Summary
The article discusses challenges with AI image generation, particularly when representing people with disabilities. Jessica Smith, an Australian Paralympic swimmer, successfully prompted an AI to generate an accurate image of her without one arm, highlighting progress in AI's inclusivity. However, issues with bias remain, as shown by Naomi Bowman's experience with AI altering her facial features despite specific instructions.
Key Facts
- Jess Smith, a former Australian Paralympian, tried using AI to generate an image of herself lacking a left arm below the elbow.
- Initially, the AI struggled to create an accurate image, often depicting her with two arms or a prosthetic.
- Jess later found the AI could generate an accurate image of her, indicating improvements in the AI's inclusivity.
- OpenAI, the company behind ChatGPT, stated they recently made improvements to their image generation model.
- Naomi Bowman, who is partially sighted, experienced bias when the AI altered her face despite her instructions.
- Concerns over AI bias arise because it reflects societal inequalities, highlighting the need for diverse input at the data creation stage.
- AI's environmental impact is under scrutiny due to the high energy consumption of data centers powering such technologies.
- Experts emphasize the importance of diversity in the teams that build and train AI systems to ensure fair representation of all societal groups.
