Really? The kind of biases I don’t like seem to be system prompting preferences. For example, when an idea is marginal, ChatGPT will kick the can forward to keep the conversation going. Perhaps the instinct is to create a demand signal or “billable hours”. Gemini Pro will tend to be cautious. To use them effectively one needs to have some self-discipline. I don’t see these tools doing obvious motivated reasoning like people do. ChatGPT will especially follow your lead, which can be a bad lead; it is a people pleaser.
My latest “wow” moment: I was checking up on the Kerch bridge attack and took a few frames from the video and pasted them for ChatGPT. I asked if the explosion suggested poor placement by Ukrainian frogmen. It explained that blast off to the side would be typical and suggested an attempt to displace the pier laterally deep under the water. It did not imply an attempt to knock the bridge down but rather to make it unsafe to use, while also leading to incremental structural failure. From: Friam <[email protected]> On Behalf Of steve smith Sent: Wednesday, June 4, 2025 8:27 AM To: [email protected] Subject: [FRIAM] Alignment (with what?) DaveW, et alia - The Alignment Problem, by Brian Christian I would say that Christian's piece here acutely represents what I'm trying to re-conceive, at least for myself. His implications of Human Exceptionalism and a very technocentric focus which largely avoids deeper political critiques about who gets to define "alignment" and whose values are prioritized. It is a bias oft-presented by those of us who are tech-focused/capable/advantaged to reduce a problem to one we think we know how to solve (in a manner that promotes our narrow personal interests). In the spirit of "anti-hubris", I was once strongly aligned with Robert Heinlein's (RAH) "Human Chauvanist" or "Human Exceptionalism" perspective as exhibited in his Lazarus Long (LL) character's oft-quoted line: "A human being should be able to change a diaper, plan an invasion, butcher a hog, conn a ship, design a building, write a sonnet, balance accounts, build a wall, set a bone, comfort the dying, take orders, give orders, cooperate, act alone, solve equations, analyze a new problem, pitch manure, program a computer, cook a tasty meal, fight efficiently, die gallantly. Specialization is for insects." I can't say I don't still endorse the optimistic aspirations inspired by LL's statement, it is the "should" that I am disturbed by. I am a fan of generalism but in our modern society, acknowledge that many if not most of us are in fact relatively specialized by circumstance and even by plan and while we might *aspire* to develop many of the skills LL prescribes for us, it should not be a source of shame or of "lesser" that we might not be as broadly capable as implied. We are a social species and while I cringe at becoming (more) eusocial than we already are, I also cringe at the conceit of being order 10B selfish (greedy?) individual agents with long levers, prying one another out of our various happy places willy nilly. I also think the hubris aspect is central. One of the major consequences of my own "origin story" foreshadowed by my over-indulgence in techno-optimistic SciFi of the "good old fashioned future" style and particular RAH's work was that he reinforced my Dunning-Kruger tendencies, both by over-estimating my own abilities at specific tasks and narrowed my values to focus on those things which I was already good at or had a natural advantage with. As a developing young person I had a larger-than average physicality and a greater-than-average linguistic facility, so it was easy for me to think that the myriad things that were intrinsically easier for me based on those biases were somehow more "important" than those for which those things might be a handicap? I still have these biases but try to calibrate for them when I can. My first "furrin" car (73 Honda Civic) was a nightmare for me to work on because my hands were too big to fit down between the gaps amongst all the hoses and belts and wires that (even that early) smog-resistant epi-systems layered onto a 45mpg tiny vehicle such as that. And you are all familiar with my circumloquacious style exemplified by "I know you believe you understand what you think I said, but I don't think you realize that what you heard was not what I meant". While I might have been able to break a siezed or rusty bolt loose on my (first car) 64-Tbird or (first truck) 68 F100 without undue mechanical leverage it was hell to even replace spark plugs or re-attach an errant vacuum line on my Honda. And while I might be able to meet most of my HS teachers on a level playing field with complex sentence constructions (or deconstructions) or logical convolutions, the same tendency made me a minor pariah among some of my peers. Back to "alignment" and AI, I would claim that human institutions and bureaucracy are a proto-instantiation of AI/ML, encoding into (semi)automated systems the collective will and values of a culture. Of course, they often encode (amplify) those of an elite few (monarchy, oligarchy, etc) which means that they really do present to the masses as an onerous and oppressive system. In a well functioning political (or religious) system the institutional mechanisms actually faithfully represent and execute the values and the intentions of those who "own" the system, so as-by-design, the better it works, the more oppressed and exploited the citizenry (subjects) are. We should be *very* afraid of AI/ML making this yet-more efficient at such oppression and exploitation *because* we made it in our own (royalty/oligarchic) image, not because it can amplify our best acts and instincts (also an outcome as perhaps assumed by Pieter and Marcus and most of us often-times). I don't trust (assume) the first-order emergent "alignment" of AI (as currently exemplified by LLMs presented through chatBot interfaces) to do anything but amplify the existing biases that human systems (including pop culture) exhibit. Even Democracy which we hold up quite high (not to mention Free Markets, Capitalism, and even hyperConsumerism,and hyperPopulism) is an abberant expression of whatever collective human good might be... it tends to represent the extrema (hyper fringe, or hyper-centroid) better than the full spectral distribution or any given interest really. An ill-concieved, human-exceptionalist (esp. first world, techno-enhanced, wealthy, "human-centricity") giant lever is likely to break things (like the third world, non-human species, the biosphere, the climate) without regard to the fact that to whatever extend we are an "apex intelligence" or "apex consciousness", we are entirely stacked on top of those other things we variously ignore/dismiss/revile as base/banal/unkempt. Elno's aspiration to help (make?) us climb out of the walls of the petri-dish that is Terra into that of Ares (Mars) to escape the consequences of our own inability to self-regulate is the perfect example of human-exceptionalist-hubris gone wrong. Perhaps the conceit is that we can literally divorce ourselves from the broad based support that a stacked geo/hydro/cryo/atmo/biospheric (eco)system provides us and live entirely on top of a techno-base (Asteroid mining Belter fantasies even moreso than Mars/Lunar/Venus/Belter Colonists?). ExoPlanetarian expansion is inevitable for humanity (barring total premature self-destruction) but focusing as much of our resources in that direction (ala Musk, especially fueled by MAGA alignment in a MAGA-entrained fascist industrial-state?) as we might be on the path to is it's own folly. The DOGE-style MAGA-aligned doing so by using humble humans (and all of nature?) as reaction-mass/ejecta is a moral tragedy and fundamentally self-negating. Bannon and Miller and Musk and Navarro and Noem and ... and the entire Trump clan (including Melania and Barron?) are probably quite proud of that consequence, it is not "unintended at all" but I suspect the average Red-Hat-too-tight folks might not be so proud of the human suffering such will cause. Maybe those chickens (the ones not destroyed in industrial egg-production-gone-wrong) are coming home to roost? Veterans services, health-care-for-the-many, rural infrastructure development, humble family businesses, etc might be on the verge of failure/destruction in the name of concentrating wealth in Golf Resorts, Royal Families, and Space Adventurers pockets? Or maybe we are generally resilient to carry all of that on our backs (with AI to help us orchestrate/choregraph more finely)? Many hands/heads/bodies make light work even if it is not righteous (see pyramids?) Bah Humbug! - Steve
smime.p7s
Description: S/MIME cryptographic signature
.- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
