What it’s like to be an AI ethicist working to make technology safe

• Giada Pistilli, 31, is the lead ethicist at Hugging Face, helping to safely deploy AI.
• The main question that drives Pistilli’s research is, “How can ordinary people use AI for good?”
• Speaking to Insider’s Aaron Mok, Pistilli’s work as an AI ethicist includes:

This narrated essay is based on a conversation with Giada Pistilli, 31, based in Paris, France, about his job as chief ethicist at AI company Hugging Face. The following has been edited for length and clarity.

I’m a full-time AI ethicist, making sure technology is deployed safely in the world.

But don’t be fooled. I am not an oracle.

In short, I work at the intersection of ethics, policy, and law to advance an ethical framework around AI for Hugging Face and the general public. The main question that drives my work is, “How can ordinary people use AI for good?”

I previously worked as a policy adviser on human rights issues at the European Parliament. After retiring, I completed a master’s degree in political philosophy and ethics at the Sorbonne University, followed by a PhD in philosophy from the same school, and am currently compiling a dissertation on the ethics of conversational AI.

After completing my master’s degree, I worked as a research engineer at a Paris-based chatbot company, working with designers and machine learning engineers on applied ethics research. Later, I joined BigScience (a project that is part of Hugging Face, an open-source AI and machine learning resource platform) to release the GPT language model to the public.

I first approached Hugging Face to see how it could help build Bloom, its large language model. I worked with the company as part of the Legal and Ethical Scholarship Working Group and was hired as a full-time Chief Ethicist three months later in May 2022. I have been working there ever since.

60% of my time is spent reading academic papers on ethics and technical AI research, or writing my own papers on topics such as ethics charters (corporate ethical guidelines that employees are expected to follow) It is spent on research involving Spend the rest on collaboration.

I am part of a team called ML and Society, where I meet with research scientists, legal counsel, and policy directors once a week to discuss model biases, secure AI applications, and public policy.

Outside of the group, I help out on internal projects such as creating ethics guidelines for Hugging Face’s diffuser library.

We also provide ethical advice and guidance to external projects, such as introducing Stable Diffusion into our platform and researching how to safely deploy AI in healthcare with external collaborators. .

First thing in the morning, spend 30-45 minutes checking Hugging Face’s latest content moderation report. After that, I do research and write new papers until my lunch break.

Around noon, I help my colleagues with specific projects, such as updating Hugging Face’s content moderation policy. Recently, I advised a healthcare client on how to deploy AI ethically by asking guiding questions, such as whether the client has the proper licenses and whether to publish their models.

Depending on the day, you may also be tasked with quickly responding to emergencies on the platform, such as heated debates between users or instances of AI harm. The rest of the day will be spent on research.

The main challenge I face is that sometimes people perceive me as the moral police with the power to decide what is right and wrong. I don’t have all the answers, so this realization can be dangerous. The more I learn about ethics, the less certain I am about what I know.

For example, one journalist consulted Hugging Face after claiming that one of his articles was plagiarized by the company’s language model. Of course I was happy to help, but it’s a difficult situation to deal with.

It’s also unfair to be forced into the spot when something bad happens. Look at how big tech companies have treated their own ethicists. It was a carnage.

With advances in AI, philosophers, and humanities in general, are much needed. Ethicists understand complex questions about our humanity. It also helps us to predict, rather than predict, the ideal future of society. That makes it closer to utopia rather than dystopia.

I hope the tech community takes the role of AI ethicists more seriously.