Paperweights, 1st Edition

Chatbot Spies, AI Ethnography, and Bono

“RAND researcher Nancy Nimitz, September 1, 1958.”

Here’s what I have been reading.

1. “Computational Espionage” | Jordan Schneider
   
   Imagine Her meets Tinker Tailor Soldier Spy. That’s the future we’re in for, and we need to adapt. One of the worst recent US intel leaks seems to have come about because a National Guard enlistee with a clearance wanted to impress his friends on Discord. Now we carry the “honeypot” in our pockets. If you thought a CCP-adjacent TikTok was bad news…
   

2. “Looking for Hidden Gems in Scientific Literature” | Ulkar Aghayeva

   Everything Ulkar writes is great, including this piece on how we can better explore and leverage scientific findings that get lost in the literature. This research constitutes a kind of “intellectual dark matter,” what Donald Swanson calls “undiscovered public knowledge.” Can LLMs and other AI tools help us sort through the noise to find the signal? Hopefully…

   An excerpt:

   The first instances of literature-based discovery were accomplished by cumbersome manual screening of papers… In response to this challenge, an array of methods has been developed in the fields of natural language processing, machine learning, and artificial intelligence… A daydreaming loop in an LLM can potentially perform billions of such pairings in a much shorter time (hours?). The DDL algorithm could therefore realize the original promise of literature-based discovery – surfacing high‑value “unknown knowns,” at scale.

   This is a major use case for AI in science. Automated and cloud-based labs are promising, but literature-based discovery seems especially ripe for good R&D disruption.

3. “We Need Ethnographic AI Safety Studies” | Joanna Wiaterek
   
   I caught the anthropology bug a few years ago after reading Gillian Tett’s Anthro-Vision and beginning my work as a journalist. Technical approaches to AI safety and security are essential, but they may be easier to answer than the less tractable human questions. The “squishy” questions are hard to answer, but the social sciences have much to contribute here if their practitioners are willing to get out of the Ivory Tower and embrace the “worm’s-eye view.”
   
   An excerpt:

   We want to expand on the idea that “mathematical and computational approaches do not adequately capture the societal impacts of AI systems” (Schwartz et al., 2022) because they neglect the role of the human(s), and the diverse number of actors who design, interact and are impacted by AI technologies. We advocate for using ethnographic methods as a complementary critical tool for AI scholars to increase the accuracy of their safety evaluations, risk assessments and policy recommendations — a method that is rooted in the socio-technical study of AI systems, inquiring into both the social and technical aspects of their inputs and outputs.

   Every frontier lab should have anthropologists on staff! I do not mean the Latourian variety, however. We need researchers skilled in old-fashioned ethnography with a deep grasp of cultural evolution and managerial economics (ideally). It shouldn’t be a corporate vanity project either. Ethnography can help us develop safer AI systems if we leverage it properly.

4. “Sam Altman on Trust, Persuasion, and the Future of Intelligence” | Tyler Cowen & Sam Altman

   
   The word “year” appears 24 times in this interview — timeline enthusiasts, take note. Also, this segment stands out:

   ALTMAN: I eat junk food. I don’t exercise enough. It’s like a pretty bad situation. I’m feeling bullied into taking this more seriously again.
   COWEN: Yes, but why eat junk food? It doesn’t taste good.

   ALTMAN: It does taste good.

   COWEN: Compared to good sushi? You could afford good sushi. Someone will bring it to you. A robot will bring it.

   ALTMAN: Sometimes late at night, you just really want that chocolate chip cookie at 11:30 at night, or at least I do.

   COWEN: Yes. Do you think there’s any kind of alien life on the moons of Saturn? Because I do. That’s one of my nutty views.

   I’m with Sam on this one. Now to get an AI bio-engineered perfect, low- or no-calorie chocolate chip cookie — what I consider to be the El Dorado of AI benefit sharing.

5. Surrender: 40 Songs, One Story | Bono

   This was my fun reading, but it turns out that Mr. Bono’s wild ride has many lessons for those anxious about navigating the post-AGI world. A few takeaways: 1) get religion, 2) stay flexible, and 3) double down on what makes us human. This book is perhaps best received in audio form, narrated by the author and mixed with auditory flourishes of music and entertaining (if not always 100 percent accurate) impersonations of Bill Clinton, George W. Bush, and Nelson Mandela.

   
   We’ve become so inundated by human- and machine-generated text. Maybe the future belongs — or never stopped belonging — to the orator, storyteller, and shaman? Also, the account of how Bono helped instigate PEPFAR is a good reminder that sometimes individuals can reshape economic history in deep, lasting ways. It’s also insanely funny to hear about the Irishman’s sleep-inducing fruit and wine allergy followed by a wake-up call from President Obama after crashing in the Lincoln Bedroom.