Artificial Intelligence: Readings and Resources

Where is AI in the Everyday?

AI chats and generative search results
Image, video and text generation tools
Grammar checkers, automatic proofreading tools, and email reply suggestions
Voice-to-text and voice commands for virtual assistants or smart speakers
Social media feeds.- both what content is boosted and some of the content (images, videos)

Key Terms

Artificial Intelligence (AI): a broad category describing a computer's ability to mirror human reasoning and mimic meaning-making
Predictive AI: AI that is trained to categorize items or sort outcomes - like image recognition software
Generative AI: AI that produces human-like responses in text, images, audio, or a combination of these mediums, based on prompts
Machine Learning (ML): the process by which a computer learns autonomously. It records patterns in training data, and then applies those patterns to new data.
Algorithms: Fixed processes, developed by human programmers, that make AI and ML possible.

Ethical Issues

Original creators (writers, visual artists, and others) whose work went into the AI model are often not credited or compensated.
These tools rely on massive amounts of invisible human labor to label initial data
AI results often sound very confident, whether or not results are accurate, which clan lead to confusing and/or convincing misinformation
Adverse environmental impacts such as water and electricity consumption

What about Privacy?

Generative AI tools save every interaction you have with them, and those conversations may be used to train future models, or sold to data brokers. They may store those conversations along with data associated with your user account.

How does it work?

The programming that produces artificial intelligence products is complicated! But as laypeople, we can understand it as being, at its core, a process of finding and replicating patterns that already exist.

Machine learning uses several steps of high-powered statistics, and then human reinforcement, to go from the initial data to something that produces the desire outcome - as the comic artist Randall Munroe satirizes in the panel here, published in 2017.

https://xkcd.com/1838/ “Machine Learning”