Ugh, Tommy keeps talking about AI

Hank Green's recent video The AI Question that No AI Person Asks had such a good thought about application of AI as "intelligence" vs "wisdom". It's under 9 minutes long, but you're a fast reader, and here's my favorite bits:

... power in the physics sense is the ability to basically manipulate atoms.

(Heat them up, move them around, put them here or there, etc.)

... lately, I've been thinking of intelligence as the ability to manipulate information. Mix [it] up. Make new things out of it. Push some over here. Bring a bunch of it to the top of a mountain and throw it off.

We have way more ability to manipulate information and to have it do stuff for us than we previously had. That's very powerful.

But intelligence is separate from what I'm just going to go ahead and call wisdom. And I don't think that we have a way to mass-produce wisdom.

So perhaps we have been moving -- throughout my lifetime -- from a world that was intelligence constrained to one that is wisdom constrained.

... intelligence would help you get what you want, whereas wisdom would help you want what you should want or the right things.

It's the ability to figure out which problems are worth solving and then to solve them in ways that don't create worse problems in the process, which is not easy.

... ... designing a more effective slot machine is an application of intelligence.

I don't think that you would call it an application of wisdom.

(Paraphrased; Emphasis mine).

---

If you're interested in longer form, more technical discussion, Casey Muratori has been doing a series (with expert Demetri Spanos) called Wading Through AI (YouTube Playlist). Latest one is on ethics. Each episode kind of stands alone, so don't worry about missing out.

AI vocabulary words

(Still trying this new thing. Tell me if you love it or hate it)

• Training Data Training data is the "raw material" AI "learns" from. Text, images, audio, or other information that's fed into a system during its development. E.g., last week's "thousands of labelled images of cats and dogs".
• Model A "model" is the trained system that does the actual "thinking". (There's no thinking, btw 😉). The end result that emerges after the previous steps of "machine learning" on "training data". This is why you might hear that some models are better at "code" or "reasoning" or specific jobs; or that some are cheaper or faster at certain tasks -- like identifying cats vs dogs in photos (from our examples). When you hear "a new model was released" or see names like GPT-5.5, Claude 4.7, etc, they're referring to specific models.

Music

I'm seeing Cory Wong again! (I promise I go to other shows).

It can't possibly do justice to the feeling in the room, but here's a 1h39m full show of Cory Wong, live in London, from 2023. 👨‍🍳

---

I had a fun chat with a friend about some of my own music. Specifically Cuatro Cuatro Cero. I still haven't found a one-sentence "pitch" (no pun intended?!) for that project, or what its purpose is. Open to suggestions!

Home

• E is playing soccer on a team. She enjoys it. We scream a lot for her to watch the ball 😅 😬 The lord is working on my patience, in some way. Haha! The coaches enjoy her, and she's getting to do "team stuff". So far, so good!
• Toby (the new pup) gained about 1.5 pounds since the last email, but still fits under the love seat. Just slows down a bit on the way under. 😅

Replies

J! Thank you for sending me the Veritasium video on Extreme Ultraviolet (EUV) lithography machines. I kinda fell off on Veritasium for a bit, but I do love the nerdy deep dives. Hadn't heard of the other channel, but I'll check it out soon!

Something practical

Alton Brown telling you how to make the perfect baked potatoes. Less than 2 minutes long!

👋🏼 See you at Church?