Tweets
-
Replying to @dmarcos
Indeed🫡 I believe so too, hence organizing https://fabien.benetou.fr/Content/SelfHostingArtificialIntelligence hoping to show that a lot can be hosted.
Eager to able to clarify all that though from a license standpoint, with FSF in particular https://www.fsf.org/news/fsf-is-working-on-freedom-in-machine-learning-applications and also contrasted with OSI https://opensource.org/ai/open-source-ai-definition
(original)
-
Replying to @utopiah
Same trend… https://x.com/xenovacom/status/1852251703791288469 technically very cool shows that interfaces, and their underlying APIs, are not truly accessible.
A lot of Web “hacks” are done, ironically enough, thanks to ARIA, cf https://www.w3.org/WAI/standards-guidelines/aria/
(original)
-
Replying to @utopiah
More directly https://x.com/utopiah/status/1498394674993717251 i.e. is the resulting representation more a graph than a map, in which conditions, and thus does this teleportation in a document (again, it is instantaneous) changes the underlying representation?
(original)
-
Vi /keyword (1-keystroke search) is like teleportation.
Sounds trivial yet means, the “shape” of code, how long your file is, how you indent it, etc, matters little. It works on the same line… or the entire file.
To put in perspective w/ cognitive maps https://www.science.org/doi/10.1126/science.aat6766
(original)
-
“In the era of generative AI, the Infinite Monkey Theorem, and its finite version, perhaps also challenge readers to consider philosophical questions around the nature of creativity, meaning and consciousness, and how these qualities emerge.” https://phys.org/news/2024-10-monkey-unable-hamlet-lifetime-universe.html https://x.com/utopiah/status/1852245351324889135
(original)
-
Replying to @GeorgeCrudo
Yep… I’m trying locally for a while now https://fabien.benetou.fr/Content/SelfHostingArtificialIntelligence on pretty much any topic and it’s just good enough to generate slop. It’s sufficient for spam and scams but for anything requiring actual quality, across domain, it’s still low.
(original)
-
Replying to @utopiah
One could argue that AI gives more “flexibility”, no need for a well defined JSON or whatever, to make it work. True but it’s :
- destroying the environment (orders of magnitude more computing required, with negative side effects)🌍
- unreliable⚠️
- comparatively very expensive.
(original)
-
Exactly, cf https://x.com/utopiah/status/1848971481536921865 when there is “AI in the middle”, between humans trying to interact, surely there can and should be a more direct way.
That’s what interoperability is for! https://x.com/danbri/status/1851960395659120749
(original)
-
Replying to @cosmoscharf
cc @liquidizer
(original)
-
Replying to @utopiah
Think GenAI vs Club of Rome’s The Limits to Growth.
(original)
-
Replying to @utopiah
…This only considers the infinite limit, with either an infinite number of monkeys and/or an infinite time period of monkey labour."
Sounds pointless, yet in a world of finite resources grandiose claims must be checked against actual bounds.
(original)
-
“A numerical evaluation of the Finite Monkeys Theorem” https://www.sciencedirect.com/science/article/pii/S2773186324001014
“The Infinite Monkeys Theorem has long-established the eventual certainty of the complete works of William Shakespeare being reproduced by a monkey randomly pressing keys on a typewriter…
(original)
-
Replying to @JordiClive, @karpathy and @obsdmd
Once it’s setup, it becomes a single command, can even become a shortcut. Could also scp all new files once connected to the WiFi at home (they are very small, even if you write a lot) so that even with the device off you still have it all.
(original)
-
Replying to @JordiClive, @karpathy and @obsdmd
Funny you say this, posted just earlier today https://x.com/utopiah/status/1852006413179965468
So… they do propose their cloud as a commercial offering but one doesn’t have to rely on them. You can directly ssh/scp to the device, get the raw data, convert to e.g. SVG and integrate in your workflow.
(original)
-
Open workflows are underrated.
Going from my (Linux) e-ink tablet (reMarkable) to my desktop (also Linux) in seconds.
Think, sketch, build, iterate.🔄
(original)
-
Damn… that’s powerful. What it takes to be a pioneer :
“you’re not a carpenter and he’s not a fucking blacksmith.
You are pioneers, and that’s all you are until you get there. You have no home, no job, no farm.
You have the journey, that’s it.” (extract from 1883)
(original)
-
RT @FrontYoungMinds: For #halloween2024, dive into our article on scary #evil characters and why they make us connect better to stories wit…
(original)
-
Replying to @Azadux and @kfarr
maybe you know someone who could help?
(original)
-
Where does thinking take place?🧑🤝🧑🧠✍️
In the brain? Behind the computer? On the wall? While discussing standing up in front of a board? At the intersection of all those people and tools?
Photos from “Inside @CERN” https://www.lars-mueller-publishers.com/inside-cern
(original)
-
Replying to @Dilmerv and @PuzzlingPlaces
If correct @liquidizer this could also means the work could be still more inclusive.
(original)
-
Replying to @Dilmerv and @PuzzlingPlaces
Ah, that was precisely my question. I assumed that because it’s relying on CloudXR it runs a Windows desktop in the cloud. Thus it means it should be possible to run Chromium and thus work with WebXR. Can you then confirm that WebXR content works on it?
(original)
-
Replying to @jonititan, @ImmersiveWebW3C, @WolvicXr and @aframevr
Pulling @felix_trz in as he’ll know better.
Theoretically WebXR is cross-platform but of course, like any specifications, some features are platform specific, even device specific. Always good to query for feature availability.
(original)
-
Replying to @rcabanier and @hybridherbst
Video of the problem. I think I’ve isolated the bug to the Browser because switching back to another browser the hands are back.🙌
(original)
-
Replying to @rcabanier and @hybridherbst
Even quitting the browser before removing the headset does not help. It happens regardless of WebXR framework used.
(original)
-
Retrospectives of exploration series.🤔
Regularly specific prototypes lead to plenty more. After a short period of time, say up to a month, of accumulating those resulting explorations, I should step back & describe (text, video, notebooks, etc) the what & why it was fruitful.📓
(original)
-
Replying to @haagch42 and @kfarr
And https://x.com/utopiah/status/1188113086483447809 at @FossXR 2019 too!
(original)
-
Replying to @haagch42 and @kfarr
Indeed I discussed with KaiRo/Robert Kaiser about his project during a Mozilla DevRoad event in Austria a while ago. He’s mentioned in https://github.com/ctrlw/osm4vr?tab=readme-ov-file#credits
(original)
-
Replying to @zotero
Looking for something like https://genomebiology.biomedcentral.com/articles/10.1186/s13059-017-1205-3 but with anything that can be parsed (e.g. BibTex, JSON) and applied to any topic, genomics and otherwise.
(original)
-
Replying to @rcabanier and @hybridherbst
Loosing my hands in WebXR frequently. The only way to get them back is to reboot. I don’t recall this happening before getting the pen. I might try to remove it for few days and try but curious to know if others are experiencing it.
Hands always work in the OS but not in WebXR.
(original)
-
Replying to @weareagog, @cgiller and @AmySeidenwurm
Actually @kentBye might be there, if you don’t already know each other might be interesting to meet!
(original)
-
“Technical Dimensions of Feedback in Live Programming Systems” by Josh Horowitz https://www.youtube.com/watch?v=lTh8BwoSlnM at LIVE 2024 https://liveprog.org itself at SPLλSH (Systems, Programming, Languages and Applications: Software for Humanity) 2024
(original)
-
Replying to @utopiah
Also makes me wonder, does @Zotero or other bibliography managers support this somehow? Can they automatically download the materials?
(original)
-
Hey researchers, in your domain, is there a convention for supporting materials?
I’m unable to sport a pattern (beside random URLs), even less a convention, never heard of a standard.
I believe it is hurting scientific reproducibility. Any pointer for on-going efforts?
(original)
-
Replying to @utopiah
Same topic, now imagine “For disaster survivors, generative AI could act as a single front door to various programs” https://www2.deloitte.com/us/en/insights/industry/public-sector/automation-and-generative-ai-in-government/leveraging-ai-in-emergency-management-and-crisis-response.html where you just survive a disaster, only to talk to chatbot rather than a human. It’d better work damn perfectly. Not my experience so far.
(original)
-
Replying to @benz145
No, it’s the past, it’s been done already, e.g. “Hacking how we see - A way to fix lazy eye?” https://media.ccc.de/v/35c3-9370-hacking_how_we_see which I had the change to see back then, nearly 6 years ago already.
(original)
-
Replying to @utopiah
I’m at a loss.
This is seriously wrong.
(original)
-
Replying to @utopiah
To give back a bit of context, this isn’t about picking another color for your car, or different background music for a video clip, it’s about saving, or not, lives.
You can NOT trust the “intern”. You can NOT trust generative AI.
Why include it in scenarii where truth matters?
(original)
-
Replying to @utopiah
Now imagine that the person giving the training was an intern genuinely trying to be helpful yet role-playing as an expert.
Would you trust them? They would stick to the manual, trying hard to do it right… but is it enough?
(original)
-
Replying to @utopiah
It’s driving my nuts to read this kind of non-sense because it’s more than non-sense, it’s actually dangerous.
I did first aid trainings and if there is ONE thing you need it is to TRUST the information given because you MUST act NOW. Right fucking now.
(original)
-
I’m genuinely trying hard not to be so negative but… “Generative AI can provide real-time information, basic first aid, and health knowledge to affected communities.” https://www.nextgov.com/ideas/2023/10/generative-ai-set-transform-crisis-management/391264/
Generative AI already told people to eat rocks & cook with deadly mushrooms.
So, no!
(original)
-
Replying to @kfarr
I did some related work a couple of years ago https://x.com/utopiah/status/1535153496009342976 but this is way more convenient.
(original)
-
Replying to @utopiah
Details on the component itself at https://github.com/ctrlw/osm4vr , kudos to ctrlw, thanks to @kfarr for bringing it to light and contributions.
(original)
-
Another beautiful example of composability! Bringing @OpenStreetMap content straight to #WebXR thanks to @AFrameVR component loading the 3D buildings of a target location.
Life-size, large plan on the floor or mini map or a floating table, it’s all there in 2 imports and 1 line! https://x.com/3dstreetapp/status/1851300749978058944
(original)
-
Replying to @utopiah
PS: hands 🙌 have an offset. Not sure why but would be interesting to investigate.
What’s surprising though is… how quickly one can adapt! The hands are NOT where they are supposed to be the offset is constant so… I can still grabs objects easily, go-go-go gadget arms! 😅
(original)
-
Replying to @utopiah
If you don’t feel like tinkering and building an entire browser, then there are also add-ons or extensions (non Chromium building only for now AFAICT) so you can still change the behavior of the browsers to adapt it to your needs. Very exciting!
(original)
-
Replying to @utopiah
IMHO what’s exciting isn’t that it works… but also that you can check HOW by going to https://github.com/Igalia/wolvic … and if you don’t like it, you can suggestion how to do better.
(original)
-
#WebXR isn’t about “just” relying on stock browsers. WebXR is about XR content working on any and every browser that supports @ImmersiveWebW3C specifications!
Here testing the latest @WolvicXR build on my @AFrameVR experience in AR mode.👌 https://x.com/WolvicXr/status/1851281638426050903
(original)
-
RT @dmarcos: A-Frame in da house at @github Universe
One of the only 10 projects selected for the open source zone. Sharing the floor with…
(original)
-
Replying to @utopiah
Basically… either you provide value, or if you accept you are purely exploratory, which is also great IMHO but it’s a different process with different requirements and outcomes.
Providing value is building a product, explorations is improving learnings.
(original)
-
Great article, I’d generalize the conclusion :
“trying to solve a non-existent problem with VR [& any other emerging tech] is pointless.
… VR[&more] development takes longer than mobile dev. You might end up spending more time on the wrong idea in this space & burn yourself.” https://x.com/tom_krikorian/status/1851003764036997199
(original)