@balajis's Latest X Threads

Investing in a bank is a high risk activity. Like buying PayPal stock. But merely depositing in a bank is *supposed* to be a low risk activity. Like using PayPal. Now, there is a school of thought that says that this isn’t the case. That any money you have in the fiat banking… t.co/iFWXl0PzZv

In a talk given four days ago, the FDIC chair noted (a) the potential for bank runs due to huge unrealized losses and (b) how “complex & challenging” it was to operate when “interest rates change to the extent they have.” In other words, the Fed did it. t.co/cZ8w1o2mHq t.co/kn80MCqTNL t.co/KqCWhaBYgq

The fundamental question is: aligned with whom? If AI is decentralized and open source, then everyone has access. In the limit we all have a personal AI, aligned with us, on our person at all times. Like a loyal smartphone dog, which depends on us for charging. t.co/4iiygmqVGw

One scenario is that the economic impact of AI on white collar workers is so substantial that you get enormous counter-AI techlash. At which point AI engineers may have to (a) leave hostile areas or (b) turn AI into a military thing to gain points with nationalists. t.co/jFkq33YPUI

Hungary doesn’t want to be a pawn in another geopolitical contest between east and west. That’s a tired role they’re all too familiar with. Instead they’ll connect to as many willing trade partners as possible. A kind of networked state, if you will… t.co/O3EKhZS30J

LLMs often generate factually incorrect instructions. Following such instructions in the physical world means running into an issue. And the longer the list of instructions, the more likely an issue eventually arises. So: plausible texts fail empirical tests.

College, company, and country admissions are already highly correlated. Many visas are student visas. Many more are based on company sponsorship. And yet more visas are awarded for founding or investing in companies. t.co/KQGLcKlV6T

Yesterday, your AI wasn’t enlightened enough, so it needed to be centralized. Today, your AI isn’t safe enough, so it needs to be centralized. And tomorrow, your AI won’t be patriotic enough, so it will need to be centralized.

This is the Linux move that’s proven to counterbalance Microsoft. t.co/l9rDB3cMF2

The fact that LLMs are bootstrapped from human language — and not from, say, sensor data — means they arguably are more human-like than a genuine alien. It’s trained on Reddit posts, not born on a remote planet.