Video Of The Week: Data Driven NYC Talk

A few weeks ago I did a 40 minute talk with Matt Turck at the Data Driven NYC meetup. Matt's a super sharp guy and talking to him was a lot of fun and we got into a bunch of interesting stuff.

Data Driven NYC #18 – Fred Wilson from Matt Turck on Vimeo.

#VC & Technology

Comments (Archived):

  1. awaldstein

    Data and NYC is a great topic.So much is being done. And note it was just in 2011 that the city had digital signs in the subway telling us when the next train is coming! I posted on the recentness and the possibilities of this a while back (http://awe.sm/aGoqy ).Making the web work for us on the street, this idea of an urban web is a great sector.

    1. bsoist

      Thanks for the link. Great post. A lot of possibilities!

      1. awaldstein

        Thanks.We are literally inventing tomorrow every day. Making the web and data work for us as we walk around in our day is as good as it gets.

        1. bsoist

          Interesting you use the Jetson’s reference. I think about that all the time when I see what’s going on. The technical side of it has never amazed me because I understand how it all works, but the fact that we are living through it still freaks me out a little.

          1. awaldstein

            Right on that it isn’t about the tech or the tools. It’s about us, our behaviors, culture and community.Freaky maybe but lucky us for being on the curl of the wave.

          2. bsoist

            I’ll have to check your blog more often. Seems like every time you share a link it’s something I really wish I hadn’t missed. :)I’m going to watch Fred’s video and then read your post again – see if I come up with any ideas.

          3. awaldstein

            Thanks!Great little note to ease into my Saturday.See you in the comments perhaps.

          4. bsoist

            See you in the comments perhaps.All signs point to yes.

    2. Matt A. Myers

      that’s what it’s all about, connecting people, to eachother, to ideas, to informationsubway system here still needs some improvement, e.g. i got a bit lost yesterday … once people make things simple enough where it’s easy me to use, they’ve reached some kind of milestone..

      1. awaldstein

        4.3+M ride it every day.It ain’t changing much very soon.It, like much of New York, ant colonies and bee hives, is I fear self organizing from the ground up. (Yes channeling a bit of Stephen Johnson here.)Just embrace it!

        1. Matt A. Myers

          dammmnnnn i didn’t realize it serviced that many. that’s great.i feel part of self-organizing could lead to simple innovations to it occurring – which i guess are happening anyway.does the city put any money into incentivizing private-solution improvements?i am thinking of something that could perhaps be crowdfunded – maybe just getting the city to promote the campaign, and if people think it’d be a good improvement, then they can help pay for it for everyone to benefit

        2. bsoist

          and I love the way it just “evolves” – trains just not running sometimes, not stopping at certain stops, etc. šŸ™‚

          1. awaldstein

            Hey–you are right of course!

      2. bsoist

        [Monk’s CafĆ©, the whole gang at a booth]KRAMER: All right, Coney Island. Ok, you can take the B or the F and switch for the N at Broadway Lafayette, or you can go over the bridge to DeKalb and catch the Q to Atlantic Avenue, then switch to the IRT 2, 3, 4 or 5, but don’t get on the G. See that’s very tempting, but you wind up on Smith and 9th street, then you got to get on the R.ELAINE: Couldn’t he just take the D straight to Coney Island?KRAMER: Well, yeah…

        1. Matt A. Myers

          i probably have done this..

  2. takingpitches

    I’ve gone to a number of sessions of this meetup (of course missed your session!). Matt does a great job curating it.I love every time you talk about indeed. Screw 200 million dollar rounds!

  3. Abdallah Al-Hakim

    Great session. Your earlier description of why software alone is not a protection of your business is terrific and says a lot about those companies obsessed with patents. Looking forward to learning more about USV investment in big data – Health care field

  4. Guillermo Ramos Venturatis.com

    The secret sauce is about monetizing data. Google does it easily because they don’t almost even need data as their users tell them what they are looking for in a little box; for all the rest, they have to collect data, guess what their users want and try to monetize it ….

    1. Matt A. Myers

      most people try to monetize data in the wrong waymost data can’t actually be monetized, not in the traditional sense

      1. pointsnfigures

        Network effects allow you to monetize data more efficiently than just selling it.

        1. Matt A. Myers

          network effects how? can you offer an example?

          1. pointsnfigures

            I think Fred showed with many of their portfolio companies. If you simply have data, someone can come up from behind you and take you out. Bloomberg is an example. It’s a desktop machine with a bunch of data. They sell information. It’s the inside chat and messaging that give it network effects to make it sticky and put up a barrier to entry.

          2. Matt A. Myers

            in this case it sounds like you’re not actually monetizing the value of the data you have, you’re using the value of the network effects holding people there, and then just seling to the peoplevalue of data is separate from monetizing it

      2. Guillermo Ramos Venturatis.com

        Agree. There are two big questions: how you monetize data and which amount of the data you own has any value

  5. Richard

    Anyone know the history of the “cookie and tagging”?

    1. ShanaC

      it is a feature from web 1.0 designed initially to remember shopping cart data.

  6. pointsnfigures

    Agree with you on this. Health care is big for all the reasons people are afraid of it. Big players, high regulation. Network effects can cause the tail to wag the dog in health care. Think of it this way: In the old days, you’d go to a doc with a problem. You wouldn’t know anything walking into the office. Now, you can prediagnose online before you even get there. You tell the doc and then they either confirm your prediagnosis, or come up with a new diagnosis. Of course, they still run tests to cover their ass.Chicago has a massive amount of data science talent. It’s why the Obama tech team located here.

    1. awaldstein

      You prediagnose online then confirm with your doctor??Three URLs of health centers where this is standard practice please.I’m living in the dark ages downtown it seems.Interested…

      1. Cam MacRae

        My doctor hates it. People anchor to the pre-diagnosis, and get shitty when he refuses to prescribe antibiotics for a cold.On the other hand, physicians as a group have notoriously bad calibration, discrimination, and consistency, yet maintain all kinds of positive illusions. Expanding the frame with a little pre-diagnosis might not be all bad.

      2. pointsnfigures

        My very close friend was feeling like crap. Couldn’t get better. Went online and plugged his symptoms in at webmd.com. Then he went to a couple of other sites. He concluded he had one of three things-and one of the things he could have had was pancreatic cancer.He went to the doc, and the doc made the same conclusion. They ran some tests and he was diagnosed with Stage 4 pancreatic cancer and given 3-6 months to live. Inoperable, spread to his liver.He is still alive today and down at the Illini game with his freshman daughter. However, he has gone online to find treatments other than the chemo and radiation they are giving him.He has found some stuff, and started communicating to see if he is qualified to try it. He’s pretty motivated!Twenty years from now, a lot of our health care will be demand driven by consumers. Already, I am seeing synergistic intersections of big data, devices, sensors, coming together.It’s yet another reason Obamacare is stupid. The rapid technological change, the decreasing costs of things like DNA research, the efficiency of remote labs to do research are changing the medical landscape.One of these days, someone will invent a drug or device, and give it away for free-and monetize the network and data that comes out of it. Just like Google.

        1. LE

          You are describing an outlier case for sure. (Sorry for your close friend of course).That said I’m not sure what your anecdote proves. Medicine involves judgement just like business does. Thinking that a layman with no experience and a narrow window can do the work of a physician with training I simply don’t agree with.It’s the same as trying to decide that you can read and decide if Amazon AWS is better than your own server colocated and/or rackspace and/or some VPS provider. Ask 10 nerds get 10 different opinions. Go ahead and read all you want. Talking to someone who, most importantly, asks questions and uses judgement based on experience over time can’t be easily replace by online information.

        2. awaldstein

          Any one who doesn’t take control of their own health, wealth, and life is simply foolish.The information is there. There are experts who are really open to having you participate.I bring info to my doctors all the time. I choose them because they are the best, they are my partner in my health and I have constant access.There is no need to wait.

        3. ShanaC

          I like this guy for what he did.

      3. LE

        “You prediagnose online then confirm with your doctor??”Don’t do that. I strongly disagree.

        1. awaldstein

          No one said I was even considering it. I was questioning it.

          1. LE

            I parsed what you said differently then.It was the “living in the dark ages” combined with the “Interested” that made me think you were considering it.Now if you said “wow lansdale pa sounds great” followed by “I’m living in the dark ages in NYC” and “Interested” I would have thought differently. I guess the fact that you focus on health and healthy living is what threw me.

          2. awaldstein

            No problemo.I’m a huge advocate of owning my life and health.Part of that is info, part of that is having the best people to guide and support me.Do I bring in things especially vitamins that I want my doctor to look for in a blood test that he doesn’t know about–absolutely. He’s glad to do it.

          3. pointsnfigures

            @domainregistry:disqus, brings up the topic. “Medicine involves judgement”—>there’s your opportunity. Predictive algos that constantly scrape electronic data from devices, sensors and records can yield better judgements and save lives. I am seeing this happen daily in different businesses.

          4. LE

            Re: Your breath test example.Most men can determine when their wife is having her period prior to the event based on behavior. I happen to have an especially good nose. So I can tell with really good accuracy a few days before by her breath.I can also tell a few days before when I am fighting something in my body by my breathing patterns when I exercise.I’m also sure you know about those dogs that can sniff out cancer.It makes intuitive sense that if you could monitor people more accurately you could developer more markers for some future event.

        2. ShanaC

          depends.One of the things I’ve learned is you have to be your own advocate and to know when to question your doctor.I’m actually leaving a doctor at a well respected institution over this.

    2. Matt A. Myers

      most doctors, when they leave the room for a minute, are checking the same available sources we all have – just to remind themselves to see if something else they notice or pick up from what you’ve said – and then they’ll direct you to testing if is warrantedi am sure this has reduced the burden on healthcare practitioners, though unsure if that means the systems cost any less – though having people making important decisions, who need to be good listeners, that are less stressed is good – or at least not overwhelmed

      1. LE

        most doctors, when they leave the room for a minute, are checking the same available sources we all have – just to remind themselves to see if something elseI don’t think that is correct. Most doctors deal with the same 95% things every day (figure arbitrary). As such they can rattle off immediately what needs to be done because they go through same stuff all the time. Just like when I deal with someone I can finish their sentences I don’t even need to hear the entire spiel because it’s the same stuff day in and day out.When my wife is on call at night we are usually watching TV. Call comes in (tv paused) Patient gets admitted and the resident gives the intake she replies instantaneously with what the resident should do with the patient and corrects the residents mistakes and/or gives the course of treatment. Back to the TV (I have an ipad mini handy for the interuptions). When we go to bed she takes the calls at night. Woken up in the middle of the night from sleep she does the same thing over and over again each patient different but it’s all the same stuff. Never a need to consult anywhere (so I’ve given it 5% for safety margin). So wrote you can do it being woken up. At the hospital for example a doctor can easily see 25 patients in a shift. They simply wouldn’t have time to look things up. Anymore than a pilot can or needs to. I have a tenant who is a doctor (ENT) who sees 60 patients in a day which is iirc 4 or 5 hours.That said there are cases where what you are saying is correct. I am just taking issue with the “most doctors when they leave the room for a minute” part.

        1. Matt A. Myers

          it more likely depends on what level and severity of what you’re dealing with, and depending on the experience level of the doctor

    3. LE

      I think people think they can pre-diagnose but in actuality the error rate on this is probably pretty high. An example of a little bit of info being dangerous. Not to mention the fact that the anxiety level created by that info online is not beneficial at all. And the old saying “ask 10 doctors get 10 opinions” comes into play as well. Given the same info there can be different conclusions or courses of action.I had a script to get the shingles vaccine. I didn’t get it filled. When I went into the pharmacy finally they said “oh you only need that above a certain age”. When I dug a bit deeper I found out the reason was there was a shortage of vaccine so they changed the cutoff (forget whether that is actually correct I’m using it to prove a point I actually haven’t fully vetted it yet). In that case I don’t care I want the vaccine I don’t care about the shortage. I only care about the risks of taking the vaccine vs. the benefits. Of course, they still run tests to cover their ass.You say that as if there is something wrong with doing a test in order to make sure you are correct. Personally I’d rather have the doctor do an extra non invasive test if he feels it is necessary. And from my recollection they did this back way before personal injury became a cottage industry.Guess what? When I handle a domain sale I go to the extra effort of checking “back history” on a domain to make sure it’s not stolen. It’s not something that I even tell the buyer. And I have zero chance of getting sued over that type of thing as well. I just do it because it’s the right thing to do the same way when you craft any product you put the effort in because your name is on it. When we do hosting for someone we do backups even though we say we don’t do backups and the customer is responsible. Because people do foolish things and you have to protect them from their folly. (Oh yeah, and to cover our ass!)

      1. pointsnfigures

        I know of a Doc that can use a breath test to diagnose whether you will have a heart attack in the hospital. Everything has been tested, peer reviewed. Problem: Algo for diagnosis is on paper and it’s impossible to get good data out of the software the hospital uses to use the algo.There is going to be a lot of innovation in this space and big giants are going to get toppled.

        1. LE

          That’s interesting. (Any links I just googled and can’t find anything on that?).Who do you see as getting toppled with that innovation though?Would imagine a machine would have to be for the home usage so someone could monitor similar to blood pressure.

        2. ShanaC

          do you have a link, I just want to see this

    4. ShanaC

      HIPPA is a nontrivial problem since it actually is more strict than the way people think of their privacy.And of course Chicago does for similar reasons NY does. Futures and Options markets šŸ˜€

  7. cfrerebeau

    To answer your question of where the data “hacker” or DIY community hang out on the web. At least for the machine learning side of it, many of people hang out at Kaggle. Obviously data is hard to find and Kaggle offer them the opportunity to get access to proprietary set of data and trying to solve real life problem.

    1. fredwilson

      kaggle is interesting. i did not realize it had gotten a lot of traction.

      1. David Semeria

        I would agree, Kaggle is where the party is for data scientists.

  8. Cam MacRae

    Interesting chat.With respect to the hype cycle of data science, big data, etc. I think we’re still marching up the peak of inflated expectations.On the way down we should find comfort in Amara’s Law: “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

    1. andyidsinga

      I recently heard (achem avc) a similar quote attributed to bill gates “we tend to overestimate what we can do in a year and underestimate what we can do in a decade” šŸ™‚

  9. sigmaalgebra

    Four remarks.First Remark — Defensibility”The only defensibility you can get with data isdefensibility around your own data. If you’regetting data from someone else, it’s not defensiblebecause they can give it to someone else as well.”Well, yes, but that throws out the baby and keepsthe bathwater. Why? Because there is an implicitassumption that given the data, everything else isroutine, and that’s badly false both in principleand already in some significant cases in the past.A major case has to be the work of James Simonswhere likely others have all the data he has, butthey don’t make the money he makes, and that’sbecause of what Simons does with the data othersdon’t know how to do.Likely if Simons wrote out what he does with thedata, what he would write would be a lot instochastic differential equations (more popular inRussia, France, and Japan than in US puremathematics) that few others on Wall Street couldeven read and understand. Yes, there’s a book byKaratzas and Shreve, but the prerequisites are aboutsix years of college and graduate school pure math.Similarly for the book by Chung and Williams. Andthere are more books, e.g., by J. Doob on potentialtheory and probability. Also Blumenthal and Getoor.The number of people in the US who know thismaterial is quite small. And there will be morethat is crucial and not in Karatzas, etc.The data itself has no value, only what is done withit has value. A lot that is done with the data isroutine and, thus, not defensible.But, in principle, and in some significant casesalready, what can be done with the data need not beroutine (basically exploiting some mathematics andnot just routine software development) and quitevaluable and defensible. Software based on themathematics can be locked up inside a server farm in’the cloud’ and, thus, better defended.Likely an old example was the fast Fouriertransform: Richard Garwin was sitting next to JohnTukey at a meeting of a presidential advisorycommittee. Supposedly Tukey was taking meeting noteswith one hand and writing Fourier theory derivationswith the other, and Garwin asked if he (Tukey) knewsomething Garwin didn’t. Well, yes, Tukey did.Garwin had Cooley program the work, and the resultwas a famous paper by Cooley and Tukey and then abig industry cutting across large areas ofelectronic engineering, audio, radar, X-raydiffraction, processing of seismic signals, etc. Itbought me a nice Camaro 396 hot rod I used to enjoyseeing shift from 2nd to 3rd at 100 MPH.What Tukey showed Garwin was sufficiently obscurethat, had Tukey just kept the secret, some quitesignificant businesses could have been built withthe embedded fast Fourier transform quitedefensible. Ah, could have been a good way to addto the Princeton endowment!Similarly for the RSA work on encryption.Second Remark — The FutureFor the future in information technology, what willbe more important is not just the data but what isdone with it via some mathematics, possibly originaland, from outside a server farm, just forbiddinglyobscure, next to magic. Mathematics will be moreimportant than computer science or softwareengineering.Want to be good with data in the future? Okay,major in pure and applied math, not computerscience.Third Remark — Search”Google benefits from network effects and the datathat they have, “in their search business, almost aninsurmountable advantage”No on two grounds:First, just from the user side, it does appear thatthe data Bing has is close enough just for Internetsearch such as Google and Bing do.Second, as for the work of Simons and Tukey, there’sa lot more to do in search. E.g., recent threadshere at AVC have discussed the difficulty in findingblogs and video clips. Why? Because the dataGoogle has and what they do with it is poor forblogs and video clips. It’s also essentiallyhopeless for still images, music, long tail Websites, Web cams, etc. So, what Google/Bing aredoing is poor for a large fraction of Internetcontent, searches people want to do, and resultsthey want to find.Is it possible to do better for this poorly servedlarge fraction? Yes, but what is needed is somedifferent data and, then, some good, originalmathematics for saying what to do with the data.Any “insurmountable advantage” here for Google hassmall value.Fourth Remark — Open Source Threat”Three entrepreneurs start a company, and they comeup with a new kind of enterprise software …”Bloomberg Data funds two, young 22 year oldprogrammers, and those two programmers basicallyknock off the enterprise software ….”A bunch of hackers create an open source projectsto create a version of this software”To me that shows the problem of thinking thatsoftware is any kind of defensible advantage”For routine software, yes.For some crucial, core, original applied mathematicswhere the software does the data manipulationsspecified by the mathematics, likely no.So, if the three guys who wrote the first version ofthe software had some crucial, core, original,obscure mathematics providing much of the value andkept that software locked up in a server farm in’the cloud’, then the two guys at Bloomberg or theopen source hackers would have been hitting theirheads against some brick walls.Another way to be defensible: Just exploit somepowerful, relatively advanced, mathematicalprerequisites. Some of the advanced results arebeyond belief; no one without the extensivebackground could even guess that such things couldbe true; they are true, but the proofs are oftenchallenging even if are told the results.Why defensible? Because due to some strong and old’cultural issues’, the number of people with themathematical prerequisites and also interested ininformation technology entrepreneurship is justtiny. Moreover, next to no one in US business iscomfortable hiring and managing such a person doingsuch work. E.g., likely Bloomberg would not be. Nodoubt part of the success of Simons has been that hefinds it easy to hire and manage mathematicians withadvanced backgrounds. Why? Simons is a goodmathematician and, as Chair at Stony Brook has someappropriate ‘management’ experience.Mathematics is a special field: For nearly everyresult of importance in the field, there is a rocksolid proof or counterexample. Some experience as agraduate student or teacher of mathematics shows inclear and stark terms just how difficult it can befor nearly anyone to find and establish a new resultor to find a proof or counterexample for an oldresult. It can be easy to beat a head against awall for weeks for a proof that takes only a fewlines.There is little so obscure, and “defensible”, as theworkings of some software, locked up inside a serverfarm in ‘the cloud’, that is based on some originalmathematics.

    1. LE

      Well, yes, but that throws out the baby and keeps the bathwater. Why? Because there is an implicit assumption that given the data, everything else is routine, and that’s badly false both in principle and already in some significant cases in the past.I think this relates more to data which you get which you can either lose or can be arbitrarily cut off from. For example if I “mine” the USPTO database of trademarks and make money off of that, I have to worry about the USPTO changing the terms and conditions of access in addition to someone else getting and offering the data.But if I buy the data from someone who is the one who “mines” the data from USPTO I also have to worry about that miner changing the terms and conditions on me as well or cutting me off entirely. And of course what the USPTO decides also. And anyone else. [3]It’s more or less a general principle in business that covers a very broad area. I want to own my real estate not rent because I don’t want to have a landlord decide what I can do and when I can do it. I want flexibility. [2] I want a server racked at a colo place. I don’t want a piece of someone elses VPS and I don’t want to buy the server from the colo place either. I want to put in the extra time at the start that will cover me for many years w/o having to worry that someone else did the right thing.The farther you are removed from the “raw materials” the more chance that you are dependent on someone elses arbitrary actions which will prevent you from making money. [1][1] Yes, as with anything it’s always a balance of the risks vs. the benefits.[2] One of the things I decided early on was to own and buy machinery as opposed to using someone else who had machinery and simply collecting a commission or tacking on profit to what they would charge me.[3] You say you have a contract that is airtight? Contract shmontrack. Golden rule. He who has the gold rules.

      1. sigmaalgebra

        Yes.My ‘server farm’ will be at first just amid-tower case at my left knee, maybe an 8 coreAMD processor at 4.0 GHz, 32 GB of main memory,and a few hard disks at 3 TB each withMicrosoft’s Windows Server and SQL Server fromtheir BizSpark program.There’s a lot of room in my house for muchmore. Or clean out a bedroom, get some wirerack shelf units (1.5 x 4 x 7 feet, $100), havean electrician run a 240 V line, put in awindow A/C, get a UPS, put maybe 12 mid-towercases on one shelf unit, and keep going.I can get 15 Mbps upload bandwidth just bymaking a phone call. Half fill the 15 Mbps 24x 7 sending Web pages of 400,000 bits per pageand three ads per page at, say, the averagefigure from Mary Meeker (KPCB) of, as I recall,about $2.25 CPM, and finally have somethinggoing, revenue of,2.25 * 3 * 15 * 10**6 * 3600 * 24 * 30 / ( 2 *1000 * 400,000 ) = 328,050dollars a month. Then 10 months of that andI’ll have funded my own ‘Series A’, just as asolo founder and sole proprietor of aSub-Chapter S or LLC.I’ve just GOT to quit posting at AVC and writethe last few of my Web pages (all simple), andthat’s the last of the significant softwareneeded to go live. More software? Sure, writea little code to extract data from various logfiles. Simple. Then get some initial data, doa few routine things (e.g., tax ID, trademark,domain name), and then DO go live.Fred and others have made the situation justtotally crystal clear: I just need to havesomething good working, and for that I knowthat for the code I just need to write a fewmore Web pages.Some random, exogenous things got in my way,but I’m able to write Web pages again now.I do want more than just something to show: Iwant users and revenue, both significant andgrowing quickly. For now, write the darned Webpages.Today, take the instance of the class I havefor a user’s ‘session state’, an instance I gotfrom the session state store I wrote (worksgreat and on a 3.0 GHz single core should beable to do the session state store work neededto send 1000 Web pages a second) or, via atricky ASP.NET facility, from a previous Webpage for that user (I have that working), takean instance of a class in that session stateclass and ‘serialize’ it to a byte array (havethat working for the session state store work– it’s easy to do), and send that byte arrayover just TCP/IP ‘raw sockets’ and some codehave working for the session state store workto a server (command line program in a consolewindow started with a simple script) for somecomputational code (running, and based on mymath), get back the results, and display themon some really simple Web pages with some dirtsimple HTML tables with some dirt simple CSSand no JavaScript! Back to it.

        1. LE

          get a UPS, put maybe 12 mid-towercases on one shelf unit, and keep going.Here is what you need to do with that. Much more power, easy to do and lasts longer. And fun.Get Interstate batteries big ass industrial batteries 8D-MHD or even larger:http://www.interstatebatter…|8d+phd|3|,Arpp~12,A~Part+Number,N~4294784989-4294784419-4294784418-2147384903-4294784417-2147384906,Nr~AND%28P_unique_id%3aInterstate+Batteries+8D-MHD+8D-MHD%29- Triplite LC1800 (or equivalent) line conditioner/stabilizer- APC 1250 Power inverter- Cables and connectors from any auto supply store.- Transfer switch if you have to take it offline (goes back to line AC) (so an Apc ups hot swap out box)- Have electrician run a line from your basement or garage where the batteries will be to where the servers are. [1]In 1996 I had a bunch of heavy power drawing servers at the office. UPS were expensive (and still are) for the punch they packed. So I hooked up 6 150 lb. industrial batteries in parallel to the inverter and filtered the power through the line conditioner. Worked through 10 years or so of power outages until servers were moved to colocation. Much much cheaper than going with a UPS solution (because they give you sealed batteries). And lasts forever. And you can easily add more batteries just bring one in and cable it to the existing setup.Interstate will deliver the batteries to you (local gas station is just a dealer.) For a fraction of the cost of a commercial system I built this from scratch and it worked. I even built a dolly to cart the batteries around and floor trays for them.I bought one of those power measuring devices from radio shack to determine how long this would last. Conservatively I could power everything for at least 24 hours. Never had a single case where the power interuption exceeded the system it worked great. Built it entirely myself from scratch with no plans or anything. (This was back when there was little on thenet about stuff like this it’s not like today). It was fun as well.Of course if you want to go the sealed route it’s still much cheaper and will last longer than an off the shelf UPS.[1] If you use sealed batteries they cost more and pack less punch (but you can use them anywhere). I had a warehouse so had no issues with the venting etc (and I did check up on the msds of all of that so I wouldn’t have any issues talked to battery engineers etc.)

          1. sigmaalgebra

            Thanks. Great!How to get the battery voltage, at about 12 V,up to the 120 V the server farm boxes want?Now I get it! The inverter also does voltagestepup, likely by using some power transistorsessentially to put the 12 V ‘in series’ severaltimes. Then it does the actual ‘inverting’creating roughly a square wave that hopefully the line conditioner makes into a smooth sine wave.So, will need either some detection of lossof utility power and a switch over to thebackup power or just run the farm off thebackup power and use the utility to chargethe backup batteries.Thanks for the idea. I’ll Google the topicand look into the EE details later — alllooks doable. One of my college summerjobs was designing and building a goodpower supply!I’d want to put the batteries where there is someventilation, but that would be no problem.

          2. LE

            http://www.powerprosinc.com…apc 1250 is an inverter. So obv. it takes 12 to 120v. I added the line conditioner just as a safety measure. Also because when I was straight to utility it had surge protection.The link shown is a much newer model than I used. Looks like the plugs are different the ones I have you simply wire in with a wrench. I had two actually so if I needed to service anything I would use the transfer switch and bring the other one on line. The spare one only had 1 battery (vs. the main having 6 batteries).The reason the solution is good is because it scales so well. You can simply distribute the power and put an unlimited amount of batteries in front of the inverter.I also had this running about 400 feet to the demarc (in the office condo) where the telco put it’s fiber optic equipment. They had this dinky battery on that so in order to keep that up I hired a guy who dragged a cable through the building to the demarc. No permits etc. Just did it. Got the office managers in the other offices to let me go through their units when the boss wasn’t around. [1] If I had asked it would have been totally nixed.[1] Shit not taught in school.

    2. LE

      It bought me a nice Camaro 396 hot rod I used to enjoyseeing shift from 2nd to 3rd at 100 MPH.I shift my car from 2nd to 3rd at about 60mph. But then again it has 7 gears.

      1. sigmaalgebra

        The car has a Turbo 400 automatictransmission and a 2.56:1 rear axle ratio. The rear axle ratio iswhat lets the engine be turningonly 5000 RPM in 2nd gear at 100MPH. The Turbo 400 torque converter,with about 3:1 torque multiplication atat stall, is what lets the car have goodstanding start acceleration.The 2.56:1 rear axle ratio is whatlets the engine turn only about 2560RPM in third gear at 70 MPH andin long distance driving at 90+ MPHget 18+ MPG.I figured all that out after having aFirebird 400 four speed with a 3.23:1rear axle ratio and a 2.52:1 first gear:Then the standing start accelerationwas okay, but the engine RPM at70 MPH was too high and the fueleconomy sucked. The 2.52:1 firstgear was with the ‘wide ratio’ fourspeed; the ‘close ratio’ four speedhad, as I recall, a 2.20:1 first gear –total bummer.I thought that 4 gears were plenty,especially given the broad, flat torque curve of the engine, but Iwanted a first gear at 3+ to 1.The Turbo 400 gives me a standingstart ratio of 7.5:1 or so in the transmission.So, it’s easy to smoke the tires ata standing start.Good thing about the Camaro: It’sa street legal place to put essentiallyany engine could want. Another goodthing: The 6 cylinder version had curb weight 3000- pounds! And anall aluminum small block weighs lessthan the cast iron 6, and an allaluminum big block weighs less thanan all cast iron small block! Can stillget an all aluminum 454 cubic inchbig block crate motor from Chevy!If take a pass by, say, Gale Banksfor two turbo chargers, could have,say, a 3200 pound Camaro with700 or HP. Or, for 3000- pounds,put in an all aluminum Eaton supercharged small block crate motor fromChevy, about 600 HP.It’s still sitting in mygarage. Hmm, when I need somecash for my server farm …!Yes, at times the Turbo 400 wasused in full sized cars with a2.29:1 rear axle ratio! Apparentlythat transmission could be usedto pull a freight train.Yes, once I was made a Full Member of the SAE!

        1. LE

          My brain has always preferred not to focus on numbers but rather the more qualitative factors which my brain derives pleasure from.But there is definitely a segment of the male population that can and do rattle off facts and figures and read popular photography or road and track or car and driver much the same way some people read rolling stone. I never read the playboy articles, just the pictures. I just care about how the music sounds.You said this “The car has a Turbo 400 automatictransmission”A slush! Shifting is fun. Not only that but you can control the vehicle entirely by modulating your right foot. And no those paddle shifters are just not the same experience.

          1. Ryan Frew

            Wait…do you have a seven speed manual?

          2. LE

            Yep:http://www.youtube.com/watc… [1]http://www.youtube.com/watc…[1] Ignore the part about the pdk it’s a slush.

          3. Dave Pinsen

            You missed out skipping the Playboy articles. A college roommate had a subscription. I read a thought-provoking interview with Stephen Hawking in one issue, and some great short fiction (by Updike?) in another.

          4. LE

            You are right.Playboy is really fascinating on several levels to me.It’s this place where women are put at ease ala “we don’t want you to do anything you’re uncomfortable with” and afaik they pulled that off by using women who were already members of the cult to make the new women feel comfortable. Social proof. People nicer to you than your own sisters. (Same way cults operate, right?).And they had/have this great safety net. You rarely hear anything bad about someone who appeared in playboy. They seem (once again afaik) to watch each others back and take care of the girls lest a black mark go on the entire enterprise. To me that’s impressive. Or perhaps they simply vetted well on intake.Locally I was at a wholesaler and there was a girl that worked there. Something about her seemed different. She was attractive in a way that you don’t get in this area. And she was really on the ball, returned phone calls, sent me a hand written thank you I was really impressed. I googled her later and found out that she had appeared in Playboy in the early 90’s a few times.Anyway she ended up getting engaged to a local celebrity here. In every news article they mentioned she was “a playboy playmate” (she wasn’t she just appeared in a few features). The celebrity moved to NYC and they are engaged. But when I called the place the other day assuming that she had moved with him she hadn’t. She was still working locally hard everyday not depending on him.

    3. LE

      Another way to be defensible: Just exploit somepowerful, relatively advanced, mathematicalprerequisites. Some of the advanced results arebeyond belief; no one without the extensivebackground could even guess that such things couldbe true; they are true, but the proofs are oftenchallenging even if are told the results.Why defensible? Because due to some strong and old’cultural issues’, the number of people with themathematical prerequisites and also interested ininformation technology entrepreneurship is justtiny.To highlight again the principle I talked about previously mathematics is “digital” and entrepreneurship (and business) is analog.Consequently the number of people’s who’s brains can operate at a high level at both will be small. Further people who are really stellar in math and sciences at an early age will be pushed and encouraged in that direction and and won’t really be surrounded by the analog things like business and entrepreneurship.

      1. sigmaalgebra

        > Consequently the number of people’s who’s brains can operate at a high level at both will be small.Right, an opportunity.

    4. Dave Pinsen

      For a long time people assumed Steve Cohen’s SAC Capital generated its returns with fancy math. Now Steve’s being sued by the SEC, and has a US Attorney breathing down his neck.The investing landscape is more complex today than in Benjamin Graham’s day, but this quote by Graham (who was offered a faculty position in Columbia University’s mathematics department at age 20) comes to mind:”Mathematics is ordinarily considered as producing precise and dependable results; but in the stock market the more elaborate and abstruse the mathematics the more uncertain and speculative are the conclusions we draw therefrom. In forty-four years of Wall Street experience and study I have never seen dependable calculations made about common stock values, or related investment policies, that went beyond simple arithmetic or the most elementary algebra. Whenever calculus is brought in, or higher algebra, you could take it as a warning signal…”

      1. sigmaalgebra

        Ben Graham was a long time ago, and we can suspectthat more was discovered in the interim.I’m no expert in applied math for the stock market:I have the background to read Karatzas and Shreve(Karatzas is at Columbia — I got the backgroundpartly from a star student of E. Cinlar, long in’financial engineering’ at Princeton and long aleader in probability, e.g., random measures, in theUS), enjoy reading the book when I get time — whichI won’t have more of until my startup has made moreprogress.While via Buffett the Graham track record isterrific, on the potential of mathematics for thestock market I would take the Simons track recordover the Graham remark. I don’t believe it isirrelevant that Simons is a good mathematician.Yes, at various times I’ve touched on differentialgeometry, but I’m a long way from understanding theChern-Simons result.But a simple view of some of the potential (no, Idon’t like puns) is the Black-Scholes option pricingmodel. Until everyone caught on, that was a licenseto print money. The E. Thorpe work was much thesame and earlier.Well, the Black-Scholes model is a ‘first passage’,a ‘stopping time’ (with a cute technicaldefinition), problem for Brownian motion.If take the ‘efficient market hypothesis’ as afirst-cut, then much more complicated portfolioanalyses are again first passage problems forBrownian motion, connected with an observation of S.Kakutani in the 1940s that could use Brownian motionto solve the Dirichlet problem — all of which looksfully natural in the stock market.Then, yes, with Brownian motion we are soon intostochastic integration, e.g., as in Karatzas,Shreve, Doob, Chung, Williams, etc. Indeed, Chungand Williams have the Black-Scholes model in theback as a simple application!I can’t tell you just what Simons did to pay himselfballpark $1 billion a year for some years.Broadly, what can be done with a mathematicalapplication depends greatly on how carefully theapplication is made. Here’s a don’t do: Don’t justtake a formula, equation, or result and apply it andhope. Instead each result needs hypotheses — justexactly as in proofs in high school plane geometry.Exactly. Ignore the hypotheses at one’s peril.At times one reason for using especially ‘advanced’mathematics is that it tends to be more general,that is, to use weaker hypotheses.So for a little diagram, with a ‘transportation’analogy, start at little town Real Question and wantto get across country to little town Real Answer.Well, can use intuitive methods and strike outacross country, through muddy swamps, etc. Or cantake a short ride on Justified Assumption taxiservice to larger town Mathematical Question, catchpassage on Mathematics Airline to town MathematicalAnswer, and take Justified Interpretation taxiservice to little town Real Answer.So, the two taxi trips are explicit and can becarefully considered, and the airline hop makes partof the journey more reliable.For a simple description of how to apply math,that’s the best I’ve been able to do — once taughtit to MBA students.For applying math, take this little analogyseriously; do be careful about the two taxi ridesand do check the hypotheses of the math. Else usesomething more severe and careful.That simple description does not mean that goodapplications are easy. I very much doubt thatsuccessful applications of math on Wall Street areoften easy.Uh, the work is not just throwing stuff against thewall to see if it sticks. I wouldn’t bet on anysuch stuff.I was making good progress on getting my Web pageswritten this afternoon.So, for a little background music, listened to someof what poor knight Lohengrin was explaining to hispoor, new wife Elsa about her unfortunate questionabout Lohengrin’s origins. Finally it dawned on methat the Lohengrin aria was intense music and,maybe, with some better performances on YouTube.Yup, there ishttps://www.youtube.com/wat…by Peter Anders.With just a little of the story and some subtitles,the intensity of the aria is clear. There areEnglish subtitles athttps://www.youtube.com/wat…with a Placido Domingo performance.So, poor Elsa blew it — had just gotten married toa perfect knight of the grail with the best weddingmusic in all of history and, then, right away ruinedher marriage so that Lohengrin walked away from her.So, in this aria Lohengrin is walking away from thegirl he just married. Big time bummer.In high school I was in love with a girl; we had amisunderstanding; she did something she should nothave done (although maybe her intention was good ifher actual effort looked like it was not), and nerdme walked away from her, leaving two broken hearts.I wasn’t the first person to do such a thing. Ifperfect knight Lohengrin couldn’t patch up thesituation, no wonder nerd I couldn’t.I could now, easily. There was a chance. Maybesomeday I’ll write a book ‘Girls 101 for Dummies –Boys’ and tell boys what I wish I’d known then –a big secret, i.e., what a romantic relationshipreally is all about (largely in E. Fromm, ‘The Artof Loving’), how to be a good leader, how tocommunicate to avoid misunderstandings, and how tosolve problems without walking away and while makingthe relationship stronger. I could have done it.My father didn’t know that stuff, either — in hisgeneration all he had to do was play a role; notnow. Now romantic relationships are broken,literally, since the birth rate is so low we aregoing extinct, literally. I’m not the first guystruggling against the tsunami.I wish I’d married her.For the Lohengrin aria, search for just the firstfew words”in fernem land”The words of the aria are intense in an indirectsense — he never says anything personal to the girlbut just explains the forces he is under that are sotragic and force him to walk away from his new wife.I’m no knight of the grail, yet still that ariacaptures the hell I felt when I believed I had towalk away from that girl. Since the opera was in1850, I’m not the first instance covered by this’universal art’. So, I was not all of the cause.This is one more lesson in understanding the humancondition.Back to my Web pages.

        1. Dave Pinsen

          You may be too smart to be a businessman.Regarding Thorp, I think his example suggests the opposite of your point. He is a genius and was a pioneer in thinking quantitatively and sophisticatedly about a new asset class, and he made a bundle because of that. And yet Thorp’s hedge fund today is probably smaller than Simmons’s annual vigorish. Because Simmons knows some abstruse math that Thorp can’t grok? It’s possible, but I’m skeptical.There’s a lot more brain power and computer power being brought to bear in the markets today than when Thorp started, and as a result, uncovered arbitrage opportunities are quite rare, particularly at the scale a Simmons (or a Buffett, for that matter) is dealing with.I lost the thread a bit when you segued into opera but will watch those video clips.

          1. sigmaalgebra

            We can cover the items in your comment withoutusing anything difficult.> You may be too smart to be a businessman.It’s possible to be too dumb to be somethingbut not too smart. Indeed, it would be dumb tolet good talents or knowledge get in the way ofimportant progress. Being “too smart” would belike running too fast to be a football tightend, having batting average too high to play inMLB, or swimming too fast to be on the USOlympic swimming team.In my mind, there’s considerable question abouthow “smart” anyone is or is not. ‘Smarts’ aretoo difficult to characterize or evaluate inany very comprehensive way. I’ve been aroundsome high end academics long enough to haveseen some people do astounding feats of mentallegerdemain yet have difficulty using some ofwhat they should have learned, and maybe didlearn, in middle school. I’m not that way.In high end academics, the main criterion of’smarts’ is just published research; otherwisemostly no one cares. So, people look at workaccomplishments and not ‘smarts’ alone. Afterwe see the accomplishments maybe we attribute’smarts’.Math is a special field: Nearly each result isrock solid from either a rock solid proof orcounterexample. Commonly it is assumed thatmath is the most difficult college major. Ithought that making sense out of Englishliterature was more difficult, but the savinggrace was that it was never very clear thatanyone could make much sense out of it so thatmy poor attempts didn’t look so bad!But people who have not been through, in finedetail, a real problem, some original math fora solution, and the solution itself will have atough time having any confidence in the role,relevance, or power of math. In particular,information technology venture capital isconvinced down to the center of the cells atthe center of its bone marrow that the’technology’ in ‘information technology’ isjust routine coding such as can be learned inmiddle school. Totally convinced. Thankfullyfor US national security, the US DoD has for70+ years done much, much better.This fact can affect my prospects of seedequity funding but likely not a Series A sincethat funding would likely be based mostly juston ‘traction’ significant and growing rapidly;I wouldn’t be required to mention the math; theVCs would assume that the software was justsome routine coding, and I wouldn’t have totell them differently.Of course, since I’m a solo founder, my ‘burnrate’ is low, and any significant ‘traction’will put me comfortably in a ‘lifestylebusiness’, with three Corvettes, one each ofred, white, and blue, and I’d be reluctant evento walk across a street for a Series A check,especially since it would come with a Board ofVCs attached, likely VCs who would upchuck atthe first Board meeting where I presented plansfor the next step in the core technology.Of course, a standard remark is that venturecapital is not for everyone. Yup.But the economy and business are darnedimportant. Business isn’t just for the poorstudents anymore. Although actually in mostrespects I wasn’t a very good student, but Idid do better as I went farther in school andfound my dissertation research to be fast, fun,and easy. Same for each of my publishedresearch papers (I don’t like publishingpapers; it’s like giving away work for free)and for the math for my startup. Fun stuff.In my startup, I’m just taking what is a fairlyobvious problem that applies to nearly everyInternet user in the world. In short, theproblem has been stated often enough here atAVC, most recently by Fred as he noted howdifficult it is to find blogs he’s interestedin and video clips he wants to watch. One ofthe issues is that the Google/Bingkeyword/phrase search technique works poorly onthose two content ‘types’.Why? For video clips there isn’t much in textto match. For both blogs and video clips,what’s crucial, that is, what the user wants,is ‘content’ with some ‘meaning’ the user willlike. E.g., to please Fred with blogs have tohave some appropriate ‘content’ in, say,business, technology, maybe parenting andtravel, and keywords/phrases make just hash outof such meaning.Or, too often at Google/Bing, get millions ofhits and then have to accept the givenpopularity sorting; e.g., in music that wouldresult in Top 40 instead of Bach, Mozart,Puccini, Verdi, Wagner, R. Strauss which wouldbe lost in the last 10% of millions of hits.There have been many efforts at solutions fromLast.FM to Pinterest, etc., but, of course,none workx very well.To be more clear, the problem I’m attacking isnot mathematical; the only role for math is thecrucial, core, internals of how some data ismanipulated deep inside my server farm. Again,I’m definitely not going to people who havemath problems, know that, and want mathsolutions. Instead, I’m going for 1+ billioncommon men on the streets around the world.Again, the problem is totally from the realworld and known to 1+ billion people; to them,it’s not a math problem, and they will have noidea that any math was involved. Nearlyeveryone will assume that the ‘technology’ inmy server farm is just routine software thatjust any software developer could write.So, seeing the real problem and what didn’twork well, I worked up a quite differentsolution. Yes, buried inside two chunks ofserver software is some code implementing someoriginal math I worked out using some advancedprerequisites, but the user experience and userinterface give no hints about the math, andusers will ‘catch on’ how to use my Web siteintuitively quickly.That the users won’t know anything about thecore math is fine — they won’t become afraidof the math.If it all works, while my site will be usefulto nearly all Internet users, how well userswill like my site will vary a lot: Someone whowants just to be a couch potato in front of aboob tube won’t much care — they don’t muchcare about anything else either. But peoplewho want ‘more’ might become quite good users.If Fred still wants to find the stuff hementioned, then he should become a frequent andhappy user.Some statistics on traffic to Google, YouTube,blogs, etc. show that easily a billion uniqueusers a month care about ‘more’.So, I’m doing an ‘information technology’ (IT)startup. Lots of people are, and successfulones are rare. E.g., Andreessen-Horowitzclaims that they get 5000 business plans ayear, and recently Andreessen claimed thatthere are only about 15 projects a year worth aSeries A.So, successful startups are ‘exceptional’.Then what has been or is ‘usual’ for ‘business’is a very poor guide to what is exceptional –won’t get much idea what Formula One race carengineering is like by watching morningcommuter traffic, what Bach, Mozart, Wagner,Puccini, R. Strauss wrote by listening to theTop 40, what Ferran AdriĆ  cooked by eatingtapas, how Michael Jordan played by watching myefforts at practicing free throws (the ball toooften ended up over the fence into the yardnext door — nice yard in a sense since the guyand his several siblings built one heck of anice business important in at least threestates, while, unfortunately his wife had onechild and then retired to a back bedroom tolive on chocolates and gain 150 pounds — moreexamples of the ‘human condition’).So, for a better solution, I stirred up somemath. It’s an advantage, stands to yield amuch better and the first good solution to theproblem.And the role of math is at least quite naturaland maybe in the future very common. Why?Because quite generally in computing, softwaretakes in data, manipulates it, and reportsresults. Well, the manipulations arenecessarily mathematically something,understood or not, powerful or not. For morepowerful, understood manipulations, proceedmathematically, i.e., with theorems and proofs.And proceed as I outlined earlier in thisthread with my transportation analogy of townReal Question, taxi Justified Assumption, townMathematical Question, Mathematics Airline,town Mathematical Answer, taxi JustifiedInterpretation, town Real Answer.Why more powerful manipulations? To get theanswers Fred and 1+ billion other Internetusers want and can’t get now. But Page/Brin,Gates, and Marissa Mayer don’t have to beat upon their employees to try to guess what I’mdoing in part because I’m using some differentdata — I couldn’t do very well with just theirdata.Generally does such a detour through some mathcircumventing the obvious slog through muddyswamps give better solutions to the coretechnical problems? Sure. Evidence: Radioand TV engineering, aeronautical engineering,transistors, GPS, stealth, etc. Evaluate thesurvivability of the US SSBN fleet under aspecial scenario of global nuclear war limitedto sea? Sure: Start with a WWII Koopmanresult and see a continuous time, discretestate space Markov process and apply MonteCarlo — the US Navy got their answers in thetwo weeks they wanted complete with a technicalreview by mathematician J. Keilson, e.g., as in’Green’s Function Methods in ProbabilityTheory’.I saw, created, a list of such applications ofmath early in my career. It can be done.Although unpleasant news to a lot of B-schoolstudents, making such progress is much of whatthe math departments are there for and, indeed,why Congress votes funds for NSF for thosedepartments. Or, a bit dramatic but not reallywrong, in the opening of the movie about J.Nash, “Mathematics won the war [WWII].”. Bythe way, there’s a nice elementary proof, as Irecall from Lemke, of Nash’s result inT. Parthasarathy and T. E. S. Raghavan, ‘SomeTopics in Two-Person Games’, ISBN0-444-00059-3, American Elsevier, New York.Sure, a dream is that just anyone can think ofan ‘idea’, that is, a new product/service, thatthey can program in an accelerator for mobiledevices and make $10+ billion. But we alreadyhave some good data on the chances; that is,just how often such a thing has happened in thelast 20 years of the Internet. So, for a poorguy losing sleep doing such work, it’s alottery shot.We need some better odds and, from that, somemore powerful methods. I see no reason the 15Series A projects a year might not be 150,1500, …. But for such a future, a guy inmiddle school who just taught himself (or evenherself) C++ has one heck of a handicap. E.g.,he’s missing out on the advantages of the workof Lebesgue, Weiner, Kolmogorov, von Neumann,Doob, Feller, and others.It’s tough enough to learn that material ingood courses in a good graduate school;otherwise reinventing that stuff is justimpossible — no one is nearly that bright.E.g., the Lindeberg-Feller result is tough evento state, and the proof is grim. The Doobdecomposition is astounding and, thus,difficult to guess (with the right backgroundit’s easy), but the martingale convergencetheorem will blow a mind, and the proof is atour de force of originality and cleverness.No way can one person just reinvent all thatstuff. Similarly a lone country doctor 100years ago had no hope of inventing the Salkvaccine.For Thorp, I read his books on blackjack andthe stock market. Okay. For “arbitrageopportunities”, one of those is essentiallywhat Thorp found.In simple terms, E. Fama might claim thateither the efficient market hypothesis holds orthere are more arbitrage opportunities. So,since I have a tough time believing that themarket is fully ‘efficient’, I have to believethat there are arbitrage opportunitiesremaining. How Simons made so much money, Idon’t know. Similarly for SAC. I doubt thatthe secret is some guy on a motorcycle askingprivate airport ramp personnel about thedestinations of Gulfstream jets.In Thorp’s book on the market, in the back hehas that an argument is just routine “measuretheory”. So, before graduate school, I got agood head start on measure theory from H.Royden’s famous book. In graduate school,since I was hoping for a Wall Street career, Imade sure to get a good background in measuretheory, e.g., through Sierpinski’s famoustricky exercise (yes, in Halmos, ‘MeasureTheory’), regular conditional probabilities(cute stuff — saved my tail feathers once),martingale theory, ergodic theory, etc., wrotemy dissertation on stochastic optimal control,circulated my resume around Wall Street, butgot nowhere. Had not yet heard of Simons. Didget a letter back from Fisher Black (right,Black-Scholes) saying that he saw noopportunities for math on Wall Street –amazing. So, do a startup.There’s a fundamental ‘organizational’ problem:Business still wants to be organized much likea Henry Ford factory where the supervisor knowsmore and the subordinate is there to applymuscle to the thinking of the supervisor. So,a subordinate can’t do math the managementchain above doesn’t understand. Sure, Simonsis an exception; so is academics: An assistantprof who wins a Nobel prize is no threat to thecareer of his department chair.Similarly in venture capital: As reported byFred here on AVC, on average venture returnsover the past 10 years suck. But the venturepartners, and their limited partners, wouldrather give up on venture capital than investin anything with technology they didn’t see inbusiness school except for software where theywant the technology to be essentially justroutine that they can evaluate just by ‘playingwith’ the user interface.That is, the VCs and LPs are just terrified offlim-flam ‘snow jobs’, ask where are the yachtsof the mathematicians, laugh, and don’t want tothink about Simons. Also a lot of VCs arearrogant, even if they are not making muchmoney, and want to think of themselves asmasters of the universe and the smartest guysaround with the entrepreneurs as just the”little people” and get their little egos hurtby any mention of anything they don’tunderstand.Spend much time around high end academics andfind that everyday are awash in things don’tunderstand and quickly know that no one canknow everything and no one can carry thelibrary around between their ears. It’s socute that so many VCs have such headaches atsomething they don’t understand — poor littlechildren! If they really understood how muchthey didn’t understand, then they’d have atough time getting out of bed ever again! Goodthat few VCs are pretty girls; I’d feel sorryabout the girls; but not for the boys!The Dodo birds didn’t do very well, either.And ordinary kitty cats gone wild are having afeast in Australia!A guy opening a Main Street pizza shop may wantto look a lot like successful pizza shops he’sseen elsewhere. But Facebook is not anotherGoogle is not another Yahoo is not anotherCisco is not another Microsoft is not anotherIBM. For exceptional successes, can’t beafraid of bringing something new to the space(especially if the thing being brought hasabout the most solid pedigree in all ofcivilization).The opera is some insight into the ‘humancondition’ which now is so bad that the peopleof Western European descent in the US andEurope have birth rate so low they are goingextinct — literally. We’re seriously messedup — e.g., going extinct. It’s not just theVC average ROI that sucks; the whole thingsucks, badly.For Wagner’s opera ‘Lohengrin’, at http://www.rwagner.net/libr…is a collection of really nice Web pages withthe libretto in German and English and links tosome of the score and to MIDI and more for someof the music.The aria “In fernem Land”, that’s in act three,scene two, and, thus, in H:DATA05PROJECTSMUSICWAGNERLOHENGRINact_three_2.htmThe full score is in http://imslp.org/wiki/Speci…and the aria starts on page 353.The music starts in a curious way: It’s justin A major, and the vocal part as it starts isnearly all from just a fourth below the open Astring on a violin (the only A on the mainlines of the treble clef) to a fifth above.The opening from the orchestra is just dirtsimple, just the chord A-C#-E except the’voicing’ (picking in what octave to play oneof the three notes), ‘orchestration’ (pickingwhat instruments to play what notes), anddynamics make the music sound ‘ethereal’. Sothe instruments are violins, in unusually manysections, flutes, and obos. The dynamics aresimple, just a few markings, with some of theinstruments fading in and out. It’s reallysimple but really effective.So, with that opera and that aria, there’s someinsight into the human condition, some of the’universality’ of art, where many in theaudience can say, “There were things somethinglike that in my life, and, thus, I’m not theonly such person suffering with this part ofthe human condition and not all the problemsare just my fault alone.”. The music, story,and libretto all get the art across nicely.Knowing just a little about the story and thelibretto, the aria is intense and effective.So, Lohengrin, a perfect knight of the grail,gets married to super sweet Elsa; she makes amistake; and Lohengrin, even with all hispowers, is forced to walk away from his bride.Bummer. At one time in high school, I was inlove and felt forced to walk away from mygirlfriend, and Wagner’s music tells me that Iwas not the first.Since the opera was first performed in 1850,the story and its part of the human conditionare old, not nearly new. So, young people needto know this story because there is a goodchance they will be in such a situation andneed a better solution than perfect knightLohengrin had. Although to some people thelessons were clear enough in 1850, my fathernever knew them, and I learned the lessons toolate to do well with the challenge I had inhigh school — bummer. Ah, maybe someday I’llwrite the book I mentioned.The Anders performance I linked to is good; hisstyle of singing is good for that aria. He’squite careful with each note of the music andeach syllable of the libretto.Several other arias from that opera are famous,and two preludes and the wedding march are someof the most famous music in all of classicalmusic.To me the music and its intensity reallycommunicate the agony of what it felt like forLohengrin to walk away from Elsa and for me towalk away from that girl in high school — hugebummer.Yes, of course, as we can easily hear, here andelsewhere in his work, Wagner was justfantastic as a composer, beyond belief. Butalso as is well known, in some respects outsideof music he was a wacko. So, just appreciatehim for his music.Music has a big advantage — can heavilyappreciate and value it without digging throughall the details, e.g., reading the score;mostly not so for math. So, very few peoplecan appreciate or value math.

          2. Dave Pinsen

            “It’s possible to be too dumb to be somethingbut not too smart. Indeed, it would be dumb tolet good talents or knowledge get in the way ofimportant progress. Being “too smart” would belike running too fast to be a football tightend”There are usually trade-offs when it comes to attributes. Someone who is extraordinarily fast usually doesn’t have the size to be an NFL tight end (there are occasional exceptions — Jeremy Shockey was one early in his career). As for being too smart for business, some of these folks would seem to be examples of that.No need to defend math — I have plenty of respect for its utility. I just expressed skepticism about attributing Simons’s returns solely to it. I agree that Simmons probably wasn’t relying on a young Charlie Sheen for his returns. But I think a Simmons or a Cohen could think up more sophisticated ways to tilt the table in their direction than Oliver Stone. Not saying they necessarily did, but I remain skeptical.BTW, why do you use a pseudonym?

      2. Tom Labus

        Great quote and more accurate today than ever with all the noise around the market and finance.

    5. ShanaC

      We’re doing: somepowerful, relatively advanced, mathematicalprerequisites. Some of the advanced results arebeyond belief; no one without the extensivebackground could even guess that such things couldbe true; they are true, but the proofs are oftenchallenging even if are told the results.As not me, I expect it to work just based on that. The knowledge plus the ability to technically support the knowledge is not simple.

      1. sigmaalgebra

        I can’t quite parse what you wrote, butthere’s much more to a successfulapplication than just what you quoted.The quote, however, in case the applicationdoes work, will help with ‘defensibility’,a major theme at AVC today. Indeed,clearly the quote was about defensibility.

  10. William Mougayar

    Finally got to it. There were some new insights into your thinking. 2 big themes:1) Big data + healthcare is a big deal. Iā€™m waiting for it. I think when we start to aggregate a lot of patient data directly from the source, we will gain much better insights. Currently, the ā€œdata chainā€ is very fragmented. If you can disrupt that, we will have big insights, solve some big puzzles, and make patients happier.2) Bitcoin as the financial Operating System of the Web. Yes to that. I get it. And I agree that we need to start seeing Bitcoin Apps on top of the infrastructure. You were lamenting that the focus is still on Bitcoin middleware. But Bitcoin is going to follow the same evolution as the Internet or any foundational technology; that is to go through the 3 pieces sequentially: 1) Infrastructure goes first typically, then 2) itā€™s followed by Middleware, and finally 3) innovation around Applications will emerge.

    1. sigmaalgebra

      For 1), be careful. Back when I was working in’artificial intelligence’ (AI) at Yorktown Heights,I learned that the AI community had finallydiscovered the importance of what they called ‘deepknowledge’, and it promises to be just crucial formany if not most hoped for medical applications from’big data’.Here’s what’s going on: Let’s roll back tosomething simpler and better understood than thehuman body, say, cars. Well, the AI communitydiscovered that to do good diagnosis, etc. of cars,need actually to know are working on cars and notsay, just sewing machines, toasters, or washingmachines.So, for cars, need to know that actually there is anengine, clutch of some kind, a transmission, usuallysomething in the way of a drive shaft, adifferential between two driven wheels, half axlesfrom the differential to the driven wheels, etc. andhow these things connect. And need similar detailsfor the brakes, air conditioning, cooling system,electrical system, etc.It was all those ‘details’ of how the machine’worked’, e.g., what connected with what, thatconstituted some of the necessary ‘deep knowledge’and much different for cars than for, say, washingmachines.Of course with enough ‘big data’ can hope to dodiagnosis with just the ‘big data’, and, yes, canmake some progress that way, but the AI communityconcluded that a little ‘deep knowledge’ wouldeasily replace and improve on some huge oceans ofjust dumb ‘big data’. I believe that here the AIcommunity was correct (maybe the only time).So, back to medicine, a physician has, is just awashin, genuine ‘deep knowledge’ about the human bodyand, indeed, what is unique about individualpatients.The combinatorial explosion of circumstances mightbe so large that, really, from the view of justdata, each patient is unique, that is, in the senseof statistical ‘machine learning’ a case of numberof samples n = 1, which messes up most approaches tostatistics. Net, the way a physician cuts throughthis combinatorial explosion is his version of ‘deepknowledge’. E.g., if the patient has been partinghis hair on the left instead of the right, rightaway a physician can judge with high accuracy ifthat fact is significant, but doing the same withjust statistics and ‘big data’ nearly guarantees toask for more data than has yet been observed for allof humanity. Combinatorial explosions are the ‘bigbang’ of data — they get big in a big hurry, bigenough to make ‘big data’ look tiny.E.g., from a short calculation, it doesn’t take avery big enterprise linear programming problem tohave more extreme points than there would beneutrons if all the visible universe were packedsolid with neutron star material. I know; I know;those 3 TB disk drives look really, really ‘big’.Right, Virginia, 3 TB is really big, plenty for alist of even your Facebook friends!More broadly, the problem with AI is that it’s longon the A and short on the I. So far, AI has a longway to go to catch with the I of, say, a six monthold kitten. Or, the problem with AI is that stillit doesn’t have even a weak little hollow hint of atiny clue about how the I works, not even inkittens, birds, mice, etc.The useful work I saw in AI was just good appliedmath and engineering done in quite traditional waysand only called AI. If there is a useful case of AI,then wipe off the surface dust and find some work inapplied math or engineering that was just called AI.For the ‘field’ of AI, what is not just good appliedmath or engineering, call that the empty set.There’s nothing there. AI is not a ‘field’; maybeit is a goal, for now more like a dream.Every now and then when I get too optimistic Ientertain myself thinking that I have a good, highlevel ‘architecture’ that would yield real I for,say, just text I/O. Yes, the AI people long saidthat they needed to attack the ‘representationalissues’, and for that my dream is something close tojust a natural language as if real I needed and wasclosely based on a language — so start with wordsand language and build from there, of course, tryingto imitate how a child learns language.Right: Got to be smoking some strong funny stuff tohave any faith that such an ‘architecture’ has anyvalue, and I’ve never smoked anything, funny orotherwise.For now, for the crucial, core technologyapplications of information technology andcomputing, just stick with what we actually know howto work with, applied math and engineering. This isgood news and says that for all that time might havewanted to spend on ‘learning AI’, just do somethingmuch more important, say, pet a kitty cat and beamazed at real I.Once I did some applied probability for anomalydetection in server farms and networks. I stillthink it is useful and could be the core of avaluable company. Eventually I found a VC in Bostoninterested, and he understood enough to ask alsoabout the next step after detection, that is,diagnosis. Since I didn’t want to lead both of uson a long walk on a short pier, I just advised himthat if anyone comes to him promising to do anythingsignificant or very useful in diagnosis, then justbreak out the laughing gas and have a good timebecause that’s about the best that can done withsuch a promise. That remains the case for medicine.In short, detection, under some circumstances wherehave some appropriate mathematical assumptionsjustified, can be just a cute problem in appliedprobability; anything significant in diagnosis,however, needs to know a LOT, e.g., ‘deep knowledge’about the system being monitored for detections, andthat ‘deep knowledge’ is forbiddingly more work.I made some progress on my Web page today; got downto where I need to do the serialization and use someof my old software to send to the server for thecomputing!

      1. William Mougayar

        I know that doctors don’t like more informed patients, because they ask a lot of questions, and there’s a part of medicine that is still an art where the answers and outcomes aren’t as clear cut as one would like them to be. It is what it is.

  11. ShanaC

    Maybe because I’m starting a company in the data field, but data by itself is rarely really interesting or defensible. What you do with it is usually both.

    1. andyidsinga

      I started thinking about the value of large data collections after Alexa was purchased by amazon in the 90s for like 200+ mm.(afaik) They had a huge DB of crawled websites and had created data about the links between them.My continuing stance on this is that building a massive database on just about anything is always a good thing and because you cannot do it *fast* it becomes a defensible and salable asset.Once you have data – you can throw a lot of brain power and compute power at it to process it quicker (or slower), put APIs around it ..whatever.

  12. Daniel Bachmann

    Interesting to listen to Matt’s great video, who built his blog 10 years ago, which is a real inspiration for me to keep on blogging, while I started 10 days ago! Big Data is clearly still a growth area and there are many different aspects to it, but also pulled further by the social and mobile enterprise!

  13. andyidsinga

    the discussion about nsa / spying / radical transparency are very interesting. We’ve been talking about that at Albert’s blog too (cc @albert:disqus ).The problem I keep having with a completely radically transparent future is its relationship to “what you can’t say” ( see: http://www.paulgraham.com/s… ) and moral fashions. I keep thinking that we need to be able to have some sort of privacy/secrecy in order to test, with trusted people, what is morally un-fashionable.