Kan AI me helpen mijn belastingaangifte invullen? Ja hoor! Toelichtingsdocument (132 blz, 59k woorden) en voorbeeldaangifte (14blz, 8k woorden) gedownload van https://t.co/ikqFHiyMGD en geüpload naar Google storage. Gemini 1.5 pro (die tot 1m tokens kan verwerken) doet de rest!
Wrote a summary blogpost about my experiences working on real-world AI projects:
Beyond the Hype: 7 Reasons Why AI Projects Fail
https://t.co/lCPo2OQE2C
After working in AI for 8 years, first as a researcher and then as a consultant, I have finally made time to collect some learnings. In this blog post, I look at the major reasons why most AI projects never reach production. https://t.co/L0GAKKIHAc
What do you do when you have a laptop from the Windows Vista era collapsing under the weight of Windows 10? You go on an adventurous quest to Ubuntu land!
https://t.co/g3iVJm0aqE
Big release that got under the radar 📡📟: last week @ml6team released Fondant-25M: dataset of image-text pairs with a @creativecommons license
And that's just the tip: they are working on a 500M one 🤯
Blog: https://t.co/HgJ9BbHOLa
Dataset on 🤗 https://t.co/fgdcm888dx
https://t.co/QW26ssdeOi has released a 25m Creative Commons Image dataset to build an image generation model that respects copyright. More info here: https://t.co/DrG3LqxAlW
Oh and by the way, InstructBLIP directly comes with 4-bit and 8-bit inference thanks to the magic of @Tim_Dettmers, enabling same performance and reducing the amount of memory by half or more!!
Hat/tip to @younesbelkada for integrating 💪
OpenAI & MS are being sued for data theft in California. The only way to solve this is to train models on data that allows being trained on https://t.co/jbt0A496SF
Wrote a blogpost on how Foundation Models are changing the way we set up AI systems. #FMOps is all about aligning model behavior through data and Ops. #mlops#foundationmodels https://t.co/JJzkfFgxbD
QLoRA: 4-bit finetuning of LLMs is here! With it comes Guanaco, a chatbot on a single GPU, achieving 99% ChatGPT performance on the Vicuna benchmark:
Paper: https://t.co/J3Xy195kDD
Code+Demo: https://t.co/SP2FsdXAn5
Samples: https://t.co/q2Nd9cxSrt
Colab: https://t.co/Q49m0IlJHD
Data preparation is going to be the #1 differentiator when it comes to Foundation Models.
Hence I've been involved in a new open-source project aimed at building & sharing reusable components for data preprocessing: https://t.co/9I7qqnLrz5
1/2
En nog meer ChatGPT in onze krant: tips om het als ideeëngenerator te gebruiken, gebrainstormd met de mensen van @ml6team: @JanLooy, @HaerensKarel en Mathias Leys.
https://t.co/hVeU28XPBB
Waar komt ChatGPT plots vandaan? Wat is generatieve AI en hoe gaat het ons leven veranderen? We leggen het uit, met veel dank aan @LaurentSorber, @NathalieSmuha en @slatre. En @janlooy en ML6 collega's ook, al zit hun input vooral in een later stuk.
https://t.co/8LjcOR73wy