Maintaining the ethical case against generative AI

Workshop review

After Kyle Wohlnut’s fascinating presentation on the Ethics of AI at METM25 in October, I was eager to hear what he had to say at this two-hour webinar. The information was no less shocking the second time around, and there were certainly enough updates on the subject to make it worth my time.
 

Kyle’s main premise was this: the very foundations of these systems are unethical, and they are causing some very real-world harms.

His first thoughts were about the dangers of anthropomorphizing AI: a collection of systems and coding that uses huge datasets to extract probabilistic correlations is much more marketable if it is described as a thinking, calculating, plotting thing.

Covering topics such as IP theft, labour abuses, the environment, cultural bias and human rights, he demonstrated how AI is more about accruing oligarchic power than making the world a better place. Simply put, for OpenAI and Microsoft, the very definition of “general artificial intelligence” is when the systems are “generating at least $100 billion in profits”.

Extensive research and references were included with each slide to support Kyle’s arguments. Ominous phrases such as “human fracking” described the labour abuses happening to exploited workers in the Global South, and “water bankruptcy” defined a world that is using water faster than nature can replenish it.

Meanwhile, the probabilistic behaviour of these tools solidifies cultural biases, propagating harmful stereotypes and spreading them to new languages and cultures, not to mention other real-world harms such as using AI for weapons, widening global inequality, and dangers to the economy with massive circular investment deals.

While we are anthropomorphizing the machines, we are simultaneously devaluing the human condition by using humans as a service.

Thankfully, Kyle had some advice for the future. As well as growing resistance in the form of boycotts and manifestos, an AI clean-up market is taking shape which is changing perceptions to the technology. Lower AI literacy tends to predict greater AI receptivity, so the best way to talk to clients is to provide them with more information. Explain that AI-generated products can’t be copyrighted, enquire if they have a corporate sustainability statement, ask them why the simplest answer would not be better than their tech solution. And remember to use the correct language. After all, that is a power we still have.