I believe you will see a chatGPT level language model (on at least some metrics) on a mobile phone next year and GPT-4 level year after 🤓
These models are barely optimised, you would be very surprised 🫨
Folk may say this prediction is dumb but really what do they know 😊
1️⃣👑The King
AUTOMATIC1111/stable-diffusion-webui 🌟93K+ Stars - https://t.co/0rR0rVTvyr
🙌Use open-source models for image generation from text, in/outpainting, training, multi-embeddings, upscaling, batch processing, and more. It's an AI Swiss army knife for your art projects!
LIMA, a 65B LLaMa fine-tuned only with supervised learning on 1000 curated examples, without any RLHF, demonstrates remarkably strong performance, generalizes well to unseen tasks not in training data. Comparable to GPT-4, Bard, DaVinc003 in human studies.https://t.co/vNuecWIP5K
Can't help but feel the AI hype is oriented in a non-optimal direction
It's almost as if we had just discovered the FFT algorithm and instead of revolutionizing telecommunications, we are using it to build Tamagotchis
P.S. I'm only half joking 😄
We're investigating this and recommend not uploading LLama weights publicly on Hugging Face until we reach further clarification.
https://t.co/OBqiDziH1i
At @huggingface, we are on a journey to make our libraries more responsible, one commit at a time!
I am very proud to announce that we are launching today the Diffusers ethical guidelines 🧨
Read more about it: https://t.co/YBIgQG3rs8
1/4
🚨Just In: Roblox is bringing generative AI to its gaming universe.
They're building:
-a tool to make “generative AI materials from a text prompt”
-a tool for generative AI to complete code
Tests for these tools will roll out in the coming weeks.
As we invent new tools to process and create culture, new forms of #illiteracy have to be fought.
See how the wider public (and not only the wider public) is tempted to use #ChatGPT as a trusted source of information.
https://t.co/Sa4AkkaT76
An open source LLM that lives in your corporate VPC probably good for this eh.
Finally my email inbox will be sorted, another Carthaginian demon vanquished.
@EMostaque It's tempting to add 'Unique technology' (R&D, secret sauce, etc) but that would be the opposite of the open approach that has enabled the current renaissance.
This is a good take.
We are working on open chatGPT (android vs iOS eh) and folk will build on both.
Value creation will be decent but nothing compared to the crazy disruptive stuff that will come in the following years.
Cost to buidl on this minimal, costs scale with revenue
“Trending on artstation” doesn’t actually work on #StableDiffusion2 as we stopped using the OpenAI CLIP as is used in DALLE2 and SD1 which over-represented on this and artists in general.
Will be interesting to see how this whole area develops with some upcoming releases.
The Rails Foundation is hiring a remote, full-time executive director to run day-to-day operations and implement our mission to improve the documentation, education, marketing, and events in the Rails ecosystem. Salary: $135,000. Apply before Dec 21. https://t.co/9FLSavutHY
@massimosgrelli Definitely worth it!
V7 introduction of the HotWIRE paradigm improved consistency even further: no more JS bundling hell.
Arguably, docs are lagging behind. But the recent creations of the Rails Foundation, with substantial assets, should fix the issue.
https://t.co/zoQS9pEmJC