Introduction

Hi everyone!

My name is Davide Eynard and I am a Staff MLE at Mozilla.ai. First of all,
I would like to thank
the ppl who made the WebAI IG possible: I am excited to be part of it (it
is also my first time taking
part in a w3c group) and I hope to bring something useful to it!

My AI background is mostly academic, dating back to reasoning and the
Semantic Web, then classic
ML and computer vision, followed by deep learning and its extensions to
graphs. Six years ago
I moved to industry where I first worked on applications of geometric deep
learning to large social
networks (Twitter) and then, quite naturally, to applications of
language models (recommenders
first [1], and now mostly agents, at Mozilla.ai).

Where I stand wrt AI+Web can be easily summarized by taking the Mozilla
manifesto [2]
and substituting "the internet" with "AI": this translates into my current
work, aimed at building
small, personalized, open-source agents and teaching people how to tinker
with them [3].

Why am I here? Even when working with open-weights models, I
found concerning patterns that
IMHO resonate with this group's interests e.g.:

- models trained to preferentially target specific websites by default
(e.g. just ask questions to
gpt-oss-20b without a search tool, it will mostly default to Wikipedia)
- potential risks to user choice and autonomy (e.g. who is in charge of the
search tool that your
agent is using?)
- sustainability concerns for data sources being accessed by agents at
scale (see e.g. [4])

Looking forward to interesting conversations with you all!

Best,

Davide

References:
[1] https://github.com/mozilla-ai/byota
[2] https://www.mozilla.org/en-GB/about/manifesto/
[3] https://github.com/mozilla-ai/wasm-agents-blueprint ,
https://github.com/mozilla-ai/llamafile
[4]
https://www.engadget.com/ai/wikimedia-announces-ai-partners-including-meta-and-microsoft-162834383.html