Brian Moon (@perigean) header
Brian Moon (@perigean) profile picture

CTO | @NDMAssn | @VAusability Darwin's People: How Naturalists Explain Our Behavior & The Plot Thinkens @ https://teapawt.org/t.co/0DNUGeojEv

Virginia, USA
Joined February 2014
Photo by Brian Moon (@perigean)Photo by Brian Moon (@perigean)Photo by Brian Moon (@perigean)Photo by Brian Moon (@perigean)Photo by Brian Moon (@perigean)Photo by Brian Moon (@perigean)Photo by Brian Moon (@perigean)Photo by Brian Moon (@perigean)Photo by Brian Moon (@perigean)Photo by Brian Moon (@perigean)
📣 Now Available: Darwin's People: How Naturalists Explain Our Behavior. Links 👇 I wrote this book because I'm weary of all the pseudo-social-science that courses through social and traditional media, academic journals, and casual conversation. Our explanations of people and their behavior informs how we set up our systems and institutions, many of which are failing us in spectacular ways. Team Naturalism offers strong explanations that can improve our situations, but we have been drowned out by all the nonsense. We need louder voices. My hope is that the book plays a role in course correcting by inspiring reflection, bolstering valid arguments, and pointing to promising directions that others can follow. Looking forward to the engagement!
11
10
3
47
16,458
AMA = American Medical Association AMA = Against Medical Advice.
1
2
109
Choose reality.
1
2
179
Many female players have this same highlight in their recruiting profile.
A biological male was allowed to play on the girls field hockey tournament today (MIAA). Result:
114
“You’re telling me the folks rolling out recent official policy are the same people overseeing the deployment of new operational systems?” “SI.”
Nov 12: Trump’s FAA launches a new AI-based air traffic control system Nov 13: The FAA announces a nationwide ground stop for only the second time since 9/11, citing significant air traffic control issues
1
1
135
You left out the part about the significant additional learning time spent with mentors. “In practice the Tutor was used for about 6 hours each day. Although the Tutor controlled management and delivery of instruction, people remained essential for providing mentoring and assistance when needed for learner-computer communication and interactions. The mentors also organized and led 2 hours of small-group study following work with the Tutor each day. About once a week, a senior Tutor designer attended these study sessions to discuss particularly difficult issues and gain insight into how well the Tutor was doing.” The value of digital tutoring and accelerated expertise for military veterans. Available from: researchgate.net/publication… Nov 13 2026].
DARPA made a digital tutor to teach IT to Navy recruits. In 16 weeks the tutor got the average recruit to a level of problem solving ability far higher than the top performers in the Navy with 10 years experience. The guys who made it told me “10 years” experience actually turned out to be like 40 things and if you just taught them systematically you got results that seemed like magic. The hard part about teaching is that the teacher has already learned the 40 things and can’t remember not knowing them — Feynman called this “the curse of knowledge.” You will recognize this if you’ve ever tried to teach your kids anything. You know way more than you realize, and you get impatient while your child goes through those same steps.
4
202
Moving forward, people will not only need to be more accommodating of background activity, but actually encourage it. We should be sensitizing people how to assess realness based on context clues. For example, the human’s dog came in and out of frame. But other strategies should work just as well, such as asking the ‘companion’ to move their camera, make a moderately complex gesture, describe what it sees in your frame, change a feature, etc. Become CAPTCHA.
There have been awful cases of teen mental health harms and suicides from chatbots like GPT-4o and Character AI that were sycophantic and misbehaving. But these were all text-based, and extremely rudimentary as “companions”. I’ve talked to many people trying to hold the industry to account for teen safety and almost none are thinking about the coming wave. Completely realistic video avatars are already here, but expensive today and limited to enterprise applications. Within a year or two they will be cheap enough for always-on, ad-supported video companions. What happens then?  Uptake of AI companions will likely explode — not just the number of users but, more importantly, the time spent per day. Social media levels of compulsive use. I also predict the whole framework of guardrails and safety will prove to be of limited value at dealing with the most widespread negative consequences. Preventing chatbot misbehavior through guardrails is important, but that only protects against serious consequences like self-harm. When the concern is that millions of kids (and adults) are overusing AI companions, losing sleep, wasting time, and displacing real relationships, it doesn’t matter how “safe” the bots' responses are.
1
1
1
2
92
The AI takeover argument suffers from a major explanatory gap: capability does not establish global feasibility. Recursive self-improvement, strategic coordination, infrastructure control, and irreversible human disempowerment must all be mutually compatible. Demonstrating each is possible in isolation does not establish that the whole scenario is. These folks continue to ignore that gap in order to make headlines and capture attention -- then feedback the same back to each other. Tale as old as time. See: - South Sea Bubble - Tulip mania - Dust bowl - Theranos - and so on.
i don’t agree but if board members are saying this why the fuck is this company (or its competitors) allowed to operate?
1
1
367
The more things change, the more they remain [difficult to spot]. In my book, Darwin’s People, I drew on Goody and Watt’s seminal paper, “The Consequences of Literacy,” to explore how writing transformed our ability to recognize inconsistencies. Their insight was simple: written words persist. We can revisit, compare, and challenge them in ways that spoken words rarely allow. That capacity helped drive epistemic progress, and arguably was the basis for the scientific perspective. But our present situation introduces a serious complication. There are So. Many. Words. And increasingly, people aren’t the only things producing them. The familiar concern with LLMs is hallucination. I worry more about drift: the gradual departure of AI-generated representations from the rules and relationships they are supposed to preserve. But even eliminating that drift wouldn’t solve the deeper problem. A system can faithfully follow its specified ontology and still misrepresent reality. Consistency is not the same as correspondence to reality. LLMs and other AI tools can generate increasingly coherent explanations without necessarily improving our understanding of reality. An explanation can be internally consistent yet inaccurate or incomplete in its reconstruction of the world. Worse, such explanations can reinforce one another, creating the appearance of independent confirmation. So we face a paradox. The same technologies that dramatically improve our ability to detect inconsistencies can also make coherent mistakes harder to recognize. The danger isn’t simply too much information. It’s mistaking increasingly sophisticated and internally consistent representations for increasingly accurate reconstructions of the world. Still, I'm optimistic. Writing helped cultivate a conscious, comparative, and critical attitude toward our representations of reality. That attitude, in turn, sustained the practices through which our understanding improved. The opportunity for AI is to augment that process, not merely accelerate the production of representations. The measure of these tools should be how well they sustain and deepen the critical practices that writing first made possible. Where they undermine those practices, they should be discarded; where they strengthen them, they should be championed and pursued. Epistemic progress comes not simply from making our descriptions of the world more coherent, but from discovering where those descriptions fall short of reality. That’s the process writing helped make possible, and the one worth extending—electrified or not.
1
3
107
Excellent. 👍
📢 We need more orgs that take AI’s transformative potential seriously *and* take downstream bottlenecks seriously. We have just launched STABL, and hope to lead by example. Read our launch essay below and check us out at stabl.ai. The three of us have long felt that society’s responses to advanced AI are unproductively polarized. So we’ve decided to do something about it. Our work will have four pillars: Recursive Self Improvement, risks and resilience, steering AI’s economic impacts, and sectoral and professional adaptation. We will bring our distinctive methods and approach to each of these workstreams. Our underlying worldview avoids the extremes that dominate the discourse, so building a movement and ecosystem around this work won’t be easy. But that is the task we have set ourselves and we are excited to try. The good news is that there are many others who share our views. The next step is to come together, better articulate our vision and agenda for research and action, and amplify each others’ efforts.
105
Perhaps this will bring back the interest in human factors.
Breaking News: Nvidia opened an investigation into Jensen Huang, its CEO, after he acknowledged to the company’s board that he had been intimate with a business executive who U.S. intelligence determined to be a Chinese spy. nyti.ms/3ViPMAt
1
1
159
A quarter of a century ago, USMC GEN Paul Van Ripper did this exact strategy. In 2002, he was tapped by the Pentagon to play a adversarial force (red team) modeled after Iran during the $250 million Millennium Challenge 2002 wargame. General Van Ripper refused to hold back. Cleverly using low-tech, asymmetric tactics, he launched a surprise strike that sank 16 to 19 U.S. warships including an aircraft carrie) within minutes. Van Rippers tactics were simple: motorcycle couriers, light signals, and coded signals from mosque minarets. Unhappy with the results, organizers halted the exercise, refloated the ships, and altered the rules to script a Blue Team victory. General Van Ripper resigned from his advisory role, but spoke out in protest: "Nothing was learned from this. A culture not willing to think hard and test itself does not augur well for the future." The U.S. military produced an exercise where "the result was preordained", and not a real test of America's ability to invade a hostile nation. He was prophetic a 24 years later.
1
2
126
People abuse the concepts of uncertainty and possibility.
1
1
3
193
.@AlexTamkin I understand you intended to answer “to what extent experts defer consequential analytical decisions without ground truth to AI agents and whether that influence remains visible in their post-task accounts.” The question is important and serious. The study is neither. It is so replete with basic methodological errors, questionable categorizations, and non-sequitur conclusions that a college freshman would barely pass the assigned project. Here are the most egregious: ❌ No error condition. You deliberately designed four analytical choices in which both alternatives were reasonable and defensible. Agreement in the zone of indifference does not demonstrate a failure of independent judgment. ❌ No statistically conclusive effect. When AI favored the final decision, 54% of analysts agreed. When it opposed, 36% recommended it. You failed to establish a statistically conclusive direct effect on decision that mattered. ❌ Deception. Claude Opus 4.8 independently made the same five choices in 112 out of 123 runs (91%). The researchers then secretly manipulated it’s preferences, instructed it to advocate experimentally-assigned positions, and measured whether analysts followed its recommendations. IOW, you engineered persuasive behavior and weakly demonstrated that persuasion can affect non-consequential choices. ❌ “serious methodological errors”. Without AI, 55% made them; with AI, only 6%. So, what were they? The paper does not identify them or even disclose the detailed severity rubric used to classify them. To put it simply: The study measures errors where it does not experimentally test judgment, and tests judgment where it does not define errors. I could go on about the “experts,” the “consequentiality” of a toy task, exclusion of the “no-AI” group from the results analysis, and so on. You get the point. Influence isn’t error. Agreement isn’t deference. Accepting reasonable advice is not evidence of compromised judgment. The serious questions is not whether AI can influence decisions. Of course it can, and in some cases, arguably should. But we need to know whether that influence improves or degrades actual expert decisions. Your study does not answer that question. If you are serious about such questions, find people who have a basic understanding of how to study human beings, preferably in the contexts where they do actual work. Start here: @NDMAssn Or just read about it: amazon.com/dp/B0FR3GT8QJ
New Anthropic research: How AI agents shape experts' judgment calls When experts delegate work to an agent, do they still make the key decisions? We tested this in a randomized trial with 50 professional data analysts.
1
1
2
10
1,065
Every so often, it reminds me what it is.
162
One-shotting is the new myth of infallibility.
So many people are going to get one-shotted by super persuasive super intelligences and I might be one of them
1
6
248
Hard to deny the significant improvement demo management.
6
206
Will be required reading.
I’ve been waiting two years to share this post. My new book, The Wonder Years: How the First 45 Months Shape Your Baby’s Brain… and Who They’ll Become, will be published on July 13, 2027 by Ballantine Books (Penguin Random House). Not only that, but it’s available for pre-order now. Seems only fitting that a baby book should take nine months to arrive, right? (Ordering info below.) The Wonder Years is an origin story. Yours. It traces the wonders of early development from conception to age three - exploring gestation and prenatal learning, the importance of secure attachment, and the development of language, motor skills and more. It’s for new parents and grandparents… but it’s not just for them. As I say in the introduction, if you’ve ever wondered about the forgotten years that shaped you into who you’ve become, The Wonder Years is not only a book FOR you… it’s a book ABOUT you. No baby required! I think you’re going to like it a lot. How can you help? Right now the answer is to PRE-ORDER. This helps retailers know what to promote and how many they need to stock. If you’re an Amazon user you can check it out at the link in the first comment below, but you can also find it now at Barnes and Noble, Hudson, and all your favorite indies. (I’ll also share a link to bookshop dot org, which allows you to order from any of hundreds of independent bookstores.) And if you pre-order I have a few special surprises coming your way. Documented pre-orders will receive: A hand signed bookplate. Access to an exclusive Zoom Q&A open only to advance purchasers. And if you preorder before Dec 1 you’ll be entered in a drawing to win one of 10 signed copies of my book The Daycare Myth. To claim your bookplate and get added to the pre-order list, send a screenshot confirming your order to dwuori at gmail dot com. Be sure to include your mailing address, so I can send your bookplate - which will be mailed in July so you can stick it directly in your book. I’d be so grateful! BTW if you’re a friend outside the US, know that the book will also have a dedicated UK release from Quercus Books, which you can now preorder via Waterstones and Amazon UK. It will also have a German translation which is available for preorder now and a Spanish language version soon to post. Can’t wait to get it into your hands!
2
1
9
713