Homepage
Open in app
Sign inGet started

high stakes design

AI is a high stakes design problem

We audited RoBERTa, an open source large language model

We audited RoBERTa, an open source large language model

We found two security vulnerabilities and a backdoor.
Go to the profile of andrea b
andrea b
Dec 19, 2022
Adversarial attacks are only part of the problem, so we also audited the hardware of this AI system

Adversarial attacks are only part of the problem, so we also audited the hardware of this AI system

Go ahead. Bring on the complexity.
Go to the profile of andrea b
andrea b
May 31
Is your AI a “Clever Hans”?

Is your AI a “Clever Hans”?

What a counting horse can teach us about evaluating AI tools & systems.
Go to the profile of andrea b
andrea b
Feb 23
What happened when we audited a deepfake detection tool called FakeFinder

What happened when we audited a deepfake detection tool called FakeFinder

TLDR: This AI tool doesn’t do quite what we thought…
Go to the profile of andrea b
andrea b
Dec 14, 2022
Do deepfakes discriminate?

Do deepfakes discriminate?

Here’s what we learned from bias testing two deepfake detector models.
Go to the profile of andrea b
andrea b
Dec 19, 2022
The high stakes of designing data products

The high stakes of designing data products

We need analysis tools that make data more accessible to less technical audiences. But unfortunately, this is easier said than done.
Go to the profile of andrea b
andrea b
Oct 18, 2019
Alice, Bob & Eve, the eavesdropper

Alice, Bob & Eve, the eavesdropper

A comic about encryption
Go to the profile of andrea b
andrea b
Jan 14
ELIZA, the paperclip maximizer

ELIZA, the paperclip maximizer

A story
Go to the profile of andrea b
andrea b
Dec 14, 2022
10 things I’ve learned about Explainable AI

10 things I’ve learned about Explainable AI

Insights from interviews with technologists & tool builders.
Go to the profile of andrea b
andrea b
Oct 28, 2019
What do people really want when they say they want “Explainable AI”?

What do people really want when they say they want “Explainable AI”?

We asked 60 stakeholders.
Go to the profile of andrea b
andrea b
May 20, 2020
The Risks of Black-Box Decision-Making

The Risks of Black-Box Decision-Making

A conversation with Krishna Gade, co-founder of Fiddler Labs, about accuracy, fairness, and the risks of black-box decision-making.
Go to the profile of andrea b
andrea b
Jan 30, 2020
The Benefit of Simple Models

The Benefit of Simple Models

A conversation with statistician Tommy Jones about the the tradeoffs between simple & complex models.
Go to the profile of andrea b
andrea b
Dec 11, 2019
Confusion Matrices & Interpretable ML

Confusion Matrices & Interpretable ML

A conversation with Nina Lopatina about understanding machine learning & why the brain might be more interpretable than some models.
Go to the profile of andrea b
andrea b
Nov 13, 2019
Explainable AI: What Is It and Who Cares?

Explainable AI: What Is It and Who Cares?

A conversation with IQT’s Peter Bronez about three use cases for Explainable AI.
Go to the profile of andrea b
andrea b
Nov 6, 2019
About high stakes designLatest StoriesArchiveAbout MediumTermsPrivacyTeams