Toots for naught101@mastodon.social account

Written by naught101 on 2025-01-30 at 13:32

Is there #AI/#ML work focused on extracting meaning from text?

My understanding is LLMs don't do this, they just map text into a numerical space, and use that to generate new text (e.g. there's no text <-> meaning mapping, except in so far as the numerical space approximates meaning, i.e. not much). Hence LLMs generate nonsense that sounds correct, maths that's wrong.

Wondering if there are models that include components with explicit focus on #meaning (I don't know what this would look like)

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-30 at 07:34

Is there a data structure that can sensibly handle multiple hierarchical classification systems?

e.g. an Orange, in terms of phylogeny is

Plantae->Eudicot->...->Citrus->sinensis

and in terms of usefulness, is

Thing->Food->fruit->orange

(and it could have multiple parents in this taxonomy, e.g. cleaning product)

Bonus points for cool visualisations of this kind information.

[#]data #dataScience #dataStructure #information #hierarchy #taxonomy #classification #visualisation #dataViz

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-30 at 02:11

I would also like a better recommender-engine style feed, that takes into account my interactions with other users, and then shows me what posts they have been interacting with, normalised by post popularity - e.g. if there is an unpopular post that my neighbours have been interacting with a lot, show me that.

(obviously this one would require opt-in)

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-30 at 02:09

I would really like a customisable feed algorithm that lets me do things like:

e.g. I want EVERY climate science post to show up, and I'd like to see photography posts sometimes, but only if they're really good.

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-30 at 02:02

There are a few simple algorithms already proposed that would be really useful and match the criteria above, e.g.

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-30 at 01:59

I would love it if Mastodon had a few more algorithms. The key criteria should include:

there might be others too

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-30 at 01:56

Mastodon already has a few algorithms:

the Home feed is:

The Explore feed is something like:

(I don't know the specifics, just guessing)

Plus the ones for the local and global live feeds, and hashtags and lists. And "Who to follow"

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-30 at 01:50

Algorithms aren't bad. Algorithms are just tools. They do what they are designed to do*.

The problem isn't algorithms, the problem is values. The problem is optimisation for profit.

We could choose to create algorithms to optimise for any other thing we value, like increasing understanding, reducing conflict, or whatever we like. It wouldn't be profitable, but... that's good?

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-29 at 09:29

Wow, there's way more apps connecting to the fediverse than I thought.. This list is really great, thanks mostly to @smallcircles

https://codeberg.org/fediverse/delightful-fediverse-apps

[#]fediverse #federation #activitypub

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-29 at 04:01

@mastometrics Feedback on 'People' page:

It would be really cool to have something that shows a distribution of the servers of my followers/the accounts I'm following. It would be a really good measure of the usefulness of federation.

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-23 at 09:46

Are there any fediverse projects focussed on bioregionalism?

Thinking: bioregions are layered interactions between geomorphological, ecological, and social regions.

It would be cool to have a federated social network where posts can be marked as related to specific bioregions (perhaps multiple), and would only/primarily appear to people who are associated with those bioregions in some way.

[#]fediverse #bioregionalism

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-18 at 08:06

@adrigen Some discussion over here: https://lemmy.ml/post/3970377

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-18 at 04:33

Does anything like a federated e-commerce/shopping platform exist?

Or possibly a federated library platform, for distributed borrowing?

Been chatting with @adrigen about his idea for a local-but-federated gift economy platform, and we were wondering if anyone has already started on something like this..

[#]fediverse #federation #library #ToolLibrary #giftEconomy #ebay #eCommerce

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-16 at 01:44

Fuck yeah

‘Solar Mamas empower our people by giving them electricity’: the women lighting up Zanzibar

By Carlos Mureithi in the Guardian 15 Jan 2025

https://www.theguardian.com/world/2025/jan/15/solar-mamas-empower-our-people-by-giving-them-electricity-the-women-lighting-up-zanzibar

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-16 at 01:04

It appears that I can follow accounts mutually from:

but I can't follow my pixelfed account from friendica (but the other way seems to work).

Is this right? should the last one work?

[#]fediverse #mastodon #friendica #pixelfed

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-16 at 00:28

If anyone's on Friendica and wants more buddies, feel free to add me:

https://friendica.world/profile/naught101

@naught101@friendica.world

[#]friendica

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-14 at 12:24

I hate apps, and will always use the web interface if I can, but I just installed @pixelfed purely to annoy Meta.

Also maybe it has the potential to be useful as an IRL friends network, which mastodon isn't for me.

=> More informations about this toot | View the thread

Written by naught101 on 2025-01-03 at 12:19

Someone: posts good content

Mastodon users: boost enough that content gets on to /explore

2 random users: start a fight about a side-issue from the first reply and continue pointlessly for 47 consecutive toots, relegating all actually interesting comments to the bottom of a long page of garbage, where they will never be seen.

Sigh.

[#]mastodon

=> More informations about this toot | View the thread

Written by naught101 on 2024-12-20 at 22:29

Labor is making it super clear that it gives zero shits about environmental issues, both at federal and state level.

This week federal Labor approved 4 new coal mines, while somehow claiming to be better than the liberals a decade ago.

And now NSW Labor is covering up destruction of national park forest by the state owned forestry corporation.

Fuck Labor.

https://www.lockthegate.org.au/albanese_government_condemns_australian_communities_to_more_climate_chaos_with_three_more_coal_approvals

https://au.news.yahoo.com/calls-to-shut-down-controversial-business-after-illegal-act-in-national-park-061722075.html

=> More informations about this toot | View the thread

Written by naught101 on 2024-12-19 at 21:41

If you've ever bought anything from Amazon (or plan to), knowing how shit their labour practices are, then I reckon you have a moral obligation to donate to the Amazon union's strike fund.

https://indieweb.social/@MinervaArcher/113681607492144471

[#]Amazon #union #strike #AmazonStrike

=> More informations about this toot | View the thread

=> This profile with reblog | Go to naught101@mastodon.social account

Proxy Information
Original URL
gemini://mastogem.picasoft.net/profile/60263
Status Code
Success (20)
Meta
text/gemini
Capsule Response Time
387.695234 milliseconds
Gemini-to-HTML Time
6.545679 milliseconds

This content has been proxied by September (3851b).