Help me stay current on our robot overlords

August 22, 2024 9:05 AM

I am a humanities librarian and have the dubious privilege of becoming an accidental generative AI subject matter "expert". I don't expect to be using this technology on the regular anytime soon, but I do teach information literacy skills to peers and users and would like to be equipped to evaluate vendors that are adding LLM functionality to their products. How I keep current on this rapidly evolving technology without drowning in nonsense?

I have been struggling to keep current on LLMs and generative AI (especially because I don't use them in my personal life). Traditional journalism, business/hobbyist/tech spaces, and resources put out by tech companies themselves seem to be a medley of hype, bull, philosophical woolgathering, outdated information, and highly technical journal articles. The demise of Google as a useable search engine is making my self-education even more of a slog.

What websites/organizations/media outlets are disseminating high-quality information about generative AI for a non-specialist audience? Format can be anything - news articles, videos, online courses, subreddits, etc. I am especially looking for resources outside the library/education sectors, though open to anything. I don't mind some technical detail, but I prefer content that doesn't require an advanced degree in machine learning to understand. Thanks!

posted by toastedcheese to Technology (14 answers total)

I reckon you could probably dip into one or more of the newsletters listed here (at least while they last), and have a line on what folks who are really into this stuff are currently talking about : https://guides.lib.uni.edu/issues-of-AI/newsletters-on-AI
posted by aesop at 9:36 AM on August 22

1) Get in the habit of using them in your personal life. Go today to chatGPT and type "How do I keep current on LLMs and generative AI?". Check out the ones that generate music, pictures, videos, programs, etc.

2) youTube. There are 15 minute videos and 1 hr videos, and 10 hr courses. Here's a start:

https://www.youtube.com/watch?v=dDUC-LqVrPU
Has Generative AI Already Peaked? - Computerphile

https://www.youtube.com/watch?v=wjZofJX0v4M
But what is a GPT? Visual intro to transformers | Chapter 5, Deep Learning
posted by at at 9:42 AM on August 22

The Batch, a weekly newsletter from Andrew Ng, the founder of Google Brain.

The Verge has a dedicated AI section with explainers, and I also like Ars Technica -- Benj Edwards seems to be focused on AI these days.

Play around with ChatGPT, Claude AI, and Google's Gemini. There is much more focus on evaluating the models themselves than applications using those models, so you'll have to learn where the blind spots are, which are pretty consistent among products.
posted by credulous at 9:59 AM on August 22

I've been reasonably impressed with some of the resources Common Sense Media has put together. I'll admit I'm perhaps biased because I know and like some of the experts they're working with. The intended audience is parents and educators of K-12 so there's not too much technical detail.
posted by potrzebie at 10:06 AM on August 22

I like this resource from Mike Edwards, which is clear & thoughtful, and from an educator's perspective: ChatGPT for Writing Teachers: A Primer.

I'm also a humanities librarian thinking about this in the context of information literacy instruction, but have found myself much more energized by the ideas, criticisms, and approaches I see from Rhetoric & Composition scholars that I see from our field.
posted by activitystory at 10:15 AM on August 22

The 80,000 Hours podcast has a lot of episodes about AI, if you are up for two-to-three-hour podcast episodes. They are ultimately about the promise and risk of AI, but they talk a lot about specific models and what they can do as well.
posted by catquas at 10:22 AM on August 22

Recommend these two prof/researchers from Princeton's blog. It's clear, critical, and reasonable

AI Snakeoil

"There are four main themes:
Resisting AI doom narratives
Debunking hype about AI’s capabilities and transformative effects
Understanding why AI evaluation is so tricky (which is one root cause of AI hype)
Helping AI policy stay grounded in the evidence."
posted by jebs at 12:02 PM on August 22

Emily Bender, one of the Stochastic Parrots authors, has a podcast. She’s also a good person to follow on BlueSky and Mastodon.

If you’re on TikTok, Morten Rand-Hendricksen is a good follow on tech things, including AI. He also has a website.

I like 404 media for tech news generally. They also have a podcast.
posted by eekernohan at 4:41 PM on August 22

Do you know about Perplexity? It's a good Google substitute: it's a search engine that summarizes its sources, with citations and links! (This doesn't answer your question, but it's a good resource.)
posted by bluedaisy at 5:33 PM on August 22

I've been using LLMs for tech support questions. They seem better at diagnosing a hardware issue or helping find a feature in an online software than the forums.
posted by jander03 at 8:24 PM on August 22

Perplexity is an overt plagiarizer -- yes, even more than the usual run of generative AI thievery. It could be classroom-useful in a discussion of plagiarism, or of techbro exploitation of journalists and creatives.

One useful education-specific AI commentator is Benjamin Riley of Cognitive Resonance (Substack, sorry). He has a summary of issues with AI in education that may be useful for opening the eyes of local lemmings: Education Hazards of Generative AI.

I validate your perception that academic librarians in the main are currently hearteyed lemmings about AI. You might, however, enjoy this non-cataloger's claims that ChatGPT can catalog getting absolutely shanked by actual catalogers in the next issue of the journal.

Me, I am talking through a potential research project with a friend, on the general topic of mania detection in librarianship... we sure did have a raft of witless blockchain and NFT nonsense too.
posted by humbug at 7:42 AM on August 23

Great folks on Mastodon: Emily Bender (mentioned above) and Timnit Gebru (fired from Google for speaking out). Both of them are hosted by DAIR. If you don't have a Mastodon account, it's worth it for them.
posted by hydropsyche at 7:56 AM on August 23

Ben Williamson.

A blog post.

A whitepaper.
posted by humbug at 5:44 PM on August 23

Don't forget about the devastating environmental impacts of these technologies. We should think very carefully about how and when generative AIs are used. In a moment when we should be reducing energy usage and conserving resources, instead we are burning fossil fuels and wasting fresh water to make frivolous inquires from AIs ... let alone the resources mined to make the servers in the first place.

AI’s carbon footprint is bigger than you think
Generative AI’s environmental costs are soaring
The Uneven Distribution of AI’s Environmental Impacts
posted by Mournful Bagel Song at 2:45 AM on August 24