Benedek Fülöp

Benedek Fülöp

they/them

I try to explain AI, and fix bikes

What is in the Box?

1 minute read

Published:

AI technology exists and is widely available. Anyone can access the repositories, download the weights, read the papers of Arxiv. There is no putting it back in the box, we will have to live in a world with AI tools. So, the question becomes how can we shape these tools. This piece will detail some of the current methods and philosophies of guiding these mathematical heaps towards humane uses.

Purpose. Is where we will have to start. What is the goal of the system, why does it exists? Should it exists?

There are many controversial uses for the technology we call AI. From Clearview.ai specialising in facial recognition for law-enforcement and “investigations”.

This post is not done yet. The content is to describe the current conceptions of explainability, contestability and refusal in AI. I aim to start from the philosophical background behind explainations and the backstop of refusal. Then once the need for trust and iterative improvement is established. Describe and detail tha different explainability techniques from, Agnostic methods, Model-specific methods, Counterfactual explanations, Example-based explanations, Interactive/human-centric methods, Concept-based approaches, Ante-hoc vs. Post-hoc distinctions, Contrastive methods and mechanistic interpretability methods. This is in order to assess which of these methods may be best integrated with convivial conceptions of AI, easily accessible, low energy, and enabling tinkering.

You May Also Enjoy

Resisting the crawl

24 minute read

Published:

The past months many people have seen many legislative attacks on our identities (by the UK Supreme Court, Donald Trump’s executive orders, Hungary’s constitutional amendment), freedom of speech, and freedom to gather and demonstrate (arrest Mahmoud Khalil, Harvard funding freeze for failing to “limit activism on campus”). Just to grab a few examples of the top of my mind, I could not possibly be exhaustive of all the explicit violence spurred and carried out. This terrifyingly comprehensive and world spanning attack on gender, immigrants, journalists, organisers, and activists by neo-liberal, corpo-bureaucratic institutions, politicians and governments may very well be only the intro to a period of catastrophe. I doubt if you have found this you hadn’t experienced the chocking, static panic of hyper normalisation. The feeling of political and societal disassociation from reality as a constant stream of news and live social-media feed on tragedies, cruelty, hate and lately unabashed state violence unfolds without much objection, or response from power and leaders. Maybe complicit silence and shuffling. Crawling up from between your shoulder blades to encircle your throat, the rage of disability to witness atrocity without agency.

Another AI is possible?

32 minute read

Published:

2022 November 30th OpenAI released ChatGPT and with this launching machine learning/AI into the popular consciousness. Since then machine learning permeated more and more of people’s lives. Breaking out from the niches of industry to a truly ridiculous array of applications as corporates from big-tech to bakeries rushed to ride the hype. More importantly people are using AI in all sorts of ways. Students use it to write their assignments, employees use it to answer emails, create reports, and write queries. These uses motivate the all too pressing questions of safety, biases of models, and what we consider ethical as machine learning models present mirrors to our lives, companies, and societies. This article considers a more fundamental question: How are we building it? And more importantly: How else can we build it?

Fixing bikes and Explaining AI; The Unexpected Parallels

6 minute read

Published:

Bike Kitchens, at their core, are about empowerment. They teach people to understand their bikes, work on them, and regain confidence in their ability to move through the world. As I spent more time fixing bikes and teaching others, I began to notice surprising parallels with my academic work in explainable AI (XAI).