In either case, you're ranking people based on WHO they know, not what they know. Worst kind of ranking possible (it keeps competent outsiders out).
In politics, the organized faction always trumps the general interest. Say Faction A is 1% of the population. The faction wants to preserve a barrier to entry that costs every voter $100 and nets each faction member $10,000. The members of Faction A will be single issue voters over the barrier to entry. But to the rest of the public, the issue is lost in the noise. Average Joe does not even realize how much the barrier to entry costs him, because no interest group will fund ads to tell him. Thus the politician always supports the faction in order to prevent the 2% swing. ( I've spent time in both the state house and Congress, and this is exactly how the decision making process works).
The life cycle of any government policy ends up being the following: 1) Program is created to benefit the public good 2) The program inevitably directly benefits a small group - employees of the agency, contractors, etc. 3) The beneficiaries become dependent on the program. They organize in order to protect themselves politically. 4) In any policy battle where the interests of the beneficiaries collide with the public interest, the beneficiaries generally win. 5) As the beneficiaries win battles over time, the program ends up serving the beneficiaries at the expense of the public.
This is basically the life cycle of every program that has ever come out of Washington. The once mighty NASA now only exists to provide employment to NASA engineers. The school system exists mainly to provide employment to teachers. The AMA exists to erect barrier to entries to the medical profession. Etc, etc.
Your algorithm proposal is amusing because it is basically the same solution to the problem of corruption/factions that the progressives tried during the 20th century. The idea was that "scientific policy" based on algorithms and formulas, designed by academics and civil servants, could replace control by politicians. Needless to say, it was a miserable failure. A classic example of policy via algorithm is the creation of Nationally Recognized Statistical Rating Organization's who used algorithms to dictate which loans regulated funds were allowed to invest in. It did not take too long for Wall St. to figure out a way to structure the riskiest loans imaginable in ways that would pass through the algorithm. The rating agencies are a high profile example, but there are hundreds of other similar failures. In practice "scientific public policy" ends up being the worst of both worlds. It combines the downsides of algorithms (lack of intervening personal judgment when the algorithm gets gamed) with the downsides of politics (corruption/factional power struggles).
And no, it works terribly for universities ( or rather, it works great for the universities, terribly for the public). For example, the number of medical schools in the country has actual gone from 162 in 1906, to 129 today, despite a three-fold increase in population. The artificial barrier to entry created by accreditation constricts the supply of medical schools, resulting in higher tuition and higher medical costs.
Or consider law schools. The cheapest law school in Massachusetts costs 38K per year for tuition. If you wanted to create new a law school as a startup, think you could create one just as good for much cheaper? I bet you could. It's not like law schools need expensive equipment. But you're not allowed to, thanks to accreditation laws.
Accreditation started off innocently enough - it was just a good housekeeping seal of approval. But then the professional organizations managed to make it illegal to practice at all without a degree from the accredited institution. The bar associations lobbied for laws that made it impossible to practice law if you simply studied on your own and passed the bar, or if you went through a correspondence course. In 2008, Abe Lincoln is out of luck. Now he has to take on 150K of debt in order to practice law.
History shows the slow is indeed slippery. What if your startup accreditation body follows the path the law schools, and in 30 years ends up lobbying Congress to ban all non-accredited startups?
Currently, there are zero investors whose money gets you you a visa. Expanding from zero to an unnecessarily small group would still be an improvement, even it if wasn't optimal.
And even if school accreditation were a net lose, that doesn't mean accreditation in general is always a mistake-- that it's impossible to do right. Whenever you try to build any kind of filter, people will try to game it, but that doesn't make filters pointless.
Recently I was talking to friend who had just left law school. He described the process as: "We spend all semester in class where we listen to boring professors who just like to hear themselves talk. Then we do internships where we get paid to learn." I have close friends in architecture school who say the same thing. Almost all the practical stuff about code and structures students learn on the job. The school teaches stuff like "aura" and deconstructionist design. Compare the current Boston City Hall the Old Boston City Hall and you see the results of this wonderful, accredited training. There is simply no evidence that law school or architecture school is a net positive, much less that is worth the cost. (Medical school is a tougher case as care has improved. But this seems to be the result of improvements in science and technology, not schooling).
I hear the same exact stories from people graduating other schools - education schools, meteorology programs, business school, etc. It's pretty clear that the primary result of establishing an accreditation body as legal gatekeeper to a field, is to create jobs for professors and increase wages for professionals. ( note - computer science classes tend to be more sound, perhaps precisely because there is no legal credentialing requirement to enter software programming, and thus the departments must prove their worth the old fashioned way)
If the purpose of credentialing laws was merely to protect consumers, the laws would only care about the results of the schools teaching ( whether graduates could past the Boards or the Bar ) rather than the process ( whether the school was accredited). The law would also allow people to get around the defaults. For instance, your web browser warns you when you access a potentially malicious site, but it still allows you to access it if you really want to. But you cannot do the same with lawyers; you a non-accredited lawyer cannot represent you in court.
And even if school accreditation were a net lose, that doesn't mean accreditation in general is always a mistake-- that it's impossible to do right.
In the context of the American political system of 2008, it is impossible. Washington does not work. You can expect to make a reasonable policy suggestion, and have the actual result come be close to what you wanted. The only thing we can do is keep the poison from spreading through Silicon Valley as long as we can.
Accreditation for colleges is much more, well, collegial. There's no big motivation to restrict supply. It's hard to think of any really good educational institutions that can't seem to get accredited for political reasons. The list of over 1000 accredited colleges in CA starts off:
* A-Technical College (LA)
* ABC Adult School (Cerritos)
* Abram Friedman Occupational Center (LA)
* Absolute Safety Training Paramedic Program (Oroville)
* Academy Education Services (Oxnard)
so it's not just the Stanfords that can manage accreditation.
http://www.ice.gov/pi/news/newsreleases/articles/080409losan...
It's hard to think of any really good educational institutions that can't seem to get accredited for political reasons.
Three years of on the job training is better education than three years of schooling in practically every field. Plus you get paid rather than have to pay. Yet on the job training does not count as education that allows you to legally be a lawyer, doctor, teacher, parole officer, military officer, etc.
The problem with government sanctioned accreditation is not that the specific decisions about which schools meet the requirements get politicized. The problem is that accreditation allows the intuitions who reap the tuition money to define what education itself means.
In CS, this is mostly false. You learn very different things in school and in industry. The former is useful in the latter, but usually not vice versa.
The most interesting thing about it, to me, was that the engineers from the private sector just couldn't seem to understand the political constraints that the government managers work under.
I've only ever worked at small to medium-sized private-sector software companies. It's fascinating to get a look into public institutions and the way civil-servants make decisions.
As a hacker, there's a real danger of only considering engineering constraints, and dismissing organizational or political constraints on a problem or program. I often see bright techies make suggestions that are just naive from a political or organizational perspective. Suggesting a plan that would work great "if only the government weren't so clueless or corrupt" is no plan at all, if you actually have to work with the government.
This is not a direct criticism of pg's essay. I think he's on to a cool idea.
The discussions I've had on similar topics in the past, though, have been really eye-opening. I may be hot-shit technically, and able to hold my own in a meeting, but there's this whole other world of bureaucracy that I must admit I know almost nothing about.
P.S. I found the NIH parts of Darwin's Radio fascinating. The book doesn't have much action. In fact, a lot of the character interaction is between government scientists, and happens in meetings. Nonetheless, I felt as though it gave me an insight into this other world. It was fascinating.
This is a good idea because the expected outcome of adding 10,000 new people to the "pool of workers" by existing companies would be to see an increase in unemployment claims (even though not all unemployed people would file).
So by adding 10,000 Founder Visas to the pool of workers and then tracking the rate of decrease (or increase) of unemployment, a metric of immigrant-based job-creation startup foo could probably be determined.