Javier Fuentes Alonso

AI at @miro. Ex-Head of AI @uizard 🎨 Tweeting about applied AI and building AI-powered products. 🤖

[Javier Fuentes Alonso @JavierFnts]

🚀 Autodesigner is here! 🎉

Today is a big day in our journey of democratizing design through AI at @uizard. Now, anyone can transform ideas into prototypes with a simple prompt.

Don't believe it? See these jaw-dropping examples 👇

[Javier Fuentes Alonso @JavierFnts]

Autodesigner 2.0 is here! 🪄

Proud of the team and the AI-powered product we have built! 🙌

At the same time excited to keep empowering all the creative people out there to test out all their ideas. 🦾

[uizard ✨ by Miro Labs @uizard]

✨ Next level generative UI design is here ✨

The most popular UI generator just got even better! With Autodesigner 2.0, type out your ideas in plain text to design, prototype and iterate in seconds.

#autodesigner #uizard #generativeui

[Javier Fuentes Alonso @JavierFnts]

@bencium I am glad you enjoyed it! Let me know your feedback if you get to try Uizard yourself!

[Javier Fuentes Alonso @JavierFnts]

Excited to speak at the Autonomous Innovation Summit this Wednesday! 🗣️

I'll showcase our latest work on AI + Design at @uizard , demonstrating how it's breaking down barriers and enabling anyone to transform their ideas into attractive, functional prototypes. 🦾

[Javier Fuentes Alonso @JavierFnts]

And you thought that the long tail of your problem was difficult to model.

[Danielle Baskin @djbaskin]

Week of Cone RETURNS for halloween. This time, witches adorned with traffic cone hats perform a ritual on a waymo

[Javier Fuentes Alonso @JavierFnts]

@julien_c @AdeptAILabs Or just join @uizard if you are interested in UI understanding 😁 https://t.co/9XaGjUrm0a

[Javier Fuentes Alonso @JavierFnts]

*Cries in European*

[Anthropic @AnthropicAI]

We’re rolling out access to https://t.co/RxKnLNNcNR to more people around the world.

Starting today, users in 95 countries can talk to Claude and get help with their professional or day-to-day tasks. You can find the list of supported countries here: https://t.co/PbMuaqJcjU

[Javier Fuentes Alonso @JavierFnts]

Really interesting to read ChatGPT's prompt for DALL-E 3.

From the strong focus on diversity, to the format, to the language used...

[Javier Fuentes Alonso @JavierFnts]

👀👀👀

[uizard ✨ by Miro Labs @uizard]

👀 Something magical is coming👀

🧙‍♂️ Hold onto your hats, because a big update is right on the horizon...

#uizard

[Javier Fuentes Alonso @JavierFnts]

Great thoughts about designing AI products

[Sam Stone @sjstone1987]

At @magicaltome, we've been thinking a lot about how user experiences and interfaces for generative AI tools should differ from classical software. I've put some of the lessons we've learned into this new blog post, The Design of Everyday (AI) Things. https://t.co/RdDvudRF8e

[Javier Fuentes Alonso @JavierFnts]

@xhluca @lvwerra Fair point, though the dataset is still yours at that point, so you can technically move to a newer more capable model easily.

But I understand the possible risk.

[Javier Fuentes Alonso @JavierFnts]

@cjc @ckaleiki @lennysan @linear How does the project organizational structure work in practice when it comes to people management? Do ICs have 1:1s, yearly reviews, etc with the project lead? Doesn't it get messy if someone is part of multiple projects in a year?

[Javier Fuentes Alonso @JavierFnts]

Image generation problem: ✅

[Nathan Shipley @CitizenPlain]

Played around with DALL-E 3 this morning!

Here's a little screen capture of my how my "cloud made out of dogs" prompt evolved into... the Sky Dachshund franchise 🤣

#Dalle3 @OpenAI

[Javier Fuentes Alonso @JavierFnts]

A great overview of the different approaches in agents research https://t.co/G9WQ4bYmj5

[Javier Fuentes Alonso @JavierFnts]

@hrishioa Really good job with the guide, really on point 👏

About how to structure the examples for few shot learning, I haven't seen much difference between having them as part of the system prompt vs "fake AI messages".

Is this something you have seen working better consistently?

[Javier Fuentes Alonso @JavierFnts]

A really nice overview of how to currently make LLMs 's prompts to work in practice.

[Hrishi @hrishioa]

Some version of this guide exists at every AI company I know.

This is everything new I've learned from using Language Models in production.

In one place. I'll keep updating as I learn more.

https://t.co/X87XPgOhG5

[Javier Fuentes Alonso @JavierFnts]

This looks promising 👀

[AK @_akhaliq]

.@AdeptAILabs just released Persimmon-8B

blog: https://t.co/9bDCdU6psl
github: https://t.co/hwXswJ4hdb

Persimmon-8B, permissively-licensed language model with <10 billion parameters

trained it from scratch using a context size of 16K. Many LM use cases are context-bound; model has 4 times the context size of LLaMA2 and 8 times that of GPT-3, MPT, etc.

[Javier Fuentes Alonso @JavierFnts]

Cool stuff. Automatically generate relevant visualizations based on an arbitrary dataset.
https://t.co/SZQcPdDrJD

[Jeremy Howard @jeremyphoward]

It looks like @johnowhitaker &amp; I may have found something crazy: LLMs can nearly perfectly memorise from just 1-2 examples!

We're written up a post explaining what we've seen, and why we think rapid memorization fits the pattern. Summary 🧵 follows.
https://t.co/CUOWyxRJBT

[Javier Fuentes Alonso @JavierFnts]

@jeremyphoward @johnowhitaker Super interesting. Has anyone tried to reduce the dataset to a handful of examples? Does the loss still look the same? 🤔