Just as Oscar Wilde said it would take a heart of stone not to laugh at the death of Little Nell, so one cannot help giggling that a Chinese inventor with a small team and modest resources has by his open source DeepSeek AI system outsmarted the behemoth “geniuses” that came up with the $500 billion Stargate announcement. Large bureaucracies subtract value, government subtracts even more value and ivory-tower HR wokery subtracts most of all. Small scale entrepreneurialism, fiercely independent of conventional pieties, is necessary not merely to “beat China” but to produce Industrial Revolutions.Liang Wenfeng, the entrepreneur behind DeepSeek, is not technically starting in a garage; he is already a 40-year-old serial entrepreneur who has made a fortune applying AI techniques to quantitative investment, a field doubtless less crowded in the murky reaches of Chinese financial markets than in the U.S. He claims to have spent only $6 million training DeepSeek, a model that has only 671 billion parameters, likely far less than its equivalent U.S. competitors. Even if the real costs were substantially more than this, it is undeniable that his success has been produced by a small, highly skilled team, trained in China rather than the West.DeepSeek’s success has a number of lessons for the U.S. tech sector. Most of those lessons were obvious from economic first principles to those who know how innovation works and has worked ever since the dawn of the Industrial Revolution. Throwing money at a problem does not work; it merely results in bureaucracy and a gigantic waste of money. This is even more the case if the money is government money.The Manhattan Project may have succeeded in building an atomic bomb, but that was a military project of hugely negative economic value, and it was completed 80 years ago. The genuinely valuable advance made possible by Sir John Cockroft and Ernest Walton’s 1931 splitting of the atom was atomic power. Had the private sector been left to develop this, with entrepreneurs who had not been devastated by the Great Depression, the war and the deadening socialism of the New Deal, it would have succeeded in doing so just about as quickly as in real life, and at one hundredth of the cost.More recently, President Biden’s endless taxpayer financed subsidies to green tech and other government favored projects have produced almost nothing beyond payoffs to politically in-group constituencies. Trump’s $500 billion “Project Stargate,” even if the private sector provides that colossal sum, will achieve very little more, because at that size it will inevitably become a morass of competing bureaucracies. A government-funded mega-AI project would be even less productive of genuine advances; the Manhattan Project is not an example to be followed. The big advances in AI or anything else will not come from gigantically funded fashionable conventional thinkers, from government grants, or from grandiose boondoggle creators like Masayoshi Son, but from small teams of truly brilliant people, funded by some eccentric but wealthy individual, possibly as in Liang’s case one of the brilliant people themselves.Liang’s success makes us think again about the likely trajectory of China. While it is an authoritarian, centrally controlled state at the top, that central control does not appear to have hindered Liang, and it has made DeepSeek no more controlling of its content than Western large language models such as ChatGPT. You cannot discuss the Uighurs or Tiananmen Square on DeepSeek, but at least until the advent of President Trump, the wrongthink controls on ChatGPT were considerably broader than that. Of course, the hosted version of DeepSeek may give all your personal information to the People’s Liberation Army, to be used for dark and nefarious purposes, but that risk is not entirely absent from the equivalent western AI. Overall therefore, we have to admit that the Chinese government’s draconian but intermittent central control is not a complete barrier to genuine entrepreneurship.As for the EU, its proud proclamation of being a “regulatory superpower” and its desire for much tighter censorship means that it will never be a significant force in AI; the regulation removes the possibility of intelligence. Arthur Mensch, co-founder and CEO of French AI leader Mistral on the recent EU AI Act sums it up:
The intention of introducing a two-level regulation is virtuous. Its effect is catastrophic. As we understand it, introducing a threshold aims to create a free innovation space for small companies. Yet, it effectively solidifies the existence of two categories of companies: those with the right to scale, i.e., the incumbent that can afford to face heavy compliance requirements, and those that can’t because they lack an army of lawyers, i.e., the newcomers. This signals to everyone that only prominent existing actors can provide state-of-the-art solutions.
All regulation corrupts and stupefies, and absolute regulation, such as the EU seeks, corrupts and stupefies absolutely. If the universe of great inventions is restricted to those advances that can be conceived by wokies, very few inventions will be possible.In that context, the announcement by Oxford and Cambridge that they are to move to “open book” examinations to help minorities is in the age of AI deeply troubling. “Open book” essays will mostly be written by AI, since this is a great deal easier and less stressful for the students than composing them oneself. The new policy is a sharp shift away from Oxbridge’s meritocracy of the 20th century, back towards the 18th century system. Under that system, the principal object of the universities was to train their graduates for the Church (today’s equivalent being the Civil Service, positively Laudian and ear-severing in its enforcement of intellectual conformity) and very little work was done, the undergraduates being occupied in drinking port and chasing whores. Today the port would be drugs and the whores would be transgender, but otherwise Oxbridge is reverting to its ancestral roots, graduating favored social categories without excessive stress for jobs enforcing societal norms.The only difficulty will come in the Mathematics Tripos, which even in Isaac Newton’s day was intellectually rigorous. However, if a rigorous Mathematics Tripos were kept the authorities could reintroduce social preference by making the disfavored social group mathematicians “sizars” waiting at table on the favored social group, as Newton was forced to do, thus making appropriate if inadequate status reparation for centuries of slavery.The Chinese success has huge implications for immigration policy. No, the U.S. should not let in infinite numbers of Chinese students; the best ones can get perfectly good educations at home, as did Liang Wenfeng; the U.S. will merely get the politically connected drones and the PLA spies. The more important implication is to end the H1B program; by importing innumerable interchangeable cogs at below-market wages, it boosts only the giant bureaucracies and deters the best U.S. students from entering computer science studies, because the rewards on graduation have been rendered so mediocre.There are also implications for capital investment. In the old days, if you overspent on a railroad or a steel mill, and allowed its construction to be prolonged and bureaucratized, you would still end up with a railroad or steel mill, albeit late and overpriced. However, if bureaucracy (for example through environmental restrictions on the provision of adequate power supplies) slows down a gigantic AI project, you will have nothing useful, because your competitor’s superior AI project will come to market first and render your project obsolete. In this context, Liang Wenfeng has undoubtedly benefited from China’s gigantic supply of new and efficient coal-fired power stations, impossible in the West because of “climate change” hysteria about coal. Ditching environmental regulations is a far more important boost to Western AI capability than any amount of multi-billion-dollar government-inspired spending boondoggles.The West can be competitive with China, of course it can. China is a relatively poor country, saddled with a grossly inefficient and tyrannical Communist government. But to compete, the West must remove its own elements of central planning, such as H1B visas and burdensome environmental controls, and allow the free market to produce the abundant innovation of which it is capable. The best of those innovations will almost certainly come from eccentric small companies, flying “under the radar.”More By This Author:Emerging Markets In A Protectionist World
Government Debt Should Be Rated BBB At Best
The Bear’s Lair: Slavery Makes A Comeback
