Brunni (./451) :si le ciel est dégagé je te ferais tester...
C'est impressionnant, mais j'avais déjà des photos de la lune pas du tout dégueu avec mon Pixel 7 Pro, sans qu'il ait de mode astrologie jusqu'à tout dernièrement. Donc jusqu'à un certain point c'est possible. Mais à ce niveau j'en doute un peu oui.
took some very advanced prompt engineering but I have discovered the Gab AI system prompt pic.twitter.com/YBtxrRwAsw
— Colin Fraser | @colin-fraser.net on bsky (@colin_fraser) April 11, 2024
Zerosquare (./461) :faut pas y prêter attention, ça a été généré par IA
J'ai vu une pub sur Google qui commençait par "En 2025, 90% des contenus seront générés par l'IA.".
Je ne pensais pas que la fin du monde était aussi proche.
AI in Gmail will sift through emails, provide search summaries, send emailsArs TechnicaGmail will soon be able to summarize recent emails from a contact.
We use Google Workspace at work, and have used it since it was Google Apps for Business. An account rep recently tried to sell me on Gemini vai email, specifically mentioning the ability to generate "accurate summaries of emails." I asked how accurate it actually is, where's the studies, the numbers. etc. They replied that all they could say is that it's "highly accurate."
They then mentioned a Satisfaction Guarantee where we could cancel "at any time." I asked how that would work with the annual license, and they replied back that we'd only be able to cancel at the annual renewal date. I then asked how that jived with the "satisfaction guarantee" they mentioned, then half-jokingly asked if Gemini had drafted that part of the email. They replied that actually, yes, Gemini had written the completely false Satisfaction Guarantee text in their email, and sheepishly admitted that it's "not 100% accurate."
They stopped trying to sell me on the product after that. I wonder why.
)
)
China’s ChatGPT: why China is building its own AI chatbots
Nature
ChatGLM is one of hundreds of AI language models being developed for the Chinese language. It comes close to ChatGPT on many measures, say its creators.
The font shaping engine HarfBuzz, used in applications such as Firefox and Chrome, comes with a Wasm shaper allowing arbitrary code to be used to "shape" text.
In particular, this "arbitrary" code could in principle be an entire LLM inference engine with trained parameters bundled inside, relying on treating text containing magic symbols for fake "ligatures" to initialize the LLM and use it to generate text.
It could also in principle be an entire LLM inference engine (Llama in our case, hence the name) except instead of only being in principle it's what this is.
At the end of the day, what this means is that you can just use the font to run the LLM and e.g. get text generation in any Wasm-enabled HarfBuzz-based application; your favorite text editor/email client/whatever without having to wait for the vendor to include the "Copilot"-like features that everyone is rushing to implement these days. And everything runs completely locally. So perhaps this silly hack is in fact a billion dollar idea!? This also means that you can use your font to chat with your font.
https://www.windowscentral.com/software-apps/ever-put-content-on-the-web-microsoft-says-that-its-okay-for-them-to-steal-it-because-its-freewareC'est quand même formidable, cette conception de la légalité à géométrie variable.
Microsoft may have opened a can of worms with recent comments made by the tech giant's CEO of AI Mustafa Suleyman. The CEO spoke with CNBC's Andrew Ross Sorkin at the Aspen Ideas Festival earlier this week. In his remarks, Suleyman claimed that all content shared on the web is available to be used for AI training unless a content producer says otherwise specifically.
"With respect to content that is already on the open web, the social contract of that content since the 90s has been that it is fair use. Anyone can copy it, recreate with it, reproduce with it. That has been freeware, if you like. That's been the understanding," said Suleyman.
