Korva is a scroll of the internet’s best ideas.
Back the person who thinks differently
The power law
Investing runs on a power law: the best investment outperforms all the others put together, and the second best outperforms everything after it. AI research looks the same way. The high-risk bets are fine as long as the ones taken are super valuable if they work.
Founders who copy
A founder with a slightly different take on the same idea as the last thousand is a thin veneer. Many try to emulate being different after hearing Peter Thiel enough times, but they do not mean it. In 2015 there were very few AGI efforts, DeepMind and one or two others, while many thousands of founders were starting photo sharing apps. Today a lot of people want to start AI labs, and only a handful are doing something completely new that wasn't possible until AI got this good.
Researchers with conviction
Researchers divide the same way. Many chase whatever the last thing was that worked, and a small number have high conviction in a new idea. Someone who is wrong may be wrong in a big way, but if right, they will be really right.
How does a power law work in venture capital?
Venture returns are heavily skewed: a few investments produce nearly all the profit, and most produce little or nothing. Peter Thiel wrote in his book Zero to One that the best investment in a successful fund will equal or outperform the entire rest of the fund combined. Because of this, a venture investor is paid for finding one extraordinary outcome rather than many decent ones. The skew explains why unusual bets that could be huge matter more than safe ones.
Why did early AGI efforts stay so rare?
Through most of the 2000s, general intelligence was treated as a fringe goal by mainstream AI researchers. DeepMind was founded in London in 2010 by Demis Hassabis, Shane Legg, and Mustafa Suleyman with the explicit aim of building general intelligence. Google acquired it in 2014. OpenAI followed at the end of 2015, which made it one of only a few organizations with that stated goal.
Why do researchers chase the last thing that worked?
Academic careers reward papers that extend recent results, which are easier to publish than bets on unproven ideas. Thomas Kuhn described this in his 1962 book The Structure of Scientific Revolutions. He called the steady, incremental work within an accepted framework normal science. Breakthroughs that change the framework itself, which he called paradigm shifts, come far less often.
More from David Senra
