Pinned post

Who I am 

I'm Tom. I live on Larrakia country in Darwin, Australia. It's very hot here. I work as a coder on diverse projects, with a special expertise in "spatial technology".

Currently, I'm working on cattle station planning, geochemistry survey, and bushfire mapping software for different clients.

I have cathexes in left wing politics that are very important to me. In my forties, my main sidelines have been reading Theory and helping program Darwin's famous Deckchair Cinema.

Tom boosted

Not generally into "look at this amazing thing we've discovered is hidden inside an AI" type articles, but this is super interesting - using ChatGPT to emulate another simpler computer, connected to the internet: engraved.blog/building-a-virtu

Tom boosted

Actually, it’s a common misconception that it’s a common misconception - almost everyone knows, but thinks that almost no one knows.

Tom boosted
A cryptobro DMed me to ask if I wanted to join his mailing list.

I spent slightly too long making this to send in reply.
Tom boosted

Quote-dunking is frowned upon here, and with good reason, but I love it when @DavidHarvey gets snarky.

(Lifted lock, stock, barrel, and alt text from @inquiline's feed on the hell site).

Dynamicland, Bret Victor, spreadsheet programming 

Since it's a bit hard to navigate from the website to other stuff about the project, here are a couple of follow-up links:

Andrés Cuervo describing projects in the Dynamicland community space:

cwervo.com/projects/dynamiclan

Transcript of an Afrofuturist Podcast interview with Victor on why the technical backbone (Realtalk, a variant of Lua) is not open source:

gist.github.com/shaunlebron/4b

Show thread

Dynamicland, Bret Victor, spreadsheet programming 

Bret Victor, almost any coder's "rather inspiring role model", is working on a fascinating project called .

Not sure how to describe it, but "accessible, collaborative materialised programming" might be a start.

I'd guess it has stuff in common with other projects like and , but its take on peripherals and interaction is fascinating …

cabinetmagazine.org/kiosk/kan-

dynamicland.org/

Image alt text, accessibility, add-ons 

Some further discussion on related features upstream in Mastodon itself:

github.com/mastodon/mastodon/i

Show thread

Image alt text, accessibility, add-ons 

(This userscript makes me notice all the missing alt text, so I end up periodically reposting it for new fediverse users …)

The script for the Mastodon web client described in the linked post below highlights images in your feed which don't have alt text.

assemblag.es/@attentive/109319

User prompting for alt text on media posts is an open issue for as lodged here on GitHub:

github.com/hometown-fork/homet

Tom boosted

After a public consultation, the people have spoken ;-) it seems there's people still interested in continuing the experiment to find ways to asynchronously read with a longer time window, until February 28th.

If anyone has ideas or suggestions on how to frame the reading better (say, focusing on a certain chapter, bringing up an example to make it more relevant, etc.) please share!

Don't forget to use the hashtag so we can coordinate!

Show thread
Tom boosted

This rolling pin from the Didion estate…. This « Maybe broccoli doesn’t like you either » apron….

Tom boosted

Reverse Turing tests; replication versus discrimination 

Fine example of a case where platforms we depend on (web search) will have to evolve to automatically filter ubiquitous troves of AI-spoofed content in order for their usefulness not to decline. Will it even be possible?

woof.group/@aphyr/109458338393

Show thread

AI-generated office horror notionally after the style of Clark Ashton Smith's weird fiction 

I'm not sure how interesting the capabilities of this platform actually are, but I guess it's fun for a day.

Earlier I was trying unsuccessfully to get it to produce a pastiche of Adrienne Riche's poetry, but it seems to do better at 20s–30s style weird fiction.

Reverse Turing tests; replication versus discrimination 

This user's observation, though reasonable, strikes me as afflicted by the "is-ought problem".

We haven't got systems that imitate humans with no understanding due to the way the Turing test was framed. We've got them because researchers have accessed their bulk training data from the world-historical change we call the Internet, and because it's seemingly far easier to build such systems than an elusive AGI.

mastodon.social/@intelwire/109

Show thread

Reverse Turing tests; replication versus discrimination 

Machine replication versus machine discrimination shapes up as a non-human contest in which a kernel of human distinction ambiguously inheres.

How reliably can a hypothetical machine administer a "reverse Turing test" which discriminates through interaction between human subjects and machine replicas?

The human labour of the conventional Turing test seems likely to be untenable soon.

Tom boosted

@attentive
Will the tests of the future rate us against the median of machine intelligence?

Charles Atlas ads; marketing masculinity in 2022 

Log onto almost any website these days and I'm bombarded with ads that might as well be Atlas's "chump to champ" campaign.

On Twitter some dweeb called "the Critical Drinker" suggests James Cameron has emasculated himself in his late career by eating too much soy.

On Facebook, "Reels" posit weird "gym crush" scenarios where a short man gets the girl by putting lifts in his shoes.

I believe this stuff works on people, but I truly wish it'd stop …

Birdsite; sf-horror Midjourney image; regrettable mention of Elon Musk 

Me logging onto Twitter on a Monday morning to witness more Musk-ology

(Original source: this thread twitter.com/hankgreen/status/1)

The predicament arguably recalls Deleuze's arguments about the "dogmatic image of thought"—on many levels. A system trained on prior networked human and machine-generated text cannot readily sense what, owing to never having been named, remains imperceptible.

Processing "training data" means algorithmically organising terms that always already try to name the world. How can such a paradigm intuit new names? How can it facilitate the creation of concepts?

Show thread

As people point out all the time, technology based on networks is *not* sentient, nor does it possess general intelligence. That's why its successes raise such bleakly hilarious questions.

Another example: if push-button high school essay writing ruins prior assessment protocols, doesn't that suggest limits of language together with rote expectations of formal education … have meant these protocols have not been assessing whatever is salient (or "intelligent")?

Show thread
Show older
Assemblag.es

assemblag.es is a Mastodon instance for people interested in thinking creatively and critically about technology, in the broadest sense. The only requirement is that you keep to the code of conduct! Note: we've paused new signups for a few days while we deal with the #twittermigration influx - please do request an invite, but be aware that for now functions more like a 'waiting list' than a knock on the door, until our numbers stabilise again :-)