If you don’t already listen to The Vergecast podcast, I highly recommend it. They cover the broad intersection of technology and culture in the most engaging way possible. Recently, they had an episode where they discussed the ins and outs of generative artificial intelligence.
What is generative artificial intelligence (generative AI)? In simple terms, it’s a tool that uses artificial intelligence to create content that matches (or comes close to matching) what a human would produce. I’ve typically seen machine learning deployed in this way, but the end result can be an essay, a corporate slogan, a fantasy-like painting, or even a very realistic looking photograph.
Imagine – you want a stock photograph of a professional meeting. Normally, you’d hire a professional photographer and paid actors to capture the scene of well-dressed business types sitting, smiling, and talking around a conference table. With generative artificial intelligence, you can simply type in the scene you want captured, and the AI produces it for you.
The Vergecast episode was great in that it provided a few glimpses into some critical issues about generative AI. They get into what the various AI models are, who is actually creating these tools and models, how the models are trained, what is produced, and the ethics and economics of all the above. There are some really important questions that they highlight, like:
- the veracity of content,
- ownership of training data and the potential infringements on those ownership rights (i.e. copyright infringement),
- liability for making models available versus liability for using them, and
- protections and guardrails around these tools.
It’s all interesting stuff, and there’s so much more that needs to be sorted out over the coming years.
Additional reading: