Andrea Matwyshyn
banner
andreamm.bsky.social
Andrea Matwyshyn
@andreamm.bsky.social
Professor, Penn State Law & Engineering
Me: andreamm.com
My labs at Penn State: pilotlab.org & manglonalab.org
@[email protected]
On the Internet, everyone knows you're a dog.
December 12, 2025 at 5:37 AM
Reposted by Andrea Matwyshyn
Well of course... Wait til they encounter hooked turns & trams in Marvellous Melbourne
December 6, 2025 at 2:24 AM
Botsplaining: When you instruct an AI tool to eliminate all sycophancy, but it still defaults to wasting your time with vaguely patronizing replies, e.g. "You have correctly identified a problem!," before it attempts to correct what it did wrong.
a monkey is sitting at a desk with a laptop computer .
Alt: A monkey is sitting at a desk with a laptop computer that it shoves away from itself in disgust, channeling its dissatisfaction with the machine's proximity in what might be a metaphorical commentary on the state of the modern technology economy.
media.tenor.com
December 6, 2025 at 2:53 AM
Reposted by Andrea Matwyshyn
I think about this tweet a lot
November 30, 2025 at 4:23 AM
Reposted by Andrea Matwyshyn
last night at a bar I learned that they improved pinball by making machines that are entirely computers and they’re much better than what we had before
November 16, 2025 at 6:43 PM
Reposted by Andrea Matwyshyn
California's Judicial Council "two weeks ago issued guidelines requiring judges and court staff to either ban generative #AI or adopt a generative AI use policy by Dec. 15": calmatters.org/economy/tech... #ethics #law #courts #tech
California issues historic fine over lawyer’s ChatGPT fabrications
The court of appeals issued an historic fine after 21 of 23 quotes in the lawyer's opening brief were fake. Courts want more AI regulations.
calmatters.org
October 4, 2025 at 12:26 AM
Reposted by Andrea Matwyshyn
FTC vs Amazon opens this week over dark patterns - design tricks that make you accidentally subscribe and struggle to cancel. "The question is when design crosses the line where a reasonable consumer doesn't have a fair shot of understanding what's going on" says @andreamm.bsky.social bit.ly/4mvJV2f
The 'dark patterns' at the center of FTC's lawsuit against Amazon
This week, the trial starts in a consequential FTC lawsuit against Amazon. The suit alleges that Amazon for years "tricked" people into buying Prime memberships that were purposefully hard to cancel.
www.npr.org
September 25, 2025 at 4:17 PM
Reposted by Andrea Matwyshyn
Today is the 30th anniversary of Hackers
September 15, 2025 at 11:56 PM
Reposted by Andrea Matwyshyn
LINK ROT: 38% webpages that existed in 2013 were no longer available 10 years later.

Even among pages that existed in 2021, 22% no longer accessible just two years later. This is often because individual page was deleted or removed on otherwise functional website.

Many implications for knowledge 🧪
September 14, 2025 at 11:19 PM
Reposted by Andrea Matwyshyn
This week's comic, inspired by a study showing that AI summaries on top of search results are dramatically cutting traffic to websites.
August 30, 2025 at 6:32 PM
Have you ever said to yourself "Gee, I wonder what it would have sounded like if an early management team at Compaq had performed a version of the 1985 Chicago Bears Superbowl Shuffle?" If so... www.youtube.com/watch?v=KB6M...
PaqRap - Compaq VPs, 1990
YouTube video by CompaqVet
www.youtube.com
August 23, 2025 at 9:31 PM
Reposted by Andrea Matwyshyn
There are solid use cases for AI. But they’re not the flashy gen AI apps that get investors excited.
“MIT found the biggest ROI in back-office automation—eliminating business process outsourcing, cutting external agency costs, and streamlining operations.” Not sexy, but this is where AI can really make people’s lives easier—by building interfaces with crappy software
fortune.com/2025/08/18/m...
MIT report: 95% of generative AI pilots at companies are failing
There’s a stark difference in success rates between companies that purchase AI tools from vendors and those that build them internally.
fortune.com
August 19, 2025 at 6:14 PM
Reposted by Andrea Matwyshyn
Incredible new privacy risk: Redditor recounts doing a job interview over Zoom with an AI notetaker on the call (Fathom). After the applicant left the call, the interviewers stayed on to discuss/bash her. She then got their post-call remarks in the AI summary. www.reddit.com/r/interviews...
I was accidentally sent the interviewers’ feedback of me after I logged off our Zoom call!
www.reddit.com
August 19, 2025 at 7:40 PM
Reposted by Andrea Matwyshyn
Yet another hallucinated citations case. Sanctions this time include:

* lawyer kicked off the case
* lawyer must notify all judges in the lawyer's other cases nationwide
* lawyer must write to judges who were falsely cited
* lawyer reported to state bar

www.govinfo.gov/content/pkg/...

🤖😵
August 16, 2025 at 8:43 PM
"Employed at the Post Office Research Station in London, which had the world’s first programmable computer... Tiring of the obstacles women faced, she set up her software company with an initial investment of £6." on.ft.com/3UsOi2J Stephanie Shirley, IT pioneer and philanthropist, 1933-2025
Stephanie Shirley, IT pioneer and philanthropist, 1933-2025
As founder of one of Britain’s pathfinding technology companies, she paved the way for women in the field
on.ft.com
August 16, 2025 at 6:41 PM
"We found that routine exposure to AI in colonoscopy might reduce the [adenoma detection rate] of standard, non-AI assisted colonoscopy...suggest[ing that] AI exposure might have a negative impact on patient-relevant endpoints in medicine in general." www.thelancet.com/journals/lan...
Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study
Continuous exposure to AI might reduce the ADR of standard non-AI assisted colonoscopy, suggesting a negative effect on endoscopist behaviour.
www.thelancet.com
August 16, 2025 at 2:34 AM
Reposted by Andrea Matwyshyn
Didn’t want to believe this was true. It’s true.
August 7, 2025 at 9:13 PM
Now do DEFCON 33.
DEFCON refers to any one of five levels of readiness used by the U.S. military.

The word is a blend of 'defense' & 'condition.'

DEFCON 5 is used for the lowest perceived threat, and DEFCON 1 for the highest.
August 4, 2025 at 3:46 AM
Reposted by Andrea Matwyshyn
📢 Vacancy alert: @ivir-uva.bsky.social is looking for a postdoc in the Human(e) AI Research Priority Area. Take a look at the details: https://bit.ly/3ItbqLu. Make sure you apply before 20 August! cc @natalihelberger.bsky.social #Vacancies #Jobs
Vacancy — Postdoctoral Research Position in Human(e) AI
The Research Priority Area in Human(e) AI at the University of Amsterdam currently has a vacant legal Postdoctoral Researcher position.  
bit.ly
July 7, 2025 at 12:00 PM
"A Florida jury on Friday found that flaws in Tesla’s self-driving software were partly to blame for a crash that killed a 22-year-old woman..The jury verdict...would require Tesla to pay as much as $243 million" www.nytimes.com/2025/08/01/b...
Jury Says Tesla Was Partly to Blame for Fatal Crash
www.nytimes.com
August 4, 2025 at 2:24 AM
"Discussions about technology for its own sake, without considering its societal consequences, are impossible." CCC, April 6, 2025 www.ccc.de/en/club/unve...
CCC | Incompatibility of membership
Der Chaos Computer Club ist eine galaktische Gemeinschaft von Lebewesen für Informationsfreiheit und Technikfolgenabschätzung.
www.ccc.de
July 19, 2025 at 4:46 PM
Reposted by Andrea Matwyshyn
Oh, sweet merciful fire, two notes:

(1) Even Latham & Watkins, ya'll.

(2) The lawyer defending an AI company for using stolen IP to train their AI/LLM used the company's AI/LLM to cite check her brief, which hallucinated results.

fingfx.thomsonreuters.com/gfx/legaldoc...
May 15, 2025 at 7:49 PM
"FBI warned on Thursday that a malicious messaging campaign has been targeting government officials and their acquaintances by sending voice messages generated with artificial intelligence" thehill.com/regulation/n...
FBI issues warning about AI voice impersonations of US officials
The FBI warned on Thursday that a malicious messaging campaign has been targeting government officials and their acquaintances by sending voice messages generated with artificial intelligence (AI) …
thehill.com
May 15, 2025 at 8:35 PM
"After years of depicting Klarna as an AI-first company, the fintech’s CEO reversed himself, telling Bloomberg the company was once again recruiting humans after the AI approach led to 'lower quality.'" fortune.com/2025/05/09/k...
As Klarna flips from AI-first to hiring people again, a new landmark survey reveals most AI projects fail to deliver
Just 1 in 4 AI investments bring in the ROI they promise—but CEOs just can’t resist the technology.
fortune.com
May 13, 2025 at 4:29 PM
"An email notification system used by U.S. federal and state government departments to alert residents to important information, has been used to send scam emails...Indiana said a contractor’s account was hacked and used to send the scam messages." techcrunch.com/2025/05/13/g...
Government email alert system GovDelivery used to send scam messages | TechCrunch
The state of Indiana attributed the scam emails to a compromised contractor's account.
techcrunch.com
May 13, 2025 at 3:53 PM