior of the game formed various functors (mappings from one category to another), and that the realization that a real world version (functor, broadly speaking) was a "natural transformation" from the game we had played into the real world. A result of this first physical meeting was the formation of the mailing list and the start of a series of monthly meetings, held essentially continually from October 1992 to around late 2002 (I held a meeting at my house and 50 people crammed in). And at one point there were similar meetings in other cities....I will never forget a 3-site linkup between Mountain View (us), Cambridge (Mass.), and Northern Virgina, with the conference call DES-encrypted...this must have been around mid-1994. Hugh Daniel helped make this happen (as he did with the mailing list itself). Jude Milhon joking referred to us as a bunch of "cypherpunks," and the name stuck. To recap this section, the intense communication between Eric and myself caused a lot of ideas to coalesce, and a course of action to be taken. This "movement," or meme, had a lot of momentum for most of the 90s. It came at a fortuitous time, of course. The controversy about Clipper and key escrow, the issue of NSA surveillance and Echelon, the arrival of PGP (Version 2.0, the first reasonably robust version, had arrived the same month we met at Eric's house, in August/September. Some of the attendees worked with "Mondo 2000," formerly "Reality Hackers," and "Wired" was about to be launched. Kevin Kelly, whom several of us knew from other circles (the Artificial Life Conference, Bay Area events), wanted to do an article on our group. He arranged for a photographer, with Steven Levy (author of "Hackers") to write an article. This appeared in the second issue of "Wired," in early 2003, and generated a large increase in list subscriptions, more press attention, and more attendees at meetings. But all was not well, of course. Work on the core building blocks was slow in coming, many of the algorithms were patented (RSA, blinding, etc.). I did a version of Blacknet in 1993 which actually worked, modulo the fact that no actual payment system existed, except in trades of information. Still, it worked. And got the attention of the security apparatus, which deemed it a threat to the security of the national labs like Los Alamos and Oak Ridge. By the mid-90s, several of our group had formed companies. Sameer Parekh's Community Connection morphed into C2Net, then got funding, then hired a dozen or so list members, then morphed in other ways (and eventually was acquired by Red Hat, as I recall). Lance Cottrell formed his Anonymizer nym company. Several people worked for Phil at PGP, quickly acquired by a large company, then folded, and since reborn again with some of the same people. And Zero Knowledge formed in Canada, hiring several people closely associated with the list. But my purpose here is not to reminisce about the glory days. Life moves on. A few people made some substantial amounts of money on crypto software, many did not, and most who had been active in corporate crypto did not choose to return to the CP list as their main forum for discussion. Which is not too surprising. Meanwhile, the list was increasingly oriented toward gossip and chatter about current events...the events of 9/11, for example, had a predictable effect on themes. Surveillance technology, TEMPEST, RFID, all this kind of stuff gets talked about ad nauseum. Me, I'm glad to be out of the day to day chatter. I am glad to not argue with Eurotrash lefties about capitalism, I am glad to not struggle to correct the misapprehensions of people who haven't even bothered to figure out why blinding works the way it does ("but, um, with powerful enough computers they can figure out who spent it, no?", and most of all I am glad to be able to to research and think about what I think is important without carping along the lines of "so, how much C code have you written this month?" I outlined the rough focus of my efforts in some long messages to the list in mid-2002. Some of the emphasis has shifted. I'll say more when I want to. (Not getting committed to a position is really another way of making the point about avoiding communities.) However, for the curious, I'll mention a few general and a few specific road markers along this "I took the road less taken" path: * Logic. Not the kind of simple Aristotelian or Boolean logic so often dismissed as obvious. Rather, how lattices and posets generate Heyting and other kinds of logics. Modal logic. Constructive logic, and how this relates to financial cryptography. * Belief systems. How reputation works, as a set of beliefs ("Alice is trustworthy," "The Bank of America will not steal my savings account money," "This financial note is worth 98% of its face value."). This is much more than Dempster-Shafer belief propagation (aka Bayesian networks). This is the world of toposes, of constructive logic, of Kripke's "possible worlds" semantics, of category theory. * Which brings me to category theory. Again, I wrote quite a bit about this in mid-2002. Since then, I have of course learned more. I see it as the unifying language to more clearly state problems and assumptions in security theory, cryptography (the logic and security part, not the more pure number theory part, which is just a building block for the bigger picture), and in the nature of money and financial instruments. (For a puzzler, a koan, what is the connection between the Curry-Howard isomorphism and why money "works"?) * Haskell. Eric and I had talked about ML many years ago, as it was being used for "theorem-proving systems" and "provably-correct designs," a la the Viper processor, proofs about some chip systems, etc. We both thought, in 1992, that this would ultimately be important for financial cryptography (a complex system where the smallest of imperfections or "holes" can drain a digital system in catastrophic ways). In the late 90s I thought that OO systems, such as Smalltalk, would be useful, but I came to doubt this. So I looked again at ML-type languages, especially OCaml (which some Cypherpunks have worked on, including Damien G., the young French man who cracked a cipher challenge in the mid-90s). Then I got interested in the role of lazy evaluation (call-by-need, more generally) in belief systems, and turned to Haskell, a purely functional (with side effects limited), lazy, mathematical syntax language which was designed by a group of very bright people to be a platform for work on these ideas. (The role of laziness is debated by some. Non-lazy languages can simulate it, via promises and futures, as in versions of Scheme. The use in belief systems comes through the need for not computing or evaluating everything that could conceivably be evaluated, but only just enough. A kind of belief system version of the frame problem in AI, for which laziness and possible worlds is ideally suited.) * And so on, with excursions into Stone spaces, algebraic topology (mainly via category theory), and "topology via logic," as in Stephen Vick's wonderful book of the same title. Do I expect this stuff to lead to some kind of "Tim's Digital Money Bank"? No. But will it lead to my own better understanding of what capabilities really are, or what containment really is, or why and how people believe certain things based on evidence? Yes, I already see a clearer and more unifying view of the ontology of these things. Marvin Minsky once said that mmost of the work in AI is about various researchers and groups giving new names to the things other people have already found. I think this describes the security (and financial cryptography area, which it functorially related to) area precisely. A lot of security papers are filled with neologisms and vague word problem arguments. The whole security and financial cryptography field is crying out for a rigorous, category-theoretic analysis. Then the theorems already proved, theorems about transitivity (of trust, for instance), about commutativty of diagrams, about pushouts and pullbacks and adjoint functors, and so on, can all be used. This echoes/parallels (different metaphors for what is essentially a natural transformation) a point that Philip Wadler has made about "stealing" the best ideas out of math, logic, and category theory and using them in programming. His paper, "From Frege to Godel" (umlaut as desired, Google for his name and the words in the title), makes this point persuasively. As he puts it, often computer scientists have discovered things in programming languages...and found the mathematicians had gotten there first. Example include Curry and Howard, Hindley and Milner, and Moggi's idea that the abstract idea of a monad from category theory could solve some thorny problems with side effects in programming. Wadler took Moggi's ideas, made them work in the context of Haskell, and now monads are at the core of Haskell. "Theorems for free" is another of his papers, and makes the same kind of point. And this way of looking at things points to why scalar and even vector reputation-based systems are so deeply flawed. There is no "authority" for setting belief values, only sets of propositions. And if propositions are programs, if propositions are types, then the ontology of the world is NOT a top-down, object-oriented hierarchy. Rather, truth is built constructively, by Brouwer's "creating subject." The natural logic of computer programming is NOT Boolean logic, though individual circuits will of course by Boolean (for reasons I won't get into now). Instead, the natural logic of computer programming is topos logic, or "intuitionistic logic" (don't be knee-jerked into thinking this has do with ESP or newage mysticism). We operate in the effective topos. Whether I believe, and to what extent, that Alice will pay me back on a loan I made to her is partly-dependent on what others have said about her "reputation" on such matters (but not conclusively, as they may be part of a confidence game with her). And it is partly based on what I believe about where she lives and what her birth or official name is (but not soley, for obvious reasons). And it is partly based on my beliefs about what I believe _she_ believes. In fact, we need to replace the lattice of statements about "actual facts" with a parallel (those darned natural transformations again!) lattice where "is believed to be the case that" replaces many of the alleged facts. And in this view, the "Aristotelian" fact that "A or Not-A," or "Alice will either pay me back or she won't" is of little interest. Operationally, all I the creating subject, moving through time and through a lattice of events and established facts (proved facts, obtained or measured, tangibly) can say is this: "Yes, Aristotelian logic applies in the sense that a year from now, I will know and the world will presumably all agree on the "fact" that either Alice paid me back when the money is due, next month, or that she didn't. But this is of no import RIGHT NOW, because we live in a constructive world, a world of uncovering new facts. We live in a world where Heyting logic applies, not Boolean logic." (Beliefs about possible futures, even the largely ill-formed "THE future," are the basis for our actions, our plans. And for why we believe with 99.99% degree of belief that a twenty dollar bill will actually be accepted by a merchant for some physical objects. The fact that the bill is hard to duplicate (relatively hard, or with harsh penalties believed by the duplicators to be a risk not worth taking), the fact that most of us go our entire lives with twenty dollar bills always being accepted, and so on, form a set of propositions which we believe to some extent. This is where the "value of money" comes from, not from naively-based "reputation networks" (with ill-formed transitivity relations) or from top-down (Boolean) pronouncements.) Lee Smolin, whose "Three Roads to Quantum Gravity" I commented on at length in a long article to the CP list in mid-2002, makes this point about the likely role of intuitionistic, or topos, logic in the explication of ordinary human and financial dealings. Google on these terms and readers can find my article on this. Combine this with category theory for rigor and for "theorems for free," and with the Curry-Howard isomorphism (that propositions are types, that programs are proofs, that _constructing_ something is how we prove many things), and mix in some of the work on evolutionary epistemology (Popper, Bartley, Hayek), and we have a solid foundation for realizing why and how money works, why and how trust works, and why and how security works. And since Haskell and other functional languages tend to produce "executable specifications," with runnable code (efficiently compilable) looking isomorphic to formal specfications, the prospects for building stable, hole-free security and financial cryptography apps are intriguing. So, this is what I have been doing. And I am happy not to have any "followers," not to have plant stakes or flags in the ground and then defend them against yahoos (Trademark of the Yahoo Corporation) or trolls or just the idly curious who won't even take the time to read Lawvere and Schanuel's "Conceptual Mathematics" or other intros to the basic ideas. I doubt that these interests will ever "align with the other planets" in the way some of my crypto interests aligned with the planets of Clippe, the Web, the interest in online transactions, and the whole debate about privacy and security in the 1990s. That was a time of ferment, of proposed legislation, of vastly increased degrees of freedom. It would be nice if there were some smart people to talk to these ideas about. I've found a few, including some former list subscribers (PH, others.) A pity Eric Hughes is off in Utah, not active on any lists that I can see. The cap-talk and E folks are very bright, and I may end up over there someday....for now I avoid it for the reasons I mentioned, the tendency to have their particular causes siphon creative energy away from my deeper interests. This is what I have been doing, and what I am interested in doing. Comments welcome, but don't count on my replying. --Tim May Corralitos, CA