How AI Image Generators Make Bias Worse
Buzzfeed recently published a now deleted article on what AI thinks Barbies would look like from different countries around the world.
The results contained extreme forms of representational bias - including colourist and racist depictions, which is something that AI image generators are often prone to doing.
With AI image generators like MidJourney, Stable Diffusion, and Dall-E gaining huge popularity, it’s important that we are vigilant about the forms of bias that these technologies can fuel.
* * * *
This video was inspired by an LIS undergraduate student’s end of first year project, ‘Beyond the Hype: Understanding Bias in AI and its Far-Reaching Consequences’ by Ana Howard.
Presentation Link: https://www.youtube.com/watch?v=aWIKtz-XyAw&t=0s
In the third term of each year, students select a complex problem they personally feel passionate about. They then apply the skills they’ve learnt in interdisplinary thinking, including qualitative and quantitative research methods, in order to tackle their topic and unearth original insights.
At LIS - The London Interdisciplinary School, we believe that solutions to the world's most complex problems won't come from a single specialism. We need to bring together knowledge and expertise from across the arts, sciences and humanities.
If you're interested in our unique interdisciplinary approach to higher education, explore our degree offerings today:
Bachelor of Arts and Science (BASc) Degree - https://www.lis.ac.uk/undergraduate-d...
Master's of Arts and Science (MASc) Degree - https://www.lis.ac.uk/graduate
* * * *
Video Chapters:
0:00 Intro
0:46 Bias in Job Representation
1:55 Barbie Bias
2:24 Generative Adversarial Networks
3:09 Negative Feedback Loops
4:08 How do we stop bias from getting worse?
6:20 Collingridge Dilemma
7:22 Outro
References:
1) Europol paper on Deepfakes and Generative AI - https://www.europol.europa.eu/publica...
2) ‘Humans Are Biased. Generative AI. Is Even Worse’ - Bloomberg Technology Article by Leonardo Nicoletti and Dina Bass - https://www.bloomberg.com/graphics/20...
3) UTK Face Dataset by Zhang, Song & Qi - https://susanqq.github.io/UTKFace/
4) ‘Turing Lecture: Data science or data humanities?’ by Melissa Terras - https://www.youtube.com/watch?v=4yYytLUViI4
5) ‘Corporate Accountability’ BY Lucy Suchman - https://robotfutures.wordpress.com/20...
6) ‘Principles alone cannot guarantee ethical AI’ by Brent Mittelstadt - https://www.nature.com/articles/s4225...
7) ‘Ethics from Within - Google Glass, the Collingridge Dilemma, and the Mediated Value of Privacy’ by Olya Kudina and Peter-Paul Verbeek - https://journals.sagepub.com/doi/10.1...
8) ‘Joy Buolamwini: Examining Racial and Gender Bias in Facial Analysis Software’ by Barbican Centre - https://artsandculture.google.com/sto...
Further Reading & Watching:
’Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence’ by Kate Crawford
‘Weapons of Math Destruction: How Big Data Threatens Increases Inequality and Threatens Democracy’ by Cathy O’Neil
’Coded Bias’ by Dr. Joy Buolamwini
Buzzfeed recently published a now deleted article on what AI thinks Barbies would look like from different countries around the world.
The results contained extreme forms of representational bias – including colourist and racist depictions, which is something that AI image generators are often prone to doing.
With AI image generators like MidJourney, Stable Diffusion, and Dall-E gaining huge popularity, it’s important that we are vigilant about the forms of bias that these technologies can fuel.
* * * *
This video was inspired by an LIS undergraduate student’s end of first year project, ‘Beyond the Hype: Understanding Bias in AI and its Far-Reaching Consequences’ by Ana Howard.
Presentation Link: https://www.youtube.com/watch?v=aWIKtz-XyAw&t=0s
In the third term of each year, students select a complex problem they personally feel passionate about. They then apply the skills they’ve learnt in interdisplinary thinking, including qualitative and quantitative research methods, in order to tackle their topic and unearth original insights.
At LIS – The London Interdisciplinary School, we believe that solutions to the world’s most complex problems won’t come from a single specialism. We need to bring together knowledge and expertise from across the arts, sciences and humanities.
If you’re interested in our unique interdisciplinary approach to higher education, explore our degree offerings today:
Bachelor of Arts and Science (BASc) Degree – https://www.lis.ac.uk/undergraduate-d…
Master’s of Arts and Science (MASc) Degree – https://www.lis.ac.uk/graduate
* * * *
Video Chapters:
0:00 Intro
0:46 Bias in Job Representation
1:55 Barbie Bias
2:24 Generative Adversarial Networks
3:09 Negative Feedback Loops
4:08 How do we stop bias from getting worse?
6:20 Collingridge Dilemma
7:22 Outro
References:
1) Europol paper on Deepfakes and Generative AI – https://www.europol.europa.eu/publica…
2) ‘Humans Are Biased. Generative AI. Is Even Worse’ – Bloomberg Technology Article by Leonardo Nicoletti and Dina Bass – https://www.bloomberg.com/graphics/20…
3) UTK Face Dataset by Zhang, Song & Qi – https://susanqq.github.io/UTKFace/
4) ‘Turing Lecture: Data science or data humanities?’ by Melissa Terras – https://www.youtube.com/watch?v=4yYytLUViI4
5) ‘Corporate Accountability’ BY Lucy Suchman – https://robotfutures.wordpress.com/20…
6) ‘Principles alone cannot guarantee ethical AI’ by Brent Mittelstadt – https://www.nature.com/articles/s4225…
7) ‘Ethics from Within – Google Glass, the Collingridge Dilemma, and the Mediated Value of Privacy’ by Olya Kudina and Peter-Paul Verbeek – https://journals.sagepub.com/doi/10.1…
8) ‘Joy Buolamwini: Examining Racial and Gender Bias in Facial Analysis Software’ by Barbican Centre – https://artsandculture.google.com/sto…
Further Reading & Watching:
’Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence’ by Kate Crawford
‘Weapons of Math Destruction: How Big Data Threatens Increases Inequality and Threatens Democracy’ by Cathy O’Neil
’Coded Bias’ by Dr. Joy Buolamwini
