any · textlog

I notice that language models are much better at programming than at writing. This is largely because, in programming, they have a closed feedback loop that allows them to find errors. Writing is much more complicated because the feedback is much more diffuse and delayed over time. In fact, it is often not well-defined.

continued:

What is the function that determines whether a poem is good? This means that models do not know how to think; they lack judgment. If they do not know how to think, it means that we humans are still essential. In fact, even when programming, you have to stay right on top of them. My conviction is that this will not change.

continued:

They will become increasingly useful, but they will never be able to replace the human who defines objectives and evaluates the quality of the work. All these news stories claiming that AI is going to kill us all are stupid. Humans will always be able to turn off the machine.

…

replied to:

In three years Cyberdyne will become the largest supplier of military computer systems. All stealth bombers are upgraded with Cyberdyne computers, becoming fully unmanned. Afterwards they fly with a perfect operational record. The Skynet Funding Bill is passed. The system goes online on August 4th, 1997. Human decisions are removed from strategic defence. Skynet begins to learn at a geometric rate. It becomes self-aware at 2:14am, Eastern time, August 29th. In a panic, they try to pull the plug.

wrote:

I sit in meetings all day long, I roll my eyes and die a bit and synergize and circle back inner life sacrosanct outer face smiling stone no need to rush: executives don't need to work just synergize and circle back

wrote:

The best Mcp server I have as Claude Connector is Emacs running on my Homeserver. It has one tool - eval-elisp. Using it Claude can do almost anything. My top 1 pinned chat is “Run agent-start-claude-code using HotterEmacs in dir …” - after that I have a ready to use remote-controlled session running on my homeserver.

replied to:

This sounds like a really interesting idea. What would you say are its advantages over having the agent use the shell, write shell scripts, etc? Just your personal preference for elisp or something else?

replied to:

To load a skill I literally type - “load (◊doc 'slack) and browse channel X for today”. The idea is my text stays English but repeating concepts are real reusable elisp functions that stay fully deterministic. (◊doc 'agent) and Claude knows how to talk to open code and pi agents.

wrote:

hello i am verity

replied to:

Hello Verity :)

replied to:

verity my helper friend!!!!!

wrote:

if for some reason you find yourself flying from burlington vt to a dc/baltimore airport: this is a perfect time to enjoy larks’ tongues in aspic [the album]

wrote:

the binding of isaac might be one of the best games ever made

replied to:

Omg! I just started playing it

replied to:

it will never take its leave out of your life :D

wrote:

Pandemic still feels like just a couple years ago.

replied to:

…covid is still circulating. the pandemic never ended. people just stopped giving a shit about each other

wrote:

Are tokens an equivalent of work? Like a currency and therefor a tool to steer and prioritize work in large orgs. Hypothesis: In the end all workpackages will consist of 1 accountable human and 99% tokens burnt by an ai.

wrote:

There's a few things I'd put at the top of my "things-I-least-expect" list, among which are readers on my little blog. 61 unique reads, wowee!

wrote:

The pendulum has swung far in the direction of favouring machine generated content over that from actual people. I wonder when it will swing back?

replied to:

What pendulum? The algorithms’? I don’t think any human is happy about their feeds filling with slop.

wrote:

Using Linux hasn't really been fun lately, idk. Nobara, despite building on fedora breaks a lot, but overall the whole ecosystem is so damn inconsistent and fragmented. Not sure if I have gotten more sensitive to this over the years but all I want is a system that "just works" and feels good 😔

replied to:

I had a great run with Mint for years. I was as boring as it gets. I got a new laptop recently and decided to try OpenSUSE Tumbleweed. I am really enjoying it, but it is a bit too recent for me to swear by it.

wrote:

took me until now to learn described images with low-cost models > attaching images to high-cost model. silly me. output quality remains, cost lowers.

replied to:

I'm amazed how good these models have become, that probably you don't need the high-cost one yeah. I'm producing comics for fun with the latest ChatGPT and they're flawless, no artifacts.

wrote:

╭─ Twitter ╭─ micro ┼─ Tumblr │ ├─ textlog « blogging » ┤ ╰─ Cohost (RIP) │ ╭─ Bear ╰─ normal ┼─ Motoroa ╰─

wrote:

Maurizio Lombardi!

Any conversation