Accountability in Generative AI
What’s common between packaged food, fancy clothes and GenAI? Well, the brands in the first two are grown up and the AI folks are still kidding themselves.
The recent POV published by Vice is pretty damning by saying “Companies don’t care”. The experts have been concerned about data accountability since 2020 (or before), and the most recent Sentinel research by FTC shows the consumers are increasingly getting frustrated with the lack of it. So, @chivukulasuresh and I took a quick look at accountability needs from a typical consumer’s perspective. Read on!
If you’re buying packaged food, it is important that ingredients are clearly mentioned. Any “misinformation” or “omission” can be dangerous to the consumer. Similarly, if a faulty ingredient is used, the end-product is considered “not fit for sale”. If the workers don’t follow proper safety procedures, production is affected. Everyday, millions of products and tons of packaged food is processed, scrutinised and safely distributed in every country according to their regulations and food safety standards. When things go wrong, brands can’t blame their logistics providers, packaging facilities, food factories, or farms. They have to recall and rectify. It’s their responsibility as the company that sells to consumers. These standards apply even when you’re distributing food for free.
Similarly, the premium apparel we wear goes through a long value chain: farmers, manufacturing facilities, shippers, warehouses, distributors to retail stores. Everyday, ~275 million garments go through this chain. The brands that sell the garments are accountable for the value chain, whether they like it or not. Ask H&M or every other fashion brand for their stories.
Information from AI products (particularly GenAI services) is exactly that. It’s like the packaged food we consume or clothes we wear. Except that it is digital in nature, fluid in state and has a longer shelf-life (like plastic, it’s there once unleashed). The makers of these services are obligated to explain what goes into it, or if there are bad ingredients like the usage of biased or incorrect info. They are just as responsible and can’t get away with jargon & obfuscation. There are quite a few emerging standards by commercial and governmental bodies alike. Companies just need to care enough to simplify and implement them in a way that works for consumers.
Everyone wants GenAI magic, but wants it done right.