This piece about the UK liver transplant matching algorithm, by @randomwalker and @sayashk, is well worth a read. Excellent example that algorithms that make real-world decisions aren't inherently problematic, but they can certainly be horrifyingly bad if not designed (and tested, and transparently published, and audited) carefully.
https://www.aisnakeoil.com/p/does-the-uks-liver-transplant-matching

interesting new paper on the feasibility of using LLM agents for computational reproducibility checks by Narayanan and colleagues @randomwalker

https://arxiv.org/abs/2409.11363

Kirkus Reviews, which provides early book reviews to the publishing industry, has given AI Snake Oil a very positive "starred" review, which we're told is rare and kind of a big deal. Honored and grateful! https://kirkusreviews.com/book-reviews/arvind-narayanan/ai-snake-oil/
Preorder:
https://www.amazon.com/Snake-Oil-Artificial-Intelligence-Difference/dp/069124913X
https://bookshop.org/p/books/ai-snake-oil-what-artificial-intelligence-can-do-what-it-can-t-and-how-to-tell-the-difference-arvind-narayanan/21324674
More preorder links at the bottom of this post
https://www.aisnakeoil.com/p/ai-snake-oil-is-now-available-to
Coauthored by @sayashk, published by @princetonupress.

I'm ecstatic to share that preorders are now open for the AI Snake Oil book! The book will be released on September 24, 2024.

@randomwalker and I have been working on this for the past two years, and we can't wait to share it with the world.

Preorder: https://princeton.press/gpl5al2h

‘Will AI transform law? The hype is not supported by current evidence’, write @randomwalker & @sayashk https://www.aisnakeoil.com/p/will-ai-transform-law They also published a scholarly paper on the topic with Peter Henderson: https://www.cs.princeton.edu/~sayashk/papers/crcl-kapoor-henderson-narayanan.pdf #law #ai #tech #chatgpt

Most online speech is hosted on algorithmic platforms designed to optimize for engagement. But algorithms are not neutral. Read other essays in our "Algorithmic Amplification & Society" project series, in collaboration with @randomwalker. Learn more here:
https://knightcolumbia.org/research/algorithmic-amplification-and-society

Excited to share that we’ve started publishing the essays from “Optimizing for What? Algorithmic Amplification and Society,” our spring symposium organized with @randomwalker. Here’s a brief intro. by
@kgb. Links to the first two essays follow.
https://knightcolumbia.org/blog/exploring-algorithmic-amplification-a-new-essay-series

The "ChatGPT has a liberal bias" paper has at least 4 *independently* fatal flaws:
– Tested an older model, not ChatGPT.
– Used a trick prompt to bypass the fact that it actually refuses to opine on political q's.
– Order effect: flipping q's in the prompt changes bias from Democratic to Republican.
– The prompt is very long and seems to make the model simply forget what it's supposed to do.
By @sayashk and me, summarizing our analysis and a separate one by Colin Fraser. https://www.aisnakeoil.com/p/does-chatgpt-have-a-liberal-bias

@yingtai @karihoffman @elduvelle @vicgrinberg

Yup! https://threadreaderapp.com/thread/1199301468941160450.html

@hjonker Yes, the project ended a few years ago. The website downtime is unintentional though; I plan to redirect it to an archive.org version. Sorry about that.

@tuliotec With modern mobile OSes, surreptitious eye tracking is not technically possible. It's also a legal risk that IMO outweighs the benefits.

But face analysis has been used for recommendations https://twitter.com/MarcFaddoul/status/1232014908536938498

Other biometrics like gait and activity recognition are also used, but not for recommendations AFAIK.

Use of location data is ubiquitous, of course.

Hope that helps!

The amount of misinformation on Mastodon around Threads and the EU is a great demonstration of how motivated reasoning is not a problem only for commercial social media platforms.

It's been six months since Elon took over Twitter. I have some thoughts on the "Twitter diaspora" and the current decentralized alternatives: https://www.techdirt.com/2023/04/28/six-months-in-thoughts-on-the-current-post-twitter-diaspora-options/

Ambulances can’t reach patients before they die. Fire trucks can’t get through and house fires blaze. Pedestrians trying to cut through trains have been disfigured, dismembered and killed; a Pennsylvania teenager lost her leg hopping between rail cars as she rushed home to get ready for prom.

In Hammond, the hulking trains of Norfolk Southern regularly force parents, kids and caretakers into an exhausting gamble: How much should they risk to get to school?

https://www.propublica.org/article/trains-crossing-blocked-kids-norfolk-southern

This is the latest in the AI Snake Oil book blog by @sayashk and me. Writing this blog alongside the book has been really fun. I'll probably do something like this for all future books! Thank you to everyone who subscribed. https://aisnakeoil.substack.com/

@RWerpachowski@mastodon.green It's a great post but we're making a different point and I don't think there's any contradiction. We're not contrasting RL with supervised, but rather any type of post pre-training intervention (sorry, is there a better term for that?) with something that can maybe prevent implicit biases in the first place (e.g. a data intervention).

OpenAI mitigates ChatGPT’s biases using fine tuning and reinforcement learning. These methods affect only the model’s output, not its implicit biases (the stereotyped correlations that it's learned). Since implicit biases can manifest in countless ways, OpenAI is left playing whack-a-mole, reacting to examples posted on social media.

People have been posting glaring examples of ChatGPT’s gender bias, like arguing that attorneys can't be pregnant. So @sayashk and I tested ChatGPT on WinoBias, a standard gender bias benchmark. Both GPT-3.5 and GPT-4 are about 3 times as likely to answer incorrectly if the correct answer defies gender stereotypes — despite the benchmark dataset likely being included in the training data. https://aisnakeoil.substack.com/p/quantifying-chatgpts-gender-bias

We're at over 1,000 registrations for the @knightcolumbia algorithmic amplification symposium this Friday & Saturday. We're lucky to have an all-star cast of speakers. In-person registration is closed/waitlisted, but you can still register to attend online: https://knightcolumbia.org/events/optimizing-for-what-algorithmic-amplification-and-society

At @knightcolumbia I'm co-organizing what I think is the first symposium on the topic of algorithmic amplification on social media — April 28/29 (NYC and online). I'm told we have over 600 registrations already. We'll be moving to a waitlist for in-person participation soon. Register to hear from ~30 leading thinkers on the topic: https://knightcolumbia.org/events/optimizing-for-what-algorithmic-amplification-and-society