Tweets
-
Replying to @semateos and @SVVRLIVE
Neat, does the cabinet vibrate?
(original)
-
Replying to @utopiah
I imagine an obvious factor is the ability to livestream video content from a camera, so both connectivity and access to a camera on a phone.
Frequent in the Global North but not necessarily elsewhere, yet it’s still the “World” worth understanding.
(original)
-
“creating a highly detailed understanding of the world.”
I mean… yes but no? Those are very specific parts of the World, namely connected ones. One should be explicit about the bias because this is even more obvious than in other mapping projects, e.g OSM. https://x.com/tweetsfromasim/status/1856847688228299149
(original)
-
Replying to @utopiah
As per usual using such tool “kinda correct” :
(original)
-
Jack Ain’t Back by The Jackals https://soundcloud.com/dessous-recordings/jack-aint-back on Dessous Recordings 🎧
(original)
-
Replying to @utopiah
Piece found via https://www.science.org/doi/full/10.1126/science.adt6140 by https://x.com/MelMitchell1/status/1857195353629483472
(original)
-
⚠️ Highlights what I start to think of as “double goalposts moving” :
- dehumanize what we do, from work to creativity
while SIMULTANEOUSLY
- anthropomorphize what machines do, especially their failure.
Truly a sad squeeze.🛑🫷 Spot the pattern and reject it. https://x.com/emilymbender/status/1849802214090653995
(original)
-
Replying to @utopiah
Next time you use a cheap (VC subsidized in the hope of monopolistic rent extractions few years down the line) LLM to generate barely usable code or a low quality summary, think of the impact.
Related piece https://fabien.benetou.fr/Analysis/AgainstPoorArtificialIntelligencePractices
(original)
-
Replying to @utopiah
To get a bit more context https://x.com/AASchapiro/status/1831701005051793490
(original)
-
OK ok … but hear me out, what if AGI solves ALL our problems? Just one more batch of GPU, in a huge data center, water cooled, we’re nearly there! /$
(to be clear /$ stands for profit-based sarcasm)
We are actually fucked if keep going down the path of naive tech-solutionism) https://x.com/JimBair62221006/status/1857498272157544810
(original)
-
Replying to @utopiah
FWIW I also linked in the past with a payment platform, so that people could book a spot, pay, get the invoice and a calendar invite instantly.
All very professional and efficient.
I find this exciting that all those tools, yes even paper work related, become available to more.
(original)
-
I have my own @jitsiNews Meet server which I self-host, running meetings on it weekly, if not daily : it just works. 👌
I did link with calendar management to book spots.
This makes me think of contracts🖋️…
…and imagine a fully integrated self-hosted composable workflow.🤔 https://x.com/tom_doerr/status/1857617395189199141
(original)
-
Community notes highlight a great example on the “objectiveness” of benchmarks. Despite “showing” just objective data leading to obvious conclusions, warning about being “the same code”, it’s also wrong.🤷♂️ https://x.com/BenjDicken/status/1857449788893286484
(original)
-
Even “X” download archive function calls its result Twitter.🤷♂️
Just downloaded my twitter-2024-11-16-hash .zip today.
(original)
-
What if you want drag&drop files from anywhere else on Earth🌍? Not just your desktop🖥️ but also your collaborators from the other side of the globe?
Thanks to rclone by @njcw it also works with @Dropbox and countless other cloud storage providers! https://x.com/utopiah/status/1846807667085320685
(original)
-
RT @martinRenou: https://github.com/jupytercad/JupyterCAD
A collaborative #FreeCAD editor for JupyterLab!
@FreeCADNews #MadeWithFreeCAD #FreeCADFriday #…
(original)
-
Replying to @francoisfleuret
I was genuinely shocked that this Andreessen/Horowitz statement was public https://x.com/utopiah/status/1854068299979444269 because I was expecting them to hold that scaling belief. Seems they weren’t but I think they are the (realistic in that case) exceptions. Maybe I’m wrong.
(original)
-
Replying to @francoisfleuret
Makes me think of https://en.wikipedia.org/wiki/Rational_ignorance or arguably “worst” https://en.wikipedia.org/wiki/Agnotology
(original)
-
Replying to @francoisfleuret
I still think it holds within the AI community, again making the distinction between technical background vs not. I bet C-suite without an actual technical understanding, clearly investors, clearly marketers, are happy to believe whatever is “optimistic” and serves them.
(original)
-
Or why I egoistically focus on prototyping. https://x.com/effectfully/status/1857073716103905692
(original)
-
Replying to @francoisfleuret
Easiest way would be asking somebody around, as we all tend forget the bubbles we live within, who does NOT have a technical background, what they believe and possibly from there pull down the “why” they believe it.
I’m convinced “they” extend the “trend” but I’ll test today!
(original)
-
Replying to @utopiah
So yes, AI is chatbot, chatbots are better thus scaling works, scaling is leading to recursive improvement and thus we get to AGI.
I’m convinced it’s the popular belief.
It’s not the technical reality.
The 2 beliefs are co-existing in different populations.
(original)
-
Replying to @utopiah
OpenAI (and others since they are the most popular one, by far) are absolutely riding that since they mention AGI to the point that it’s been structural (being in the contract with M$ AFAICT).
(original)
-
Replying to @utopiah
I bet that pretty much any article on AI, that isn’t actually critical about the trend (so that’s a huge share) has at least an ending line on how “if it just keeps on improving, it could solve any problem” and that’s IMHO precisely hinting on scaling at THE solution.
(original)
-
Not sure if sarcasm or totally disconnected from the VC-fueled marketing hype machine that’s been rolling full steam ahead since ChatGPT is out.🤔
Yes, absolutely everybody without an actual PhD in CS is at least tempted to believe that because it’s been repeated relentlessly.🤷♂️ https://x.com/francoisfleuret/status/1857185503784714545
(original)
-
Replying to @GregMadison and @benz145
cc @zachernuk
(original)
-
Replying to @tom_doerr
Works in XR too https://x.com/utopiah/status/1723355347883057240
(original)
-
Replying to @utopiah
So now saved layouts can have optional perspective preview.
(original)
-
RT @AntonioCasilli: AI doesn’t build itself. Millions of unseen data workers train & update it.
On 21 Nov, 2:30 pm CET, @leilachaibi @Mila…
(original)
-
Replying to @utopiah
It’s just https://x.com/dmarcos/status/1848441421470273604 great demo, intertwine with my text/code on-going work https://git.benetou.fr/utopiah/text-code-xr-engine/ and mostly patience. The “trick” is to let the pen go to sleep, get your “hands” back, rinse&repeat.
(original)
-
So… there actually IS a way to have the MX Ink and hand tracking on the Quest… kind of,… if you are patient! And together, but not really : ✏️🤌😎 https://x.com/utopiah/status/1844350729991094318
(original)
-
Going meta : sendPerspectiveToServer() sharing the perspective, not as an “objective” data file but rather as an image from the user viewpoint.
(original)
-
Replying to @utopiah
Also if done right the setup doesn’t have to be complicated, cf e.g. https://git.benetou.fr/utopiah/text-code-xr-engine/src/branch/fot-sloan-companion/Dockerfile with already quite a few otherwise" heavy things in place.
One could also imagine that specific heavier computation (vision is demanding, even tiny models) delegating to other machines.
(original)
-
With such an architecture tagging with meta-data an image, locally thanks to vision models (e.g. @MoondreamAI) to be searchable, filterable and usable in XR becomes trivial. It’s just a single additional (optional) step in the cascading conversion process : https://x.com/utopiah/status/1857378905872609572
(original)
-
Replying to @utopiah
And on-going https://x.com/utopiah/status/1857335894006505530 clarification of the coarse architecture.
(original)
-
Replying to @utopiah
How it’s done https://git.benetou.fr/utopiah/text-code-xr-engine/src/branch/fot-sloan-companion/
(original)
-
Literally juggling with files :
from 💻 desktop to ➡️ XR🕶️ to ➡️ any other device📱.
(original)
-
Replying to @WolvicXr
Indeed thanks for the official confirmation https://x.com/utopiah/status/1855318410814693803
(original)
-
Replying to @utopiah
Hopefully much clearer version. Still “shocked” how convenient it is not to care much about the input format. Here the source file is a .odg by @LibreOffice Draw and… it’s just there, manipulable in XR.
(original)
-
It’s fascinating how expertise is invisible from the outside.🤔
I’m no ecologist, yet I bet checking “Screenshot of a demo for Sora, via OpenAI.” for barely a second an expert would spot how nonsensical this “plausibly correct” image is.
Do those animals really live together?🤷♂️ https://x.com/bcmerchant/status/1857192464534176010
(original)
-
Replying to @utopiah
Reminds me I need to consider blinders https://x.com/utopiah/status/1788926413300498792 so that I can still my laptop screen, keyboard, coffee, etc while traveling yet occluding “noise” in the broad sense (sorry other travelers).
(original)
-
Replying to @utopiah
It will be dedicated to run https://x.com/utopiah/status/1857339273449537762 fully offline.
(original)
-
Replying to @utopiah
To give a bit more context, this is all relying on https://x.com/utopiah/status/1846807667085320685 namely putting files back in the center, giving affordances in XR.
(original)
-
Replying to @utopiah
… and by updating default.layout.json accordingly, positioning “behind” the user default start position, it’s available every time the system starts.
(original)
-
Replying to @utopiah
A bit too literally but arguably already useful :
(original)
-
Replying to @utopiah
Main idea here, that the system itself serves its own documentation https://git.benetou.fr/utopiah/text-code-xr-engine/src/branch/fot-sloan-companion
Namely have a schema, e.g. SVG file, possibly embedded in the usual README md file, that is then available via default.layout.json with optionally clarification post-its
(original)
-
2:33AM notes on ideas💡🌃… not sure if a good thing or not.
Helps me get back to sleep instead of ruminating, or, maybe worst, actually waking up and work on the idea.
… but would I just NOT get thinking if I knew in advance there would be no chance to take notes🗒️?🤔
(original)
-
Preparing a minimalist travel kit to tinker on the Future of Text in XR.
(original)
-
Replying to @utopiah
Same idea https://x.com/_MichaelOswell_/status/1857004299739439387
(original)
-
AI code assistant is the “new framework” replacement.
Now imagining generating a new framework.
Faster slop. Lower value. More waste. https://x.com/wesbos/status/1856806422899822999
(original)