Ben Lund

[Ben Lund @benlund]

@Catholicizm1 It’s about making peace with your passions, and thereby gaining mastery over them.

[Ben Lund @benlund]

But the second is that SPCX will be ONLY 0.1% of the S&P500 index! (Because it’s only floating 5% of its shares.)

That's the size of the opportunity and hence the size of the supposed problem!

[Ben Lund @benlund]

On the inclusion of SpaceX in the S&P500

[Ben Lund @benlund]

However, there are two MAJOR flaws in this plan.

The first is that you’ll have to guess the fair market value of SPCX better than all the people who will be buying shares in the IPO. If you can’t then you’ll lose your money.

[Ben Lund @benlund]

P.S. And of course after you've destroyed the blue buttons you have to find and drive out the guy who invented the buttons in the first place.

[Ben Lund @benlund]

On the Red / Blue Button Question

The question bears a superficial similarity to problems of living in society, so it triggers our communal instincts. But it’s actually hijacking that instinct, as you can see via a simple reformulation:

[Ben Lund @benlund]

What button should you press? It’s totally obvious. Everyone should immediately press their red button, followed swiftly by a vigorous campaign to find and destroy all blue buttons.

[Ben Lund @benlund]

This is tongue-in-cheek in case it's not obvious. To all my friends working and using AI - please don't let this happen.

[Ben Lund @benlund]

AI strategy becoming clear. Patience is all you need. A year or two. Everyone else a burned-out low-attention-span absent-minded wreck. You're the most competent person still standing. Wait long enough you can probably just walk into any org in the world and declare yourself king

[Ben Lund @benlund]

Basic thesis: AI is divination for rationalists

[Ben Lund @benlund]

"What can the most advanced form of Artificial Intelligence have to do with the most ancient form of superstition?"

[Ben Lund @benlund]

New essay on AI and the I Ching: https://t.co/yWJJDibnXz

[Ben Lund @benlund]

6) Therefore, ASI from LLMs is a logical impossibility.

Thread with fuller explanation here, along with common objections: https://t.co/Bbc2RGCeFD

[Ben Lund @benlund]

Here is a simple first-principles argument for being skeptical about Artificial Superintelligence:

1: Given any rule or function which can be endlessly applied, it is possible to precisely and correctly derive any given set of its results.

[Ben Lund @benlund]

Today’s a good day to post this reminder:

Working from first principles, you can show that getting Artificial Superintelligence from LLMs is logically impossible.

Explanation:

[Ben Lund @benlund]

5) The term “superintelligence” surely means something that surpasses the human intelligence function (if it means anything).