Tweets
-
RT @hackylawyER: “As we celebrate the death of one toxic business model, we must keep our guard up lest we usher in something potentially j…
(original)
-
Replying to @Silka_Sietsma
FWIW https://twitter.com/utopiah/status/1645434640608419842 thus why I’m not scared, mostly frustrated.
(original)
-
Replying to @utopiah
PS: “proof” that I get it https://twitter.com/utopiah/status/1621760472461594624 showing that I did run generate code in the current environment that called for it (trivial in Javascript) https://git.benetou.fr/utopiah/text-code-xr-engine/commit/37b6e058826c9e10e62072e96446c96969c3c6be but I wouldn’t call that autonomous or self-improving. There are numerous 🫂➿🤗 not 🤖🔄🤖IMHO.
(original)
-
Replying to @utopiah
Finally why I believe relying on GPT4 for autonomy and democratization is a problem https://twitter.com/utopiah/status/1558428237281239040
(original)
-
Replying to @utopiah
IMHO https://twitter.com/utopiah/status/1640276599760211970 is more interesting but as mentioned so costly.
Despite all these criticisms if the impact, not just the goal, is democratization of access to better models and tools for all, not just privileged developers, that would be very positive.
(original)
-
Replying to @utopiah
or if it’s basically another attempt with the trendy implementation of the moment, i.e LLM, but honestly without clear example of a qualitative difference by default I remain doubtful of the actual interest and thus wary of such claims.
(original)
-
Replying to @utopiah
Self-improvement or rewriting itself has been a goal, and a concern, since the term AI was coined.
There have been art pieces, e.g Google Will Eat Itself (GWEI) - Hacking Monopolism Trilogy https://gwei.org 2005, for decades.
Would be good to pinpoint what’s different…
(original)
-
👀 “This program, driven by GPT-4, autonomously develops and manages businesses to increase net worth.”
Does it really though? Seems like asking for money while people do merge PRs, write prompts, manage issues, while the core is a commercial black box.
Where is the autonomy?🤷♂️ https://twitter.com/SigGravitas/status/1642181498278408193
(original)
-
Replying to @coderofsalvatio, @seflless, @mozilla and @djleufer
But only recently did https://twitter.com/utopiah/status/1642879603151712258 with just before https://twitter.com/utopiah/status/1642037804073926656
(original)
-
Replying to @coderofsalvatio, @seflless, @mozilla and @djleufer
That was much before “neural” chips in mobile phones or https://precursor.dev AFAIK.
(original)
-
Replying to @coderofsalvatio, @seflless, @mozilla and @djleufer
FWIW wrote about this kind of stuff, poorly, vaguely and incorrectly I’m sure, a decade ago https://fabien.benetou.fr/Cookbook/ObjectsExoBrain
(original)
-
Replying to @utopiah
PS: I’m aware of Veblen goods https://en.wikipedia.org/wiki/Veblen_good and find that in a finite world where we know resource consumption, even for virtual goods, does have a huge impact on our quality of life, destroying our very own ecological niche, is a special kind of idiocy.
(original)
-
Economical value from solely ownership without actual usage is exactly the kind of future I hope to avoid.😡
“Virtual Metaverse Real Estate Is Completely in the Toilet - The metaverse is completely screwed.” by @vtanger for @futurism
Artificial scarcity isn’t per se valuable!📉 https://twitter.com/razialo/status/1645366461056139264
(original)
-
Replying to @utopiah
Damn, buffer, there is no butter involved. Guess I need a Kouign Amann boost🧈😂
(original)
-
Replying to @utopiah
I’m relying on SantaCoder on HuggingFace but might as well rely on a local instance thanks to the recent Docker announcement https://twitter.com/utopiah/status/1641031814994186244
(original)
-
Replying to @utopiah
To clarify a bit the Vim syntax :
- V selects the current line
- with r read the following results and output in current butter
- ! runs a filter as a command from the shell (here sending the current selection to SantaCoder for completion as a shell script)
(original)
-
Replying to @utopiah
Perspective motivating this kind of efforts https://twitter.com/utopiah/status/1306566303839391744
(original)
-
Code generation in Vim thanks to @BigCodeProject SantaCoder.
No VSCode, no Copilot, no Github, no Microsoft.
These are very popular choices for developers, so why bother with alternatives? Because who owns the tools matter, not solely how good they are.
PS: hallucinated code.
(original)
-
Replying to @emicovi, @EMostaque and @iwontoffendyou
Check @BigCodeProject
(original)
-
RT @Level1_Geek: Have you ever wanted to play #DnD with kids? I worked with @rbswertfeger to create this guide for you! You’ll find tips fo…
(original)
-
Replying to @simonw
Indeed yet I’d argue with the right mindset, even a grain of rice can be a brilliant teacher. The intelligence here comes from you, not the AI. Consequently the question is how can we kickstart and support that process for all learners, including those who aren’t there yet?
(original)
-
“Grounded-SAM + Stable-Diffusion Inpainting: Data-Factory, Generating New Data!”
(can’t wait for the @huggingface demo to use its new @docker feature and be able to test with (hopefully) a 1-liner. Don’t want to fight with Python dependencies anymore.) https://twitter.com/_akhaliq/status/1645115958594351106
(original)
-
Replying to @Chris_BNEO, @aframevr and @threejs
True I do hope some will say “Damn… that’s ugly” to which I can reply “Let’s do better together!“😁 but only AFTER they get the essence of game mechanics, how we can design how we interact with that brand new world to make it interesting.
(original)
-
Replying to @utopiah
More examples to come and as usual suggestions always welcomed!
(original)
-
Kids were really into gaming so for the next workshop a new game-mechanics branch!
Starting with a Minecraft inspired example showcasing the core idea in a dozen lines of code https://git.benetou.fr/utopiah/text-code-xr-engine/commit/01f67fa7b18e270586548a2666a971d1ce66fe8d thanks to @AFrameVR & @threejs raycaster https://aframe.io/docs/1.4.0/components/raycaster.html https://twitter.com/utopiah/status/1642775410332008448
(original)
-
Replying to @rmfrt
cc @mattdesl for other pointers in case it’s not a fit.
(original)
-
Replying to @rmfrt
I would open up https://p5js.org/get-started/#sketch and play for at least 10min (literally, as there is no install required) then eventually if it doesn’t work out, list why showing the result back here and hopefully others can help better.
(original)
-
Replying to @pauls551 and @GaryMarcus
Hopefully he did see “Her” though. Much more mundane and yet with scammers https://www.theguardian.com/technology/2023/mar/08/darktrace-warns-of-rise-in-ai-enhanced-scams-since-chatgpt-release the path is rather obvious, when it doesn’t get much darker https://mobile.twitter.com/utopiah/status/1644410833198358548
(original)
-
🔥 Glad to see composability in AI grow with open source models:
“RWKV is an open source RNN with Transformer-level LLM performance”
precisely because IMHO performance isn’t everything. Yes we all want to do more with great accuracy but not by stacking inscrutable black boxes. https://twitter.com/LangChainAI/status/1644803966046466048
(original)
-
Replying to @LFeldmanBarrett and @rowancheung
Cc @djleufer
(original)
-
Replying to @PierreLebeaupin and @Coxygrue
Par curiosité est-ce que cette demande augmente ou diminue ?
Dans quelle proportion par rapport aux autres postes ?
À quelle vitesse dans le publique versus privé ?(original)
-
Replying to @AntoniaRForster, @IanXRNerd and @LumaLabsAI
Cc @karanganesan
(original)
-
Replying to @seflless
Wasn’t earning money with it in 2010 https://mobile.twitter.com/utopiah/status/15507761299197952 and wasn’t called generative AI either but proof I was studying the topic.
(original)
-
Replying to @seflless, @mozilla and @djleufer
So AI? Yes more of it actually it also better AI for a better society. I’m not sure I’m doing this right and I surely don’t want to be naively critical either. I simply find that challenge at the intersection of both technology and positive inclusive impact even more interesting!
(original)
-
Replying to @seflless, @mozilla and @djleufer
a lot more about the broader consequences. I still love technology, and even the conceptual aspects behind e.g computational complexity, deeply but I just can’t be blissfully unaware of all its costs anymore. Once the curtain is lifted we can’t forget.🤷♂️
(original)
-
Replying to @seflless
Maybe it’s anecdotal but I attribute my personal shift to @Mozilla and @djleufer mostly.
Until then, even though I cared for Linux and thus FLOS since I started interacting with computers, it was mostly a solo endeavour with impact limited to myself.
Since then it’s been…
(original)
-
Replying to @seflless
FWIW https://fabien.benetou.fr/Content/SelfHostingArtificialIntelligence
(original)
-
Replying to @seflless
So exploring AI, from generative AI to LLMs to whatever one finds interesting, absolutely.
To fuel models, both technically or commercially that aren’t aligned with what I believe to be more crucial regardless of the fascinating nature https://mobile.twitter.com/utopiah/status/1306566303839391744 no.
(original)
-
Replying to @seflless
FWIW I made money with generative AI in 2014, so before I started working on VR. I also studied formally AI before that. So the technology itself I find not only fascinating but even important regarding what creativity even is. Yet I find the buzz itself damaging to the field.
(original)
-
In case you find me too critical of the AI buzz, in particularly LLMs as able to do a lot and speculatively everything, maybes listening to DARPA instead could help bring a healthy bit of scepticism back. https://twitter.com/DARPA/status/1641833490168655876
(original)
-
Replying to @stevenbeeckman
Can’t they do a search in logs? Without having access to the logs but while reading things like https://www.vice.com/en/article/pkadgm/man-dies-by-suicide-after-talking-with-ai-chatbot-widow-says supposedly “fixed” after the event, and again to recall happening with an emotional context of somebody who is emotionally fragile, I don’t find that dubious.
(original)
-
Replying to @stevenbeeckman
Why dubious? I didn’t have access to logs but if citations from the discussions are correct (seems journalistic trivial to check) e.g his children dead or not refuting that self sacrifice could help saving the planet (one his main concern), it’s pretty obviously problematic.
(original)
-
RT @alexjmurrell: The Age of Average.
“From film to fashion and architecture to advertising, creative fields have become dominated and def…
(original)
-
Replying to @utopiah
Initial article on chatbot named “Eliza” linked to suicide https://www.lalibre.be/belgique/societe/2023/03/28/sans-ces-conversations-avec-le-chatbot-eliza-mon-mari-serait-toujours-la-LVSLWPC5WRDX7J2RCHNWPDST24/ after conversations with it from someone already depressed or at least in an emotionally troubled condition.
(original)
-
Replying to @hughhowey
It’s more than controversial, it’s both dangerous and naive. Please do listen 27min in https://twitter.com/tristanharris/status/1639426462540505088 it will just take a minute.
(original)
-
Replying to @utopiah
Sadly, here we go https://twitter.com/hughhowey/status/1643828206045216768
(original)
-
Replying to @razialo
The one that was associated (edited to remove “lead to”) a suicide in Belgium lately or the 60’s one?
Anyway linked to https://twitter.com/utopiah/status/1644548377726648320
(original)
-
Replying to @utopiah
Next time I read about anyone suggesting that psychologists time is counted I’ll point to that example. This is seriously dangerous and again this is from a commercial product made by a company which raison d’etre, its founding goal, was AI safety.
(original)
-
Replying to @rikarends
Even worst the institutions who should be role models relying on such tools. Deeply disappointing.
(original)
-
27min in example is just… inhumane. https://twitter.com/tristanharris/status/1639426462540505088
(original)