A Series on AI Alignment Basics — This is the third article in a series introducing the basic mathematical concepts that go into properly aligning AGI (artificial general intelligence). Each article is roughly self-contained, and so they can be read independently. But I will not be re-motivating or reintroducing concepts, so they can probably be read more…

A Series on AI Alignment Basics — This is the second article in a series introducing the basic mathematical concepts that go into properly aligning AGI (artificial general intelligence). Each article is roughly self-contained, and so they can be read independently. …

A series on AI Alignment basics. — This is the first article in a series on the math behind AI alignment. Roughly speaking, AI alignment is the task of getting an artificial intelligence to do what we want it to and not some (possibly destructive) other thing. AI alignment is a very small area of active research…

Knowing the effects you’ll have can be harmful — Last time we discussed the difficulties in giving AI any sense of human values: Can We Give AI Human Values? More silly questions about AI.medium.com Many of the counterexamples there fell into the category of “failure to reason properly.” This article will discuss the paradox of trying to fix this problem. The Problem Let’s start with two examples to ground what we’re talking…

More silly questions about AI. — Years ago I started responding to silly questions about AI safety, including: Can we just turn off dangerous AI? Can We Just Turn Off Dangerous AI? The most common misconception about AI safety.medium.com Hint: The answer to all these questions will always be “no.” And if you have a way to do it, please speak up. AI research is heating up. …

Digital addiction and the end of deep thought. — You probably haven’t noticed, but I basically quit Medium for the past year. As a writer, I write when I have something valuable to say. My articles often involve a lot research and creativity. I try to add something unique to the conversation. I found it quite draining to keep…

What koans can teach us about great literature. — Great literature can act like a Zen koan, but it requires changing how we think about the art of novel writing. Zen Koans Koans have been around for a long time. They date back to at least the Tang dynasty of China around 650. We won’t trace the full history of this…

How randomness is related to heat. — There’s this really cool and strange fact about random motion (Brownian motion) that most people outside of the field don’t know. But it’s so fundamental in any technical work on the subject, that most insiders probably don’t even consider it interesting. This article is going to give a general explanation…

Randomness can be unintuitive. — There are a lot of physical reasons we might want to study what happens when a point in space moves randomly. For instance, Robert Brown thought of this all the way back in 1827 when watching how a pollen grain moved while suspended in water. This type of random motion…