Live data from Hacker News

Black Swan Farming

paulgraham.com

271–280 of 321 posts

Re: Black Swan Farming

#271
Investing in startups has become a giant ponzy scheme where the idea that can be transformed in a giant hype machine can bring more results as more people invest in the hype before the bubble bursts.

Facebook is useless, Airbnb will be gone in five years when more horror stories surface, as well as dropbox, today storage is so cheap anybody can have 20GB for free when buying a cheapo hosting plan. Even GAE gives you 10GB free, all you need is a piece of software.

So, etsy will last forever, facebook won't, amazon will, dropbox won't, godaddy will, airbnb won't.

Sometimes I think, investing a billion dollars in instagram is running a freight train loaded with money over a cliff.

That money would have been better invested in a new eBay (without auctions) that would last forever, empowering small merchants and providing a complete experience for ecommerce, from payment to delivery.

We already had enough with social, but the hype machine won't die just yet, so fuck investors, I hope they lose all their money and then blame the economy.

Re: Black Swan Farming

#272
perhaps paul's intuition is right - there is something wrong here. certainly real world business success is not so binary as this.

if there is a 500 pound supposition in the room it's liquidity. if the only way an investor can achieve it is through a massive "event" such as an ipo or sell out then any otherwise successful company is a non-entity for investors.

if y combinator were holding securities with some liquidity it needn't devolve into a 20th century vc. there are many ways to achieve this, all disruptive yet also quite well suited for this institution and its community.

Re: Black Swan Farming

#273
post #232

Earlier quoted context omitted.

For me, Dropbox beautifully solves four hard problems: 1. Keeping all my documents in sync between several computers, so that I can pick up work on any one of them at any time 2. Off-site incremental backup 3. Sharing and working on a set of project files with non-technically-minded collaborators 4. Sending and receiving large files Moreover, it just sits there quietly solving problems 1 - 3 day-in, day-out, and I ne…

I have to think about it all the time. It flatly fails to serve my gaming club. We have a Gb of card images for our card games. Almost every club member cannot share these files, since they run out of space. Think about it. Sharing files between 25 club members, using our own bandwidth, our own disk space. And Dropbox thinks we should pay them big bucks for this. For what? Storing our files on their server, insecurel…

Well, OK. The fact that it doesn't solve any problems for you has little bearing on the excitement of those who, like me, find that it solves many of their problems.

The reason that storing files on their server (and using their bandwidth, as well as yours) is important is that it means I don't have to keep all my synced computers on at all times. That's a big deal for me.

It sounds like yours is another problem, which Dropbox is not well adapted to. Perhaps you'd be better off with PowerFolder or similar.

Re: Black Swan Farming

#274

Earlier quoted context omitted.

With self publishing now, it is likely that unpopular works will persist. Over time there is more of a chance that anything great that was passed over will eventually find it's audience. Whereas in the past that stuff that didn't get published may have only existed as a single or few copies and was eventually lost to the passage of time.

That would be great, but it assumes a writer with marketing and promotion skills. Even if a large potential audience exists for a book, reaching that audience can be very difficult, which is a big part of the value large publishers offer.

I am loathe to admit record labels providing any value, but they do this as well, helping good (and bad) artists rise above the noise floor to reach a mass audience.

Re: Black Swan Farming

#275
post #273

Earlier quoted context omitted.

I have to think about it all the time. It flatly fails to serve my gaming club. We have a Gb of card images for our card games. Almost every club member cannot share these files, since they run out of space. Think about it. Sharing files between 25 club members, using our own bandwidth, our own disk space. And Dropbox thinks we should pay them big bucks for this. For what? Storing our files on their server, insecurel…

Well, OK. The fact that it doesn't solve any problems for you has little bearing on the excitement of those who, like me, find that it solves many of their problems. The reason that storing files on their server (and using their bandwidth, as well as yours) is important is that it means I don't have to keep all my synced computers on at all times. That's a big deal for me. It sounds like yours is another problem, whi…

Just posting my experience, like others here do all the time. Another example is instructive.

My problem is, Dropbox scales the cost as the number of people looking at a folder increases. As an Engineer I see that as marketing, their cost doesn't increase incrementally in this case. It seems unnecessary and blocks me.

Btw you would only have to keep 1 synced computer on, some of the time. Not a big deal actually. And why keep my data around on their server after we're synced? Simpler for them I suppose, but insecure for me.

I could try and get everybody in the club to install another tool; might look into that, thx.

Re: Black Swan Farming

#276

Earlier quoted context omitted.

You talked yourself into nonsense. On waste in DoD, sure, but not the parts I was talking about. Some of the big waste was, say, getting AC working in Iraq and now getting Diesel fuel and jet fuel to Akrapistan. And the cost of the black oil to send a destroyer across the Pacific would really set one back. > Also you appear not to have any idea how much waste exists in Pentagon contracting. Nonsense. All my early car…

This is so silly. You don't have a startup idea until you have a product that you can sell to Grandma, that will work for her on a standalone basis. It can't require an Act of God to take it market. It has to be cheap enough that Grandma can buy it, and it has to be simple enough that it can be built by 3 guys in a shack in Palo Alto. Often this means taking stuff that exists already and chopping off the most expensi…

Much that you say is okay in general although nearly irrelevant to what I have written on this blog or done in my own work. You seem to be hinting that research that is new has to struggle with professors to be converted to a product, has to be too expensive to convert to a product, can't result in a project that can be done by three guys in a garage, and can't result in a cheap product Grandma can use. All of these hints are true for some research but not all.

You mentioned DSP: Okay, there was a Presidential Science Advisors meeting with J. Tukey writing notes with both hands and R. Garwin. Garwin saw Tukey writing what looked like Fourier math and asked if he knew anything Garwin didn't since he (Garwin) was struggling doing numerical Fourier transforms. Sure, Tukey knew something. Back at the office Garwin showed Cooley, and the Cooley, Tukey fast Fourier transform came alive, and that is the core of DSP.

Now all the 'research' required was just one or a few pages of derivations with elementary facts of trigonometry. Still, the possibility of the FFT was widely missed. So, all Tukey had to do was just keep his mouth shut, get with Texas Instruments or some such, work out a deal, build an FFT chip, try to keep secret how it worked, or maybe patent it, and get rich. These days he could run a SaaS site that ran applications software that heavily exploits the FFT. E.g., run software for seismic signal processing based on the FFT. If Tukey was the only guy in the world who knew the math behind the FFT, then he could have a license to print money.

Seismic signal processing? Sure, for the oil industry. Look up "The Abcs of Seismic Exploration and Processing Enders A. Robinson" and other publications of Robinson. One of his books has a nice collection of FFT and related software in an appendix. At one time I was an expert in the FFT, and that expertise got me a new, high end Camaro. Keeping his research on the FFT a secret could have made Tukey a fortune.

Well, Tukey is not the only person who can do such research! And some of us who can do such research, e.g., with just paper and a pencil and then a cheap computer, also with TeX, now can decline to publish and, instead, just build the research results into software run on a server farm.

E.g., once a guy built an FFT box, all hardware. He explained it to me complete with the logic for the different, tricky signal flow graphs between the stages. So each different graph needed different hardware. So, I mentioned to him that there was another version of the FFT that used the same signal flow graph at each stage, and, of course, right away he noticed that he could have saved a lot of hardware!

Of course, as I outlined, we don't just do the research as a shot in the dark. Instead, as I outlined, first do Step 1, that is, pick an important, unsolved problem. In Tukey's case it would have been running a SaaS shop for seismic signal processing. Then in Step 2, try to do some research to get a powerful solution to the important problem. In Tukey's case, that's where he would have done his research for the FFT. You are following me, right?

So, we don't do the research until we have a problem in mind. There's a recipe for rabbit stew that starts, "First catch a rabbit". Well my recipe for applied research is, "First get an application" or, if you will, a problem. THEN try to do the research. If the research fails, then pick another problem. However, if the research is successful, then right away you have at least on valuable application, and that fact puts your research near the top of the pile of what is relevant for making money soon.

That's what I've done for my project and what I recommended as a solution for the project evaluation struggles in PG's essay.

So, in your terminology, in my recommendation, the incentives and goals very much ARE aligned.

There's no law that says that I, or anyone who can do research, can't put my feet up, think of an important problem, review tools that might be useful for finding a solution, think of a solution, as original work, 'research', and proceed. The research need be no more than Stan Ulam's "few scribbles on a piece of paper" that change the world. That's all there was to Tukey's FFT, and that's all there is to my research for my project. But lots of people struggled with Fourier transforms and didn't see what Tukey did. And lots of people are struggling with the important problem of my project with essentially no chance of reinventing the research I did. And my research stands to be much more powerful than anything those other guys will stir up -- for just why, an explanation here would just get people to willingly misunderstand in order to argue and the HN mods would down vote me with glee.

You seem to be missing what I said about DoD research and, instead, are running off in various other directions critical of the DoD. Your criticisms may be correct, but they are basically irrelevant to what I said: My Step 2 asks for some research to find a powerful solution to the important, unsolved practical problem of my Step 1. So, a question is, can there be such research? My answer is, sure: The US DoD has been providing excellent examples for over 70 years now. This answer is rock solidly correct and answers the question. That the DoD wasts money in various other ways is irrelevant to my simple point here.

The core problem PG is seeing in project evaluation is that the projects he is considering are neglecting both my Step 1, select an important, unsolved problem, and Step 2, do some research to get a powerful solution to the problem.

The big advantage of picking an "important, unsolved problem" as in my Step 1 is that, like my example of a one pill cure for any cancer, we are absolutely certain that there will be good 'product/market fit', that is, that customers will line up. So, the first problem with the projects PG is seeing is that the problems being solved are not both unsolved and important enough. So, it's not clear if some such project will be another InstaGram or PInterest or just forgotten. As explained by PG and on this thread, for such projects, some catch on and some don't, and early on it's a shot in the dark to know. That a good, one pill cure for cancer would "catch on" is beyond doubt. Then some of the struggles PG explained in project evaluation would go away. See the importance of picking an "important, unsolved problem"?

The big advantage of doing some research, as in my Step 2, for a powerful solution to the problem of Step 1 is that do get a solution to the unsolved problem, or maybe just a much better solution than anything else, and get a technological barrier to entry. Now we come to the common remark that "Anything you can do 1000 other people can do, are doing, or have already done", etc., NO! If Tukey had just stayed quiet, then it would have been a long time until others had figured out what the heck he had done. Sure, the main ideas of the FFT had been in the literature, in one form or another, going way back; still what Tukey did was nice, high quality research. All he had to do was just be quiet. Of course, Tukey was a big shot at Bell Labs and Princeton and in government science so was the kind of guy who would rush to publish. But by not publishing, think what he could have done for the Princeton endowment!

So, net, instead of struggling trying to evaluate the projects that PG sees, pick an important, unsolved problem and then do some research to get a powerful solution. For the DoD, it shows that such research is possible.

What is research? Not easy to give a good answer quickly. A good answer, and easy here, is just to get a technical Ph.D. from a good research university. Without a good understanding of what research is, mostly people, hearing 'research', will get all confused with rivers of misunderstandings, fears, jealousies, resentments, etc.

How to apply research to solve a practical problem? Again, a good answer is not easy to give quickly. To get a good understanding, get a job in a lab, maybe a DoD lab, that does such work. Again, bridging from research to a product, in some cases, is easy enough. But unless people really understand such bridging, say, from having seen it and done it, typically, again, they will get all confused and draw lots of wrong conclusions.

Look, it's clear: It's quite clear that quite likely 2+ billion Internet users will like to use my Web site based crucially on my research. But it is also clear that even here on HN and even on Sand Hill Road or Winter Street, any mention of such research is heresy, really offensive, really pisses people off.

The flip side of this situation is that (1) there are very few people, even among college profs, who have the prerequisites I used in my research, (2) only a tiny fraction of those people have any interest in entrepreneurship, (3) even if some qualified person did reinvent my research, they would face the same head in the sand attitudes about research I've seen on Sand Hill Road, (4) so that, net, I stand to have essentially mo competition.

So, I should just quite getting attacked on HN for trying to help PG, YC, HN, and SV, mail a letter I should send today, and return today to writing the last dozen or so of my Web pages as I planned. Off to it.

If anyone wants anymore about research, by all means get a good undergraduate major in pure math and then apply to graduate school in a technical field in a good research university. I've already been there, done that and never wanted to be a teacher and certainly don't want to now.

Today is the last day for me on HN.

Stay ignorant my friends, and arrogant, too!

Re: Black Swan Farming

#277
post #181
post #12

The counter-intuitive nature of startup investing is a big part of what makes it so interesting to me. In most aspects of life, we are trained to avoid risk and only pursue "good ideas" (e.g. try to be a lawyer, not a rock star). With startups, I get to focus on things that are probably bad ideas, but possibly great ideas. It's not for everyone, but for those of us who love chasing dreams, it can be a great adventure…

> "The counter-intuitive nature of startup investing..." The hidden premise of that statement, appears to imply there exists a universal intuition to which this is "counter". I'm not sure that is completely accurate, and is perhaps more useful to think about it in terms of a bifurcated process that can roughly segment the populace into two modes of preferred intuitive operation, but possibly more. Without going into…

I think you're bluffing. For someone with normal human intuitions, the reward curve of startup investing should be counterintuitive, because it's very different from all the reward curves that were available in the ancestral environment.

Re: Black Swan Farming

#278

This paragraph read a bit confusing to me: The fact that the best ideas seem like bad ideas makes it even harder to recognize the big winners. It means the probability of a startup making it really big is not merely not a constant fraction of the probability that it will succeed, but that the startups with a high probability of the former will seem to have a disproportionately low probability of the latter. It could…

There is a tradeodd between are startups that will succeed, i.e. return their risk-adjusted capital, and those that will make it big - the subset of the prior that are able to scale. One can get return-of-capital startups in niches that cannot scale or attempt to take on Google or Fb with a very low probability of success.

Ahh....ok....I get you.

Re: Black Swan Farming

#279

I'm not a VC, but I don't see how most successful startups seemed like bad ideas. Google was entering a crowded field, but the field sucked (still does, actually) and Google was obviously superior early on. As for Paypal, do you remember mailing checks after winning Ebay auctions back in the 90s? As for Youtube, do you remember what it was like to share videos in 2004? Enough said. Dropbox was about unbreaking broken…

It wasn't at all obvious that Google could do better than the big four of the day. There were at least thirty upstarts that had good results in some way or other. My money wouldn't have been on google.

With paypal it seemed crazy that sellers would hand over control of their bank accounts to this unqualified, unregulated company. Given paypal's history of account freezes, it still does.

Video sharing sucked in 2004, damn right. Microsoft and Real and Aol and Apple were pouring big money into making it work. Could we've guessed that someone was going to make a lot of money in internet video? Sure. Could we have predicted that it would be youtube? Much tougher. Dropbox was the same; there were hundreds of other products making much the same claims about how they could synchronize your files easily. Likewise skype; we'd had netmeeting for years (even today, polycom are still in business selling software that works less well than skype for hundreds of thousands).

Can you predict that a given service could be done much better? Maybe. But picking which startup is going to win based on that is a whole lot harder.

Re: Black Swan Farming

#280
post #108
post #12

The counter-intuitive nature of startup investing is a big part of what makes it so interesting to me. In most aspects of life, we are trained to avoid risk and only pursue "good ideas" (e.g. try to be a lawyer, not a rock star). With startups, I get to focus on things that are probably bad ideas, but possibly great ideas. It's not for everyone, but for those of us who love chasing dreams, it can be a great adventure…

So, if you randomly decided which startups to invest in, would you be more successful? Can you really predict anything? If you randomly invested in 100 startups, would your returns be better than your screening process? Can you test this?

We have a good deal of evidence that our selection process is better than random. We know it's at least internally consistent, in the sense that startups that are ranked higher in the application phase are more likely to make it past the interview phase. And we also in turn have (a necessarily small amount of) evidence that the startups that turn out to be big winners do the best in the interview phase.
Post reply on HN