Super hyped to have @modal on board as a sponsor for our next hackathon! π
Participants get $250 in free compute credits and a shot at winning a cool $2500 cash prize. Letβs go! π
working on an open-source tabular data cleaner built with the new @gradio dataframe component
β¨ upload or paste your csv/tsv/txt
π§Ή clean data with AI (dedupe, strip emojis, fix formatting)
built with svelte + transformers.js
contributions welcome!
https://t.co/pnwihvyfi5
π¨ new gradio release π¨
we've entered the frontend arena... you can now use the @gradio dataframe as a standalone component in your svelte projects!
π§΅
And just like that, @OpenAI gpt-oss is now the number one trending model on @huggingface, out of almost 2M open models π
People sometimes forget that they've already transformed the field: GPT-2, released back in 2019 is HF's most downloaded text-generation model ever, and Whisper has consistently ranked in the top 5 audio models.
Now that they are doubling down on openness, they may completely transform the AI ecosystem, again. Exciting times ahead!
π¬ From Replika to everyday chatbots, millions of people are forming emotional bonds with AI, sometimes seeking comfort, sometimes seeking intimacy. But what happens when an AI tells you "I understand how you feel" and you actually believe it?
At @huggingface, together with @frimelle and @YJernite, we dug into something we felt wasn't getting enough attention: the need to evaluate AI companionship behaviors. These are the subtle ways AI systems validate us, engage with us, and sometimes manipulate our emotional lives.
Here's what we found:
π Existing benchmarks (accuracy, helpfulness, safety) completely miss this emotional dimension.
π We mapped how leading AI systems actually respond to vulnerable prompts. π We built the Interactions and Machine Attachment Benchmark (INTIMA): a first attempt at evaluating how models handle emotional dependency, boundaries, and attachment (with a full paper coming soon).
οΏ½οΏ½οΏ½οΏ½ We also shipped two visualization tools on Gradio to see how different models behave when things get emotionally intense.
After a lot of excellent profiling and compressing by @evilpingwin, we were able shrink your Gradio app to just 1/5th of its size!
It took longer than we thought it would, but I think you'll be pleased with the results!
πππ πππππππ --πππππππ ππππππ
Introducing SmolLM3: a strong, smol reasoner!
> SoTA 3B model
> dual mode reasoning (think/no_think)
> long context, up to 128k
> multilingual: en, fr, es, de, it, pt
> fully open source (data, code, recipes)
https://t.co/duszyObJsG
You probably want to add the image to the folder and then import that image in your svelte or JavaScript file. This should give you a path that you can use as a source. This way is preferred because it means that the bundler sees it and can add it as a dependency.
Let me know if this works for you!
Do you have some information about rough costs per βlearning resourceβ? It might be viable for individuals with specific goals.
Iβd also love to see more technical details (or even open sourcing so people experiment individually). You could use a non commercial or share-alike licence if you are concerned about that.
I donβt think this is an ignorant take, it is a strategic attempt to cement power.
We have seen it time and again, OAI have no moat, their models are matched by others.
This is what the giant data centre is about. Make compute (and therefore capital) the moat.
We asked @rabois about AI insights from the most recent KV Summit at @khoslaventures.
"@altcap spoke about this: there's only going to be one foundation model research lab that'll be a successful winner."
"VCs shouldn't invest there. There's more room for VCs at the application layer, or something that transcends research labs today."
"The next generation of AI will not look like LLMs or research labs."