I help organizations improve online and offline experiences | UX Researcher | Service Designer | UX Designer | Speaker | Author | Ethical Research & Design | 3Q-DO NO HARM Framework| I ❣️Books & Libraries | Remote First
Continuing with Beware of AI Hype & Harm . . . #3. Are We All Unpaid Workers in the AI Value Chain? These large language and image to text models like ChatGPT & Lensa AI are trained on large amounts of data scraped from books, webpages, and other source. We didn’t give consent to use some of the data and companies are promoting AI Utopia (where all hard tasks can be automated) to entice us to make these AI tools better. Consider these issues: · Unauthorized Use of Images: Consider the case of one woman who found private medical photos of her face were used in a LIAON dataset used to train Stable Diffusion and Google Imagen. As a copy of her release shows, she only authorized the use of her photos in her file only and not shown to anyone. (1) · Copyrights: Several artists in a class action suit and Getty Images separately have sued Stable AI, the companies behind a popular AI image generator, for copyright infringement by scraping their copyrighted images from the internet for their datasets. Some images produced from its models have the Getty Images’ watermark on them. These lawsuits raise new questions about the legality of this. (2) · Collecting Biometric Data: Lensa AI faces a new lawsuit over collecting facial data (biometrics) through its image scraping without consent under Illinois’ 2008 Biometrics Information and Privacy Act. (3) · Underpaid and Traumatized: According to a Time article, workers tasked with labeling graphic depictions of violence and abuse were paid as little as $2 hour for nine-hour shifts in Kenya. This would helped OpenAI be more marketable because it was less toxic. Workers called dealing with this toxic content “torture” and despite promises of “wellness counselors” they either unhelpful or not available as promised.(4) · Free Work: Last, but not least when we use these models, we provide more data for them to train on. From our initial asks through all our refinements needed to produce a helpful answer, we are modeling language even when the chatbots responses are wrong. This helps train the model which improves its performance and increases its value in the marketplace. - Are our rights protected in the AI value chain? - Do we want to be workers in the AI value chain? - Are we getting enough benefit from the AI value chain to work for free? ✨ ✨ ✨ 👋🏽Hi, I’m Lisa D. Dance, a UX Researcher, who helps businesses be easier to do business with through my consultancy ServiceEase. ✅I'm open to speaking, workshop, and UX Audits engagements. Key topics include: #ux #technology #customerexperience #aiethics
