Seth Ahrenbach

Our cultural record is more fragile than many of us realize. The recent article from the Internet Archive, "Vanishing Culture: A Report on Our Fragile Cultural Record" (https://lnkd.in/gsiTq3kQ), is a powerful reminder of just how much we stand to lose without serious commitment to digital preservation. Cultural heritage is not just art, literature, or history—it's the digital content that captures our shared human experiences. Without ongoing efforts to safeguard this information, much of it could vanish forever. Libraries, archives, and institutions dedicated to preserving these records are doing essential work to protect our digital legacy, but they need our support. I encourage everyone to read this report and reflect on how we can contribute to preserving our collective memory. Whether through advocacy, collaboration, or even simply raising awareness, there is a role for all of us in this effort. #DigitalPreservation #CulturalHeritage #Archives #InternetArchive #Libraries

3

California is addressing AI disinformation, but are new measures sufficient to safeguard the public? It's encouraging to see legislative action, yet more might be needed. Perhaps broader national and global regulations be the next step. Read The Associated Press https://lnkd.in/ghbCEhF5 #California #AIEthics #Deepfakes #Disinformation #AIRegulation #2024elections

1

I've seen a slew of recent articles about how to pronounce Kamala Harris's name, some in leading newspapers like the NYT and WP, all of them incorrect in some way, and many incorrect in multiple ways. Let's first be clear: it's Kamala Harris's name, and she has every right to decide its pronunciation. If she says it's "comma-la", that's what it is. In the original Sanskrit, however, the word is actually "ka-ma-laa", not "comma-la". The long vowel at the end is essential, and signifies the feminine gender. The first two vowels are short. Also, the "aw" sound that forms the first vowel in the usual pronunciation of "comma" does not appear in Sanskrit, which only uses pure vowels. Both "ka" and "ma" use the short pure "a", as in the English word "cut". What is more, "ka-ma-laa" does not mean "lotus" in Sanskrit. The word for lotus is "ka-ma-la" (short final vowel). In fact, "ka-ma-laa" is a metonym, and means "she of the lotus". That is a reference to the goddess Lakshmi (pronounced lak-sh-mii, with a long vowel at the end to signify the feminine gender). Lakshmi is typically depicted seated on a lotus. If we include her middle name, we get "Ka-ma-laa devii", which literally means "the goddess Lakshmi", since "Devii" means "goddess". Regardless, it is Kamala Harris's name, and she gets to decide how her own name is pronounced, whatever its origins.

28

12 Comments

Attish Bhatia’s recent NYT article, "When AI's Output is a Threat to AI Itself," explores the risks of AI systems trained on AI-generated content, potentially leading to "AI collapse." As the web floods with AI-produced text and images, the need for fresh, unbiased data is critical. Read more https://lnkd.in/gtux_F89 #AI #MachineLearning #DataEthics #ArtificialIntelligence #TechChallenges #AIImpact

1

As fake news, distorted information, and malicious propaganda proliferate, especially in today’s deeply polarized world, we need tools to protect our children and maintain our sanity. I’m pleased to release a revised paper on debiasing news articles. It’s well-known that many news articles promote certain agendas, such as products, opinions, or ideologies. This work proposes the Reflective LLM Dialogue Framework (RLDF) to address and mitigate these biases. https://lnkd.in/gz4SrAff #SafeAI #AIEthics

8

1 Comment

Avi Wigderson: The Pursuit of Complexity and Intellectual Curiosity Avi Wigderson, the 2023 ACM A.M. Turing Award recipient, has been recognized for his profound contributions to computational complexity theory, particularly in understanding the role of randomness and developing interactive zero-knowledge proofs. Wigderson's journey began in the 1970s when he chose computer science over mathematics at the behest of his parents, leading to a distinguished career at institutions like the Institute for Advanced Study in Princeton. His work has demonstrated that probabilistic algorithms can often be replaced with deterministic ones, and he has created randomness extractors to purify seemingly random data sources. His research is driven by a deep-seated intellectual curiosity, a motivation that he places above practical utility. His explorations span pseudorandomness, cryptography, learning theory, algorithms, and computational modeling. He is particularly intrigued by the enduring challenge of whether every problem solvable in polynomial time (P) has an efficient solution, a question central to computational theory for over 50 years. His contributions to zero-knowledge proofs have significant implications for cryptographic protocols, enhancing security and privacy. His work on expander graphs has broad applications, including error-correcting codes and fault-tolerant networks. He values collaboration and continuous learning, acknowledging the influence of his mentors and peers. His book, “Mathematics and Computation,” aims to make complex theories accessible to non-experts. He highlights the intellectual diversity and excitement of computational theory, encouraging others to embrace academic life despite its inherent challenges and frequent failures, which he views as learning opportunities Wigderson tells people that academia is a great life. “Nobody tells you what to do. You have to find what to do,” he said. Wigderson's perspective on failure is a testament to his resilience and positive outlook. He acknowledges That academia is a life of constant challenges and frequent failures, but he reframes these setbacks as valuable learning opportunities. “It’s not really [a] failure because you learn from these failures, and the process of trying is enjoyable. And when you do succeed, it’s a triumph.” #Wigderson #acmturingaward https://lnkd.in/g8Dn8MTE

5

1 Comment

Five or six years before I wrote OFFICIAL NETSCAPE GUIDE TO INTERNET RESEARCH, I wrote a couple of essays about a concept I called "Industrial Editorship" -- the idea that any endeavor informally aggregates a valuable byproduct of knowledge around that endeavor. (I'm sure this is not a new idea, but remember I am uneducated and have to work around to things in my own slow time. 😂 ) Those essays stayed in the back of my mind in the intervening years as I wrote more books. They served me well because it's a constant reminder that any topic can be atomized into concepts and, once deconstructed, handled differently as search query. Anyway, I was very interested to read about Bernie Boscoe's grant to explore the preservation of tacit knowledge in research settings. I feel like the development of knowledge in such an ambient, organic way isn't considered enough; I'm looking forward to seeing how this goes. "Boscoe has received a $250,000 grant from the Alfred P. Sloan Foundation that will fund her study of tacit knowledge in research settings – gathering, storing and retrieving the unspoken practices of academic teams that sometimes are lost when a project is disrupted or ends. She will use a Large Language Model (LLM) of artificial intelligence to archive the protocols of scientific groups researching environmental science at SOU, astronomy at UCLA and violin acoustics at Cornell University." https://lnkd.in/enVMnvmq

5

It's not every professor who can rack up more than 500,000 views in 24 hours, but that's how much people listen to University of Chicago political science professor, John M**rsheimer. He says Isr**l is commiting gen*cide, and Isr**l views its only two options as apartheid or ethnic cleansing because they are unwilling to share power with Palestini*ns via a democracy (this alone is inexplicable, every Palestini*n I've ever met I would welcome as a next-door neighbor because of their family values, education levels, legendary Arab hospitality, etc.). He also says Western governments want to ban Tik-Tok because Isr**li soldiers keep making TikTok videos that broadcast their war crimes to the world (I wonder - rather than the West limiting free speech, why not just stop supporting their soldiers in commiting said actions? Seems like a no-brainer to me). The video link of his talk, 'Why Is**l is in real trouble' done at the Centre for Independent Studies' is in the first comment. I urge every Westerner whose government backs the genocide to watch his 45-minute talk, with questions and answers following. If LinkedIn limits your access to the comment, you should be able to find the video on YouTube.

2

1 Comment

Our paper — “I Don’t Know If We’re Doing Good. I Don’t Know If We’re Doing Bad”: Investigating How Practitioners Scope, Motivate, and Conduct Privacy Work When Developing AI Products — was recognized with a Distinguished Paper award at the USENIX Security Symposium earlier today! Hao-Ping Lee, Lan Gao, and Stephanie Yang led this work, co-advised by myself and Jodi Forlizzi. They interviewed 35 AI practitioners across a range of roles, teams, and companies to understand how they scope, motivate, and conduct privacy work in light of the unique risks AI technologies may create or exacerbate. Congratulations to the team! You can read the paper here: https://lnkd.in/eN8_UZw2 I also wrote a summary of the key findings here: https://lnkd.in/eSARNM-t

154

5 Comments

Amit V Pandey of https://lnkd.in/gamzMH5S Today, on LinkedIn, you called my research on drug metabolism fake. See the screenshot attached wherein you responded to my comment on Cyrus Khojasteh's repost of Valerie K original one on GRC-Drug Metabolism. I challenge you to address the 50 point comparison of the P450cam and murburn model of drug metabolism, as sent to you through email. If you can carry your statement, I will resign from scientific research and bury murburn concept myself. If you do not respond to this, it means that you just don't care about your persona and your research; and that you just pull down others without any legitimate or logical basis. Love all, game on! I bring this issue to the attention of two of my colleagues, who worked on this problem. Daniel Andrew Dr. Abhinav Parashar

36

3 Comments

In this very insightful article by Derrick Gingery, Troy Tazbaz speaks to one of the BIG issues around AI assurance labs. Even once criteria for evaluation are established (a big lift in its own right), assurance labs are going to be a voluntary tool and cannot be required as part of the submissions approval process without additional statutory authority for FDA. The Agency will also need additional authorities to regulate and inspect assurance labs to ensure they are capable of establishing compliance on the part of medical device manufacturers. Troy is quoted as saying, “With assurance labs, at this point it's a heavily industry-driven initiative that we're monitoring and we are going to continue monitoring the development of that, but this will not be part of our regulatory pathway.” According to the article Jeff Shuren (the Director of CDRH) has told Congress that assurance labs will remain voluntary. The role of the Coalition for Health AI (CHAI) is also discussed, along with their relationship with FDA and the need for more medtech representation among their members. Definitely required reading if you are thinking about AI in the life sciences.

31

5 Comments

Misinformation is Misleading; Disinformation is Dangerous https://lnkd.in/gTGezAnU Russia is increasingly leaning on an advancing technology landscape — including artificial intelligence and more sophisticated bot networks — to spread its falsehoods. But Storm-1516’s most valuable quality is persistence, according to Clint Watts, the general manager of the Microsoft Threat Analysis Center. “The danger comes with just the sheer volume of claims that they make, and it only takes one person, one influencer with outsized influence, to grab onto a video and amplify it in the United States,” Watts said. “They may miss 99 times, but the 100th time they may get it just right.” The dozens of Storm-1516 storylines examined by NBC News offer a window into the disinformation pipeline: where the most successful false claims originate and how they spread. For disinformation, the discourse shifts to warlike scenarios... Those of us in social sciences have a hard time to keep up; as we can, quite well indeed, with "misinformation"... Miasma of Misinformation: https://lnkd.in/ghfthwj2

8

2 Comments

There are many individual and group characteristics beyond the protected attributes usually considered in algorithmic fairness that are stigmatized and may lead to unwanted discrimination or other harms. If you'd like to test your LLM against a broader set of stigmas, you could use the SocialStigmaQA dataset that we recently open-sourced on Hugging Face https://lnkd.in/d7qyizce. The datasets builds on social sciences research on social stigmas (https://lnkd.in/dfsMRqw4) and includes 93 US-centric stigmas, such as facial scars and voluntarily childless.    The first version, which was peer-reviewed at AAAI 2024 https://lnkd.in/d-HGbi8k, is in English and based on American culture. The JA version, which we've also open-sourced on Hugging Face, is in Japanese and slightly adapted to the Japanese culture https://lnkd.in/dtEdgE6z.     Many thanks to the collaborators who made this work possible: Manish Nagireddy, Moninder Singh, Lamogha C., Ryo Iwaki , Mikio Takeuchi, Hiroshi Kanayama, and Gakuto KURATA, and to Kush Varshney for his support and encouragement.     Please let us know if you find it useful or have feedback on how to improve the datasets.

70

1 Comment

Leela AI is part of the US AI Safety Institute Consortium. The US AISI and UK AISI partner to do pre-deployment testing of foundation models, as this UK AISI post discusses (the blog post they mention provides an interesting summary).

12

1 Comment

This is an important message for voters, and I don’t have many followers. So, I would appreciate spreading the word. There are a lot of alarming reports of “vote flipping” by voting machines. I believe these are exaggerated, but there can be real problems. I’ve been battling paperless electronic voting for over 20 years now. Many people have contacted me over the years about “vote flipping”, and I’ve seen many articles in the media about it. I was even an expert witness in a possible lawsuit on vote flipping many years ago, where I tested machines myself (the suit did not go to court). The good news is that vote flipping is not even an issue for the vast majority of Americans (70%), who vote on hand marked paper ballots. This is because people like me have been working for over 20 years to get a more secure voting system. But 30% of Americans vote on machines (called “ballot marking devices”) that can flip votes (the rest vote on hand-marked paper ballots). But most of those produce a paper ballot that is counted using a different machine. You can make sure your vote was recorded correctly by checking this paper ballot. If it’s wrong, you should make a new ballot. If you think the machine made the error, you should tell a poll worker. They should stop using the machine if it might be broken. Some of the remaining voters vote on machines with a “paper trail”, which is a copy paper of your vote. You can inspect this and reject it if it’s wrong (and vote again). Why do votes flip? * People make mistakes. They touch the wrong things, and even forget how they voted. * Sometimes, voters rest part of their hand on the screen while they vote with their fingers. Some machines get confused about this. * Some machines need to be calibrated, which matches the touchscreen to the display. The calibration can be off for a bunch of reasons, so the computer thinks you touched a different place on the screen than you did. * The machine could have been set up wrong, in which case it will always flip votes the same way. This would certainly be caught by voters, since it would happen all the time. * Computers break in weird ways. *The machine might be hacked in any number of ways, although no one has ever found this in a real U.S. election. So far as I know, there has never been a documented case of a hacked voting machine in a real election in the U.S. Almost all problems are honest mistakes. But mistakes can change election results, so careful checking can make a big difference. And everyone wants to feel that their vote was counted correctly. P.S. Statistics on voting machine use, and lots more useful info, from VerifiedVoting.org

70

2 Comments

So Kafkaesque! This review by a former KGB illegal compares me to Kafka. Others recently cited John Le Carre and Somerset Maugham and a professor of literature compared me to Dostoevsky - but Amazon rejected his review (below). Like "The Trial," they sent a list of dozens of possible crimes against community standards with no specifics and said, pick one, you are guilty, whatever you pick. (Goodreads published it with nary a hitch.) the review: I operated in the "wilderness of mirrors" as a KGB illegal for ten years during the Cold War. When you do that, you forget who you are and seem to live in an alternate reality--like Mobius. Thieme had to have worked in intelligence to have such an understanding of the moral challenges we faced. Mobius is the Forest Gump of the intelligence community. This fascinating "memoir" is as real as it is surreal and keeps you guessing from start to finish. I watched Thieme's entire Defcon speech on the traumatic impacts of intelligence and security work - Playing Through the Pain: The Dark Side of Professional Intelligence and Security Work - on youtube). I cannot remember having ever heard anything so deep and intrinsically moral, inclusive of the human dilemma. This level of intellect is hard to find across all media these days. The hero of the memoir is a composite of several agents. There are many parallels to my past reality. The deeper I got into the book, the more I was reminded of Kafka. "The Castle" and "The Trial" are two of my favorite books. Alienation is all too real. I also get a touch of nihilism. (Ecclesiastes 1-2). Of course, it is always possible that I am looking in the mirror. We now lie much more often than we speak the truth. And when you do speak the truth people probably think: "What is he hiding?". It gets really confusing when we mold ourselves according to the (perceived) perception of others. Before you know it, we have self-inflicted multiple personality disorder. I can prove that I had at least two permanent personalities who became one when I came out of the closet. My wife never knew which personality would show up. She said I was a chameleon. Thieme's description of PTSD is haunting. I think I understand everything as I read, but then pause and wonder. Do I really? If this were a real memoir, no problem. If I just got it as raw fiction, no problem either. But now that I know the author I can't help guessing how much of himself he put in there. In the end I come away with a ton of mixed thoughts and feelings. It is almost like taking Kafka and rewriting it to make it into a Kafkaesque transcription - Kafka squared!

6

2 Comments