lingo.lol is one of the many independent Mastodon servers you can use to participate in the fediverse.
A place for linguists, philologists, and other lovers of languages.

Server stats:

67
active users

#talk

1 post1 participant0 posts today
Stefan<p>this is a pretty good talk!</p><p>State of Godot and the Web – Adam Scott – GodotCon 2025 <br><a href="https://www.youtube.com/watch?v=u_WMJG0menc" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://www.</span><span class="">youtube.com/watch?v=u_WMJG0menc</span><span class="invisible"></span></a></p><p><a href="https://gardenstate.social/tags/videogames" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>videogames</span></a> <a href="https://gardenstate.social/tags/godot" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>godot</span></a> <a href="https://gardenstate.social/tags/youtube" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>youtube</span></a> <a href="https://gardenstate.social/tags/talk" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>talk</span></a> <a href="https://gardenstate.social/tags/confrence" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>confrence</span></a></p>
Blort™ 🐀Ⓥ🥋☣️<p><span class="h-card" translate="no"><a href="https://mastodon.xyz/@nextcloud" class="u-url mention" rel="nofollow noopener" target="_blank">@<span>nextcloud</span></a></span> releases a new video / blogpost for Nextcloud Talk for <a href="https://social.tchncs.de/tags/DataControl" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>DataControl</span></a>.</p><p>The blog post embeds Google trackers.</p><p>The video can *only* be watched on a Google surveillance platform.</p><p>The app depends on Google / Apple for notifications.</p><p>Does anyone else see the problem here? None of these are necessary.</p><p><a href="https://social.tchncs.de/tags/Nextcloud" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>Nextcloud</span></a> <a href="https://social.tchncs.de/tags/Privacy" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>Privacy</span></a> <a href="https://social.tchncs.de/tags/Google" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>Google</span></a> <a href="https://social.tchncs.de/tags/Talk" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>Talk</span></a> <a href="https://social.tchncs.de/tags/YouTube" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>YouTube</span></a></p>
Benjamin Han<p>4/n</p><p>Talk: <a href="https://youtu.be/5QcCeSsNRks?t=0" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="">youtu.be/5QcCeSsNRks?t=0</span><span class="invisible"></span></a></p><p><a href="https://sigmoid.social/tags/AGI" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>AGI</span></a> <a href="https://sigmoid.social/tags/AI" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>AI</span></a> <a href="https://sigmoid.social/tags/talk" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>talk</span></a></p>
Continued thread

3/

* Two types of abstractions
- Type-1: prototype/value-centric, continuous — transformer is good at this
- Type-2: program-centric abstraction/graphs, discrete/search — ???
- Type-2 is data efficient but test-time inefficient due to combinatorial explosion.

* Intelligent systems needs to combine Type-1 and 2 abstractions
- Type-2 can be guided by type-1
- Need global library of abstract subroutines

Continued thread

2/

* The evolution of ARC benchmarks
- ARC-1: Prior-free reasoning tasks
- ARC-2: Compositional reasoning tasks
- ARC-3: Reasoning with interactive agency

* Intelligence implementation requires two components
- Abstraction acquisition
- Abstraction recombination (application)

1/

François Chollet, creator of Keras and the ARC suite of AGI benchmarks, gave a very thoughtful talk on how we can actually get to #AGI:

* Intelligence is a process, not skills
- Intelligence is an efficient conversion from past experience to potential operating area
- AGI is not about automation, it’s about invention

* Three characterstics of intelligence
- Fluid intelligence
- Broad operational area
- Information-efficient program acquisition and synthesis

Three common mistakes I’ve seen #FOSS people and other #introverted geeks make in #PublicSpeaking…well, as an introverted #FreeSoftware geek myself, these also things I have to keep in mind when I speak.

Slow…down. Speak at a consistent, slow speed. This ensures people don’t miss things, you get time to think, and people get time to absorb the information.

Speak loudly. There’s a limit to how much the microphone will do for you. Sometimes you won’t even have a microphone (e.g. when hosting outdoor #OpenStreetMap mapping parties) - learn to make yourself heard without one.

Don’t trail off. End your sentences with a consistent volume and pitch.

So how slow, and how loud? In a lot of cases I’d say WAY slower than what you call slow, and WAY louder than what you call loud 😅

There’s more, but this is the foundation of any talk. The best content in the world means nothing if it doesn’t actually reach your audience.

A lot of this is useful in everyday life too. Start practicing! 🙂

“Fascist AI” talk at LOOPS

The New Practice space at TU Berlin runs a series of talks called LOOPS. Together with my friend Malte I was invited to talk a bit about “Fascist AI”. How capitalism, fascism and AI narratives are very closely aligned and (re)produce one another.

I enjoyed giving the talk (I rarely do those together with others) and the Q&A and conversations afterwards immensely. You can check out a recording of the event here:

youtube.com/watch?v=OMl-yZMD-w

You can get Malte’s slides from […]

tante.cc/2025/07/04/fascist-ai