Thomas Mayfield

@thegreatape
0 Followers
40 Following
93 Posts
Software mechanic, fitness nerd, SFF reader. Shaving yaks at @Bookbub
Bloghttps://thegreata.pe

Feel like I should move to a smaller instance. Nothing wrong with mastodon.social, but it's huge and is now a default server, so the local timeline is a firehose of kinda unrelated stuff. Having a slower local timeline that's related to your interests and worth browsing seems like one of the things that should make Mastodon interesting.

Bummer that you can't port all your post history when you move, though.

Outside of physical training and work, I haven’t done well recently at spending time in ways that match my values. Don’t quite have the shape of it yet, but these ideas feel like part of what is missing.

Two lifting concepts I'm figuring out how to apply to the rest of my life:

1. Autoregulation: one's capabilities will vary day to day, sometimes by quite a bit. A good program has systematic ways of assessing and adjusting what you do based on this.

2. Raising the floor: over time, not only raising the ceiling of what you can do on your best days, but raising the floor of what you can do on your worst.

Welp, looks like Mailbrew's bird site bot finally got axed. I'll miss it. A daily digest was a great way to keep up with a handful of friends over there and the best attempt I've seen at taming that medium.

I read a lot of series as they are published—it's a great way to help them get finished—and I'm really happy to see more authors putting series recaps at the start of later books. My poor memory needs all the help it can get.

(s/o particularly to @DjangoWexler , just started Emperor of Ruin)

Good morning from the dogs, who have gone back to bed
I don't begrudge anyone help clearing snow—especially today's horrible heavy wet slush—but man are snow days more peaceful without the drone of snowblowers.
At this point "we will never cram generative AI into our product" is a feature I will pay for
This has mostly reinforced my take that you should worry about an LLM taking your job only if you believe that you are, at a deep level, just a fancy autocomplete with a feedback mechanism.

Long read, but great deep dive into how LLMs like ChatGPT work: https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-why-does-it-work/

Kinda striking how many of the engineering choices are "idk why but it seems to work better like this".

What Is ChatGPT Doing … and Why Does It Work?

Stephen Wolfram explores the broader picture of what's going on inside ChatGPT and why it produces meaningful text. Discusses models, training neural nets, embeddings, tokens, transformers, language syntax.