Godel Incompleteness For Startups
skibinsky.com
skibinsky.com
This is really the same reason why people who think that software development process/methodology is all they need to be successful. They are effectively saying they have invented strong AI using people pushing bits of paper as their computing model.
Thats the whole point of business operations to take into account objective reality "full of unexpected random behaviors", find a formal system that will hide these complexities from the customer, and present him with easy, predictable and repeatable experience. Every time you get into Starbucks you get a cup of coffee, not a random outcome. Joel talked at length http://bit.ly/33fK5T how many little details are hidden in these corporate systems to produce what seemed like much simpler outcome for us on customer side.
Mistakes happen, employes are never ideal, yet that doesn't prevent creating formal systems how business operate, far from it. Its only differentiates good startups growing into good businesses. They are the ones who figure out correct way of formalizing their systems so they can grow and hire more employees, scale nationally etc.
(btw for a person who actually knows and practically uses math these observations would be self evident, yet perhaps less so for a typical comment troll )
You then need to show that business decisions/operation are sentences/statements in this formal system. Then you need to show that these formal systems that describe a business include Peano axioms for Godel's theorem to even apply.
All this to show what we freaking already know trivially. Business are not complete :D. There are other business ideas that are "orthogonal" to what any business does.
Take Joel article as example, and i can find countless others. When biz becomes big enough, the only way to retain manageability of operations is to create either explicit guidelines for their workforce to follow (precisely like "invented strong AI using people pushing bits of paper as their computing model" - although i would replace AI with wetware robot) or implicit conventions that form so called "corporate culture" yet these "conventions" are nevertheless is very clear logical deductions in whatever basic formal system that business established.
There is obvious weakness that i do not have full documentary access to any large enterprise to formally prove to you "yes, their 5000 pages of corporate guidelines, instruction manuals, employee handbooks and divisional policies are a formal system with no contradictory statements". McKinsey can do it, i can not. Take it as inherited weakness of popular blogging. Don't think that still degrades the article to "waste of time".
Another example of very explicit formal system: in another discussion, one of my correspondents points out how pre-crash hedge funds were effectively formal systems that KNEW about incoming crash, but logical following all the rules of systems they found themselves employed at were preventing them from doing anything to prevent such a crash:
<blockquote> when the collapse was coming was that all the smart people at the banks knew what was up, but didn't have a way to express it inside the firm. "Let's say I'm wrong: the ten guys at my desk make $100m each for the bank this year betting that the bubble doesn't burst, and I'm the guy who called them all idiots and then lost $25m shorting their positions. You think I keep my job long enough that when the bubble pops the year after, I'm around to say I told you so? Or in contrast, the bubble does pop, and each of the ten guys at my desk lost $500m and my shorts made the bank $2b. You think they're going to give me a $20m bonus for showing that they should have known sooner that they were bankrupting the company? Not only do they figure out a way to fuck me, but there's no bonuses that year anyway if all the traders were wrong." </blockquote>
How can anybody argue this is NOT a formal system? They have the their axioms, a finite list of rules how bonuses calculated, how performance is awarded, what action dictionary each trader has, and the deductions above are just "immediate consequence" as Godel would say of previous theorems of performance compensation.
> All this to show what we freaking already know trivially. > Business are not complete :D. There are other business ideas that are "orthogonal" to what any business does.
sounds like you live on different planet. If everybody "knew trivially" that Taleb would be a celebrity way BEFORE 2008, not after. Crisis wouldn't happen. seriously - any day you open browser and find BS like this AGAIN and AGAIN "http://online.wsj.com/article/SB1000087239639044372020457800...
"OMG! 75% of start up fail!!! Bad VCs hiding their dead! Be smart - invest in good VCs! They have the right system!", "Use this smart hedge fund - that the new system, not that hedge fund, they had bad system!". Biz word is absolutely chockfull of idiots looking for a "system", looking for mechanized recipe to follow to find "success".
But all this is somewhat beside the point, since knowing a business is a formal system to which Godels theorem applies does not let you figure out on its own if it is going to make it and if you therefore should invest in it. If anything you are arguing all businesses are formal Godelian systems therefore that can't be used as a distinguishing feature.
> What is that magic wellspring of value that year after year creates these huge opportunities out of nowhere?
Complexity is unidirectional. People move on, rationale is lost, and increasingly the bulk of institutional knowledge grows to be dogma. We do it this way because we've always done it this way.
Additionally, head counts don't naturally go down, people don't automate themselves out of a job and neither do they take kindly to being automated. Change becomes bad.
> Lets imagine, for a second, a time traveler from the future appears in front of Bill Gates in 1999 and utters something like: “I’m from the future! You should sell keywords to match search queries!”"
I am sure there were people within Microsoft who said that very thing around 1999. Also, Google employees lobbied for a social networking product to Larry & Sergey in 2003 and the result was Orkut.
The reason Microsoft didn't rise up to the occasion and grow to dominate search and Google did not seize the social opportunity the first go around is not because these things were unknowable to them under the formalized systems that governed their businesses at the time as the author suggests.
It is not a lack of information. It is more Kafkaesque. Employees had the knowledge and couldn't freely execute on it -- organizational complexity kills joy.
The advantages of working for a large corporation evaporate when you want to work on a risky new idea outside of the company wheelhouse. If anything, they becomes disadvantages.
You are just as likely to succeed independently, have more fun doing it, and keep much more of the windfall. So why not jump ship?
i think its good example of not "good enough" formalization. Afaik Orkut had a lot of operational freedom to do what he wanted with Orkut. it just nobody (the whole pool being semi-dead friendster at that point) had the answers what is actually important to build a socnet. Today we have 20/20 hindsight its absolutely terrible to have runaway adoption in 3rd world country that doesn't speak english... yet thats one of the axioms we learned "the hard way". And obviously that increased formalization (whats working, whats not) was helping Mark a lot when doing next iteration.
What I'm trying to say is that Orkut was less passonate about it and had his creativity and motivation stifled compared to Mark.
Mark obsessed over his project and had everything to gain. Orkut on the other hand - What advantage was there for him in building his project from within Google? None. And not much to gain, just a promotion.
If he really had his heart in it why not go it alone? He didn't, so he did not.
You can be very much obsessed even in building something within Google. Obsession and passion are not measured in money.
By your logic the creators of Unix weren't obsessed with it, because "what did they have to gain (from AT&T)? Just a promotion".
I'd say being passionate in the specific thing we're discussing (Facebook vs Orkut) doesn't mean a thing. What exactly that more advanced Mark HAD to show for his passion? Yes, he has the far far better business now, but where was the passion factor at play in making a better offering THEN?
What I see is business opportunities and mistakes in Orkut vs Facebook.
Not any engineering (or other) passion that resulted in a visibly better product.
Yet thats kinda the key point of the essay: the act of knowing how big/important social networking is requires somebody first to formalize that knowledge. imho Google was very good corporate host to let Orkut experiment so much as he did. It just wasn't enough.
Not so, Bell labs was a special place at the time and provided significant advantages (like access to computers for one). If the likes of Thompson and Ritchie came about 15 years later where they could develop Unix independently and the aforementioned advantages had melted away, why wouldn't they? At that point AT&T would offer nothing but constraints.
I call BS. People also do stuff for the mere fun of it, especially geeks.
For example Thompson and Ritchie had no problem developing Plan 9 for Lucent (AT & T) "15 years later" after the "advantages had melted away".
Not to mention that the example of advantage you provide ("access to computers for one") is insignificant in any case. Thompson and Ritchie could get work anywhere they wanted with computers available. At least the other commenter meant huge deals of money by "advantage", not standard job perks.
And thats exactly what big corporations are doing or dream about doing (depending on their competence) with their workforce. reading their employees handbooks or HR guides would make you feel they are 100% successful. Of course reality makes it much less deterministic, yet it is their ideal - having employees as wetware AIs repeatable and predictably implementing corporate policies. (In case of Foxconn that seems to be becoming a reality without all the inconvenience of wetware.) corporations are working that way since thats the way it maximizes their returns. Just showing the natural duality between free form exploration of new "true yet unprovable" statements and rigid late stage formalization that has to be understood by startup founders.
I am familiar with the incompleteness theorem but I didn't get quite to what the point was, although admittedly I did not perform a very close reading. However, I believe what's going on here is, and this is not meant to be overly critical, a fallacy of conflating that are incompatible. People do this with Godel's theorem all the time by the way; his theorem was very specific and yer it has been abused by all sorts of people, including postmodernists.
I don't think abuse is going on here. I think it is just an over analysis or something like that of startups. I think this could be an interesting for those unfamiliar with Godel and Turing, but then again there may be better articles ... I give a Gentleman's B. :)
These are good points, and some organization certainly become inefficient/chaotic with age. Yet do keep in mind there are brutally efficient corporations that are extremely good at extracting value just by repeating what they already do. thats bread and butter of F500. Going to extreme evil as example - how Halliburton can extract billions of dollars in what feels like an instant from a newly occupied country they never done business with? Thats power of their formalization. they know what contractors to hire, what officials to bribe, and all these 1001 things they need to do to "make spice flow" in no time.
My point is to contrast that repeatable and scalable power of formalization of legacy companies to almost free form exploration of startup. You see startup kids doing totally insane things- trying to create processes and structures so they look like "big corporation", or failing to create them when they absolutely need to start formalizing whats working, etc.
>Additionally, head counts don't naturally go down, people don't automate themselves out of a job and neither do ?>they take kindly to being automated. Change becomes bad.
very good point.
>What I'm trying to say is that Orkut was less passonate about it and had his creativity and motivation stifled compared to Mark. >Mark obsessed over his project and had everything to gain. Orkut on the other hand - What advantage was there for him in building his project from within Google? None. And not much to gain, just a promotion.
thats pretty brutal on poor Orkut :) I would be careful here just to say we dont know how motivated or not he was at that time. From what i seen at that time he seem to be rolling out updates pretty quick, it was active and clearly passionate development. You might be right that if he tried "a little bit harder" he might have gotten better result. Yet just the same i'm not going to be surprised that all the minefields Orkut/Friendster/Myspace managed to blow up themselves on were critical insights (== discovery of new formal system) that helped Mark avoid these pitfalls. and of course that was also big part of explanation: http://www.youtube.com/watch?v=zVIhUVid4fA
>It is not a lack of information. It is more Kafkaesque. Employees had the knowledge and couldn't freely execute >on it -- organizational complexity kills joy.
That what i sort of argue against in that and previous essays. Employees are far too busy making sure that stuff they already got works. they joined 9-5 jobs because they like being in predictable field, with no surprises and defining their own success as just repeating what they doing over and over. JWZ said it best "we joined netscape to build a great company, those who came later wanted to join a great company". Any startup that plans to grow up past its humble beginnings needs to understand that duality and understand what motivates its people past first 100 employes. Its the whole difference between discovering new domains and formalizing them for revenue extraction.