sebszyller.com

Opinionated takes on machine learning, privacy, confidentiality, and their intersection with law and regulation. Note that all opinions here are my own and do not reflect the official stance of any party that I'm affiliated with unless explicitly stated otherwise.

Blog

Open Weights Have Nothing to Do with Open Source

We love open source because we can read and modify the code. We love open weights because we can run models locally. They're the same thing.. wait, what?! Read more.

Learning Rust with Large Language Models (Part III): Finding a Needle in a Haystack

Low mean and high standard deviation is the name of the game when it comes to complex topics and LLMs. Read more.

Learning Rust with Large Language Models (Part II): an Outdated Manual Written by a Newbie

Imagine a teacher who has read only the introductory chapters but of all textbooks in existence. Enter learning Rust with LLMs. Read more.

Food Markets Are a Bad Analogy for Data Marketplaces

January 21, 2024

It's common to refer to data marketplaces using the food market analogy -- it's easy to understand and unfortunately, provides us with a bad mental model. Read more.

Learning Rust with Large Language Models (Part I): a Project for 2024

January 2, 2024

Bespoke tutoring is one of the promises of large language models. Let's see if they can help me learn Rust. Read more.

No One Cares About Large Language Models Anymore

December 23, 2023

Large language models are just better search. It's all hype. Everyone works on multimodal and video models now. Right? RIGHT?! Read more.

Content Provenance Needs Critical Mass

November 29, 2023

Leica adds provenance support to its latest camera, while OpenAI claims they can't reliably detect synthetic text. Is content provenance just a pipe dream? Read more.

Data Marketplaces for Individuals (Still) Don't Make Sense

November 2, 2023

Controlling your own data is nice. Decide what you want to share, and charge for it; or don't share anything at all. But other than data that compromises your privacy, do you have anything worth selling? Read more.

The Security Through Obscurity Moment of Large Language Models... and Money

OpenAI is on track to learn all the good security practices the hard way. Probably make a tone of money in the process too. Read more.

Finding the Beauty in the Imperfections of Generative Art

September 5, 2022

Generative art has taken over the internet by storm. It fails in more than one way but do we really care? Read more.

Can You Spot a Deepfake?

This looks Photoshopped. This CGI is bad. We either heard or said these more than once. There's a new cool kid on the block now. Read more.

Kosher Data for Your Ethical Needs

I like my data like I like my coffee -- sourced without exploitation. Read more.

Why Synthetic Data Is Not Private

And what's more, it's snake oil... unless done right. Read more.

Data Marketplaces for Individuals Don't Make Sense

Or at least are challenging. Who doesn't want to go against big tech. and have agency over selling their own data, right? Read more.

On the Difficulty of Cross-disciplinary Communication

October 30, 2021

I'm sorry but I don't speak corporate law. Could you explain it to me in Greek instead? Read more.

Whose Model Is It Anyway?

February 10, 2021

You gathered the data, sanitised it and trained your fancy model. Your clients love it. Can you prove that it's yours? Read more.