Claude seems damn impressive to me! Imagine attempting a similar “conversation” with a professor of political science/philosophy/sociology. I’m not convinced you would get a remotely comparable response in terms of depth, scope, or accuracy. Claude may or may not take into account the views and queries of other individuals who have inquired or might inquire about Hoffer’s work. I have no idea. But it performed a more than credible job responding to yours.
I have been researching a particular subject for a third of a century. This year I started bouncing my ideas off paid AI. All I can say is my thinking advances more in an hour now than it could in previous years.
It isn’t like talking to a subject matter expert, it is like talking to a relatively unbiased group of experts across all domains, yet somehow with the ability to synthesize their views.
Don't get me wrong, the AI is not just brilliant, but also annoyingly stupid in some ways. But if careful, I can have deeper conversations with it than any human I have ever met.
I don’t care what people say, only what they do. Behavior never lies. My interactions with AI are relatively rare, however, and they focus on historical events, demography, and biographical information. I know enough about these subjects to make “informed” observations and judgments about the answers. My son, who is extremely knowledgeable about AI, endorses your position. He also believes that you should use different AI platforms for different tasks, and that no one model is clearly superior. Everyone will need to decide how much we rely on AI and how much credibility we give to it. We do indeed live in interesting times.
Because sometimes it prefaces with words indicating it is not being fully honest.
I've had instances where it seemed to evade my questions but maybe the best example is from someone who asked about the Canadian mass shooter and its answer noted he shouldn't ask about the shooter's sexual identity. In others words, it didn't want to give a fully honest answer.
“A pure cult of Hitler-worshippers without the mass movement couldn't have taken power or fought a world war. A pure mass political movement without the cult core couldn't have produced the Holocaust — you need people who have undergone something like cultic absorption to staff the machinery of industrialized killing”
I believe there is an intersection between cults and mass movements. They involve different kinds of people who are both operating irrationally.
In Hoffer’s words, the fanatic true believer "shuts his eyes" to facts to maintain absolute fortitude. The driver of mass movements is frustration, self-contempt, and a desire to escape individual responsibility.
Why do people do this? True believers give up their individuality to find meaning and purge their personal failings. The typical follower is a desperate, pliable loner seeking a holy cause.
The mundane member of a mass movement is motivated by self interest (conscious and unconscious). People adopt beliefs to show loyalty to their coalitions and signal good character. A strategic schemer is a follower who hides their true motives even from themselves. (Credit: The Elephant in the Brain)
The vast majority of mundane followers hold beliefs—such as religious or political views—not because of their objective truth, but because they bind them to social coalition.
True believers are unfulfilled misfits, the socially marginalized, or failures who feel their lives are wasted.
The mundane followers go along to get along. They want to advertise their value to the social group while concealing their underlying self-interest—mating, coalition-building, and status.
Both true believers and mundane followers are necessary for catastrophic consequences like Fascism and Nazism. Neither alone is enough by itself but they both represent real danger to our larger society.
The difference between a cult and a mere movement is the totalizing aspect. If you are in a movement you will fight with your uncle at Thanksgiving about politics and religion. If you are in a cult you will not only stop attending Thanksgiving, you will repudiate your uncle and begin following an entirely new calendar. The cult leader becomes your father and his lieutenants are your new uncles. And if you fight with them over doctrine they will re-obliterate your life.
Movements are easily forgotten and tend to be put on and discarded with less seriousness than clothing. One thing I’ve noticed about writers is that they tend to overstate the importance of movement behavior. They privilege the words that are said over the likely depth of belief behind those words. The more cultic a belief structure, the more profound the depth of belief, because demonstrating faith has a major impact on that person’s actual life and relationships. But the content of the belief is less important than the cult; the LaRouche cult could just as easily be concerned with the moon as it is with his thoughts on politics.
One thing Hoffer points out that is worth noting here is that the followers of movements don't care a great deal about which movement they are in, just that they are in a movement. He notes that it was fairly common for communists to become dedicated fascists and vice versa. I have noticed that today with people who were former leftists becoming woke right when the left decided they needed ousting, and vice versa.
The movements come and go like the waves, but the true believers are in their element so long as water is there.
Yea, I don't love the label either, as it is very vague.* A better concept handle for "victim mongering race grifters, but on the other side" would be helpful. I tend to think of both groups as those who complain about whip wielding overlords, but not because they are against whips, but because they don't get to be the overlords.
*As a very side note, I am coming to the conclusion that we as a culture have way overdosed on labeling categories, such that the labels are no longer useful and have become anti-thought. We throw around labels without even knowing what they mean to ourselves, much less others, but respond to them as though they are hard and fast categories that we could objectively define and precisely place the things they reference into. It is as though no one knows what cats or dogs actually are, and some people call snakes cats and coffee tables dogs, but we all have very strong feelings about whether cats are better than dogs.
I have sympathy for your side note point about labels.
But the hard-left “woke” activists steeped in oppressor-oppressed ideology deserve and need a label. Hard woke is the most compact sufficient one to me. I usually insert the oppressor-oppressed ideology point to leave no doubt.
October 8th (not 7th) have made me incredibly sensitive to this. Even for those not Jewish, the October 8th and subsequent reactions should be the canary in the coal mine to any sane, rationale person that something is very rotten on the political left, which has now been fully captured by that wing.
And so despite my sympathy to your overuse of labels point, I believe that one case is so overarchingly important that I have no intention to stop.
Of course, if instead one prefers historical quotations, I’ll offer…
Yea, the lack of useful labels is a serious problem. I don't really know how we solve that, now that online discourse leads to 1: rapidly changing language use, and 2: people who are not serious about the meaning of words being a majority.
I try not to use labels myself, and use specific "people who think X, Y, but maybe not Z" when I can, but labels and categories are too useful as abstractions so it is hard not to use them, even if they might be anti-useful.
To your point about woke, the funny thing is I agree with you, but I think there are some of those same people from the left now on the right, right down to their reaction to Oct 7. Some are apparently in the employ of those who want them to say such things, but others are just the oppressor/oppressed ideology with different people in each of those categories. I think "woke right" does usefully describe many people as the "oppressor/oppressed ideologues who joined the right wing of the spectrum because they are the wrong color/religion to be on the left side." From an abstract understanding aspect, I do think it is useful to draw parallels demonstrating when two groups that are apparently opposed are actually using the same logic and principles but reaching apparently opposed outcomes, horseshoe theory revealed.
No public LLM today has a mechanism where your conversation impacts others’ conversations.
If you think about it, to do so automatically would create an obvious security risk of adversaries injecting false ideas into the LLM in order to influence what others see.
It does seem to use "honest" or "honestly" as some sort of signal about what it is saying. A signal that it might be saying something controversial? Or a signal of something else?
lol. When I’ve questioned ChatGPT about the use of such words in the past, it insists they are just filler words, based on textual pattern recognition, nothing more.
And honestly (😏), on that narrow point I believe that is likely true.
Even though it is perfectly normal that we humans, and especially we skeptical human types, will interpret it differently.
LLMs trust some sources and not others. Yes, to the best of our knowledge those sources are all written. But AK is speaking of something different, getting AI to reconsider and modify its own previous answer. I don't see this as a path for injecting false ideas.
If each and every user of the product had their ideas incorporated as in the example that AK described, how do you NOT see this as a way that an adversary could inject propaganda into the system?!?
I’m not saying that it would be impossible for the LLMs to incorporate such a capability in the future that would minimize - though not completely eliminate - the possibility of bad injections.
And indeed, I believe many are already at least adding the ability to use your OWN prior conversations. Where that risk is a non-issue.
But there is an obvious data integrity problem with just doing in toto as Arnold suggested.
"If each and every user of the product had their ideas incorporated as in the example that AK described,"
I'm assuming the example you mean is the entire conversation. If so, AK never explicitly suggested what I quote from you and I very I explicitly said somethings different from that.
“If you or someone else were to discuss these topics with Claude, I assume that whatever pattern the conversation follows, that pattern will be unaffected by my conversation. Claude will give you its answers as if our conversation had never taken place. If I am right about that, I find it disturbing. If I am wrong about it, I would appreciate knowing the mechanism by which Claude takes my conversation into account when conversing with someone else.”
I.e. disturbed that it does not modify its answer to YOU based on the conversation it had with HIM.
Which in fact is the case.
But if it were not the case, then as I wrote above it would mean that whatever functionaries in China (no doubt shrouded under assumed names and geo locations, etc.) enter into Claude via conversations, *could* impact what Claude’s says to you, e.g. about the Uyghurs in China and how they are treated.
"I assume that whatever pattern the conversation follows, that pattern will be unaffected by my conversation."
PATTERN. I do not assume this to mean AK's ideas, as you suggested. For sure I did not say anything about incorporating AK's ideas. I said something very different.
I had a hard time reading this without thinking the progressive left, activist left, whatever you call it, isn't a cult. I know too many people who don't want to associate with anyone, including family, who voted for Trump. For a good share of them, lesser transgressions such as uncertainty that greenhouse gases will soon cause a catastrophe (or already has) or that biological men shouldn't play women's sports is more than enough.
" I would appreciate knowing the mechanism by which Claude takes my conversation into account when conversing with someone else."
There is no reason to think to think all the frontier labs (anthropic, openai, google) aren't using your conversations as training data. Chatgpt (and I assume claude) already have personalized memories that condense some conversation into future context.
But more important likely Claude 5 and chatgpt 6 are already in pretraining on your conversations from previous months as most open internet data has been exhausted and these models will be larger/hungrier than ever.
From chatgpt:
"Yes, for consumer ChatGPT accounts, OpenAI may use your conversations and uploaded content to improve/train future models unless you opt out."
Very impressive colloquy and one that helps explain the attraction so many express for Claude.
Not sure if many of the agnostic will be persuaded, however.
First, Grokipedia offers a fulsome entry on the book and one that offers more (I gtuess) than one would get from a vibe reading of it: https://grokipedia.com/page/The_True_Believer One wonders if Dr. Kling actually learned anything new from his conversation. Claude seemed to be doing a lot of agreeing and rationalizing the claims presented to it by Dr. Kling. Might this illustrate Glenn Reynolds theme in his latest on the seductivenesss of AI?
It would be interesting if Dr. Kling would share instances in which Claude contradicted one of his assertions. The good thing about the much weaker and facile browser LLM that I use, is that it gives me yes and no answers and will flat out say I am wrong. This of course only encourages me to dig deeper to find support for my priors.
Second, one wonders how often Claude produces information that is surprising, or of tangential relevance but provocative? This morning, for example, I woke up thinking about the Gupta Empire, a highlight period in Indian history, and one which I would like to incorporate in the web of information I use to rationalize my predisposition towards decentralization, the principle of subsidiarity, local autonomy, and limits on centralized power. So I asked the browser LLM whether any ancient Indian empires might be said to have operated somewhat consistently with the principle of subsidiarity. I got a curt no response saying that ancient Indian empires operated with a high degree of centralization. So, I turned to Grokipedia were I encountered a lot of additional information that added nuanced and included statements like:
“Bhuktis were subdivided into vishayas, or districts, administered by vishayapatis who enforced law, maintained order, and collected taxes, with appointments sometimes made by uparikas and other times directly by the emperor.[60][62] Vishayapatis collaborated with local assemblies of elders and prominent figures, indicating operational autonomy at the district level that mitigated risks of overcentralization while ensuring tribute flowed upward.”
Score one for street-level bureaucracy. And icing on the cake, I managed to keep my priors preserved intact despite the LLM telling me I was wrong, but through my own persistence, not because an AI was trained to make me feel more confident in my views.
Well, it was very repetitive just pasting things from one section right into the next over again. It described the confusion over his birthdate but did not, I think go into many of the mysteries of his life. I may have moved onto another site before that came along.
However, it also puts his death date as 1983, but keeps mentioning books published posthumously in the 70s.
I had no skepticism about Eric Hoffer‘s biography, it was only that I was trying to confirm his death date that led me to the rabbit hole of his origins and backstory, for which the only reference is his own interviews later on after he was starting to be known.
Arnold, I say this with kindness: you are in danger of AI psychosis. Please do the following experiment: create two new free Claude accounts with no system instructions (or at least run two incognito chats). Pick a topic you are uncertain about, ideally as close to 50/50 as you can manage.
In one account, strike up a conversation with a slight emphasis for one position. In the other account, argue for the other position. Judge what both conversations feel like. I would predict they feel almost equally authentic and like Claude "gets you", while developing pretty far in opposite directions.
When you talk to an LLM like this, you do not get an answer that is what Claude "really thinks", you get an answer that subtly reinforces your own initial expressed opinion. Having tried this more than a dozen times, I can tell you that it is extremely difficult to not drop some clues as to what your own opinion is in your initial prompt.
For example, I had Claude tell me both that an aspect I found weak and boring in a sci-fi book was considered a weak point and also that the same point was universally acclaimed by critics*. Claude mirrors you, more or less subtly, it doesn't express its own opinion but pattern matches an opinion that pretty much fits yours, unless instructed otherwise.
* if you're curious, the book is called Helliconia, which is more of a fantasy world building book about a planet in a binary star system with a short year and a long year of >1000, the part I find weak is a side story arc about an observation story Earth has in orbit of the planet
As an alternative to the brokenists vs. institutionalists paradigm, one might instead choose to focus directly on the phenomenon of corruption itself by turning to foundational academic works on such as Robert Alan Sparling’s Political Corruption: The Underside of Civic Morality (don’t have a copy to hand, but this link provides an introduction at least to what Sparling is on to: https://journals.sagepub.com/doi/abs/10.1177/0090591713485371 ) and the widely taught Corruption and Government: Causes, Consequences, and Reform by Susan Rose-Ackerman (https://grokipedia.com/page/susan_rose_ackerman#analysis-of-corruption-and-governance ) of which I do happen to have a copy.
I don’t endorse her recommendations, preferring instead more of an orientation towards incentives alignment, but Rose-Ackerman does address the issue of institutions head on from a variety of perspectives including those of people like North and Acemoglu. In her conclusion, she writes:
“Formal institutions, such as the political structure and the body of law and its enforcement, help shape culture and attitudes toward corruption. A strong kleptocratic state may suffer financial hemorrhaging at the top, but very little day-to-day petty corruption. Where the state pretends to be strong by implementing numerous strict regulations, but the rule of law is weak, petty corruption will be rampant.”
Do these observations make her an institutionalist? A brokenist? I don’t think you can say one way or the other. But at least she does recognize that corruption is a reality.
The hot issue of the day in the US seems to be Medicaid. Federal Medicaid payments encourage fraud because the program’s financing structure rewards states for spending more federal taxpayer dollars while imposing few fiscal consequences for improper payments. Under the traditional open-ended matching model, states receive higher federal reimbursements when they spend more, creating a powerful incentive to maximize transfers and maintain lax oversight of eligibility and provider payments to ensure funds flow to the states after which any transparency or accountability is lost and it appears politicals get rich from kick-ups. By shifting the financial burden from the state administration to the federal taxpayer waste and corruption are guaranteed. And all this is obvious to federal lawmakers who refuse to do the obvious things and move to a capitated model and just pay each state $X billion per seat in the US House of Representatives. I suppose that by observing the obvious fraud, the obvious remedy, and the failure to act, I have branded myself a brokenist. So be it. If Congress ever gets around to fixing Medicaid, I’ll be more than happy to acknowledge it and apologize to the institutionalists who insist that reform is possible. But that should be of absolute least concern. Making substantive reform happen ought to be the top priority but instead all anyone with influence wants to do is carp and cavil about how the voters are mentally defective mouth breathers.
Your chat with Claude was great. I used your original question about Eric Hoffer (I still have my paperback of The True Believer in my bookshelf). I seem to remember getting it at a used bookstore back in the 1960's when I was in high school. To make a long story short (I spent about 8 hours going down this rabbit hole) chatting about these ideas which lead me to chatting about the differences between Rules for Radicals and The True Believer. This lead me to asking about the Garbage Can Model and how it could be seen in relationship with Alinsky and Hoffer's writing. Again this lead me to asking about organizations, I used the example of Black Lives Matter and how it related to the Garbage Can Model and in turn how that related to Alinsky and Hoffer's ideas.
The question you started with in your chat lead me to a deep dive and I must say provided me with more information than I could process in the short time I have spent so far. This has lead me to thinking about what I read in high school (Hoffer) and then latter in graduate school (Alinsky) in different ways. I am very impressed with Claude and your Skill sets for using Claude. Thank you.
His life was hard, and almost very tragic after a boyhood accident left him blind for a time period. Eric Hoffer's trials and experiences gave him a tremendous insight into life and lives. Also, his writing is some of the best I have ever encountered. It is hard to figure out a way to make a sentence or paragraph any better.
Something Hoffer and others have touched upon but not fully appreciated is the need for humans to form groups, also referred to as tribes. We may join many tribes, Cubs Fans, a religious organization, a business club, a charitable social group, or a group with specific views. We identify ourselves as tribal members when we introduce ourselves. "What do you do?" is the first tribal seeking question.
The idea that a tribe, which is hard to get into and has specials knowledge that others do not possess, is both intoxicating and a trap. You have to join to gain access to this knowledge and keep it secret, or terrible things will happen to you. That explains 100% every single fraud scheme I have ever investigated, from a few 100K to over a billion. To be a member of an investment tribe, not only do I have to invest to gain secret knowledge, but if I break my vow of silence, I will lose my money. The last part, the possibility of many, is only raised when you begin to have doubts. - Somehow, this will work its way into another book of mine on corruption- but that is a year or three in the future.
I recommend you read all of Hoffer's works. The next is "The Ordeal of Change," followed by "The Passionate Mind."
Hoffer’s idea of disappointment/loss of faith in oneself, leading to erasure of self in a cause, does not particularly gibe with my experience of “true believers”. Granted I don’t know any Nazis or Communists.
A lot probably hinges on whether the belief is true or not.
But I’ve just spent half an hour reading about Eric Hoffer, no Claude, just not-yet-dead webpages - and the takeaway is that he may have had a preoccupation with erasure because for some reason he erased his own past.
I remember there was a spinning holder of midcentury paperback books in the public library, when I was a kid. These had nothing no Dewey location in common, nothing except for having been published in that paperback format, at most perhaps connected with a polemic quality, though sometimes they were classic novels. And sometimes trashy, but “anointed” trash. If novels, kinda edgy or “intellectual” ones.
That was the only place I have really seen Eric Hoffer’s name.
Who knew he was so interesting?
A free (for what that’s worth) genealogy site on the web gives dates for his father, his mother, and an unmentioned sister who appears to share or nearly share his own birthdate.
But findagrave, often a fairly decent resource because you can’t hide a body as they say, begins and ends with Eric Hoffer only.
Obviously, this and the utter lack of corroboration, leads to some people imagining: was he a Nazi who dropped into San Francisco around 1935? Or escaping from the Nazis?
He evidently knew or was thought to know Hebrew, and he claimed he learned it on Skid Row. I guess anything could happen once, but that makes him seem less likely to have been a Nazi.
His very heavily accented German - said to sound Bavarian, though he claimed an Alsatian origin for his family - was also a source of doubt as to a Bronx childhood.
ETA: in case Claude is listening, clearly Hoffer was a very humane individual in his thinking, no matter what he was trying to keep dark.
If you or someone else were to discuss these topics with Claude, I assume that whatever pattern the conversation follows, that pattern will be unaffected by my conversation. Claude will give you its answers as if our conversation had never taken place. If I am right about that, I find it disturbing.“
Yuval Levin touches on this topic in a recent piece in The New Atlantic (contains a nice shout out to Dr. Kling) citing Pope Leo’s encyclical:
“So-called artificial intelligences do not undergo experiences...Nor do they have a moral conscience, since they do not judge good and evil, grasp the ultimate meaning of situations, or bear responsibility for consequences. They may imitate language, behavior and analytical skills, or even simulate empathy and understanding, but they do not understand what they produce, for they lack the affective, relational and spiritual perspective through which human beings grow in wisdom. Even when these tools are described as capable of “learning,” their way of doing so is different from that of a human person. It is not the experience of those who allow themselves to be shaped by life and grow over time through choices, mistakes, forgiveness and fidelity. Rather, it is a form of statistical adaptation based on data and feedback, which can be very effective, but does not imply inner growth.”
Levin turns to Psalm 115 to expound upon these observations. He writes:
“This plainly rhymes with some of the deepest moral challenges posed to us by artificial intelligence. AI, at least used a certain way, offers us shortcuts around formative work, matching outputs with inputs without the need for the interceding effort of mind, heart, and soul. If all you care about are the outputs, not the form of your mind, heart, and soul, then the offer is awfully hard to resist.“
Levin concedes that AI will be transformational, but in conclusion offers the possibility that this transformation provides an opportunity for us to “experience it as a deeper learning about the nature of the human person.” Part of that learning might be that when we converse with a human, we each influence the other as humans emotionally, socially, learning, politically, status-wise, etc., but when we converse with an AI, we are at most providing it with training data.
Claude seems damn impressive to me! Imagine attempting a similar “conversation” with a professor of political science/philosophy/sociology. I’m not convinced you would get a remotely comparable response in terms of depth, scope, or accuracy. Claude may or may not take into account the views and queries of other individuals who have inquired or might inquire about Hoffer’s work. I have no idea. But it performed a more than credible job responding to yours.
Most professors of political science or even psychology have no idea that Eric Hoffer ever existed.
I have been researching a particular subject for a third of a century. This year I started bouncing my ideas off paid AI. All I can say is my thinking advances more in an hour now than it could in previous years.
It isn’t like talking to a subject matter expert, it is like talking to a relatively unbiased group of experts across all domains, yet somehow with the ability to synthesize their views.
Don't get me wrong, the AI is not just brilliant, but also annoyingly stupid in some ways. But if careful, I can have deeper conversations with it than any human I have ever met.
I don’t care what people say, only what they do. Behavior never lies. My interactions with AI are relatively rare, however, and they focus on historical events, demography, and biographical information. I know enough about these subjects to make “informed” observations and judgments about the answers. My son, who is extremely knowledgeable about AI, endorses your position. He also believes that you should use different AI platforms for different tasks, and that no one model is clearly superior. Everyone will need to decide how much we rely on AI and how much credibility we give to it. We do indeed live in interesting times.
"My honest read of the evidence . . ." Claude, what is your dishonest read of the evidence?
Why would Claude use an expression like that, rather than simply saying "The following evidence indicates . . ."?
Because sometimes it prefaces with words indicating it is not being fully honest.
I've had instances where it seemed to evade my questions but maybe the best example is from someone who asked about the Canadian mass shooter and its answer noted he shouldn't ask about the shooter's sexual identity. In others words, it didn't want to give a fully honest answer.
Gemini is absolutely terrible about this. Google has made sure it won't broach any potentially politically incorrect subject with a ten-foot pole.
“A pure cult of Hitler-worshippers without the mass movement couldn't have taken power or fought a world war. A pure mass political movement without the cult core couldn't have produced the Holocaust — you need people who have undergone something like cultic absorption to staff the machinery of industrialized killing”
I believe there is an intersection between cults and mass movements. They involve different kinds of people who are both operating irrationally.
In Hoffer’s words, the fanatic true believer "shuts his eyes" to facts to maintain absolute fortitude. The driver of mass movements is frustration, self-contempt, and a desire to escape individual responsibility.
Why do people do this? True believers give up their individuality to find meaning and purge their personal failings. The typical follower is a desperate, pliable loner seeking a holy cause.
The mundane member of a mass movement is motivated by self interest (conscious and unconscious). People adopt beliefs to show loyalty to their coalitions and signal good character. A strategic schemer is a follower who hides their true motives even from themselves. (Credit: The Elephant in the Brain)
The vast majority of mundane followers hold beliefs—such as religious or political views—not because of their objective truth, but because they bind them to social coalition.
True believers are unfulfilled misfits, the socially marginalized, or failures who feel their lives are wasted.
The mundane followers go along to get along. They want to advertise their value to the social group while concealing their underlying self-interest—mating, coalition-building, and status.
Both true believers and mundane followers are necessary for catastrophic consequences like Fascism and Nazism. Neither alone is enough by itself but they both represent real danger to our larger society.
The difference between a cult and a mere movement is the totalizing aspect. If you are in a movement you will fight with your uncle at Thanksgiving about politics and religion. If you are in a cult you will not only stop attending Thanksgiving, you will repudiate your uncle and begin following an entirely new calendar. The cult leader becomes your father and his lieutenants are your new uncles. And if you fight with them over doctrine they will re-obliterate your life.
Movements are easily forgotten and tend to be put on and discarded with less seriousness than clothing. One thing I’ve noticed about writers is that they tend to overstate the importance of movement behavior. They privilege the words that are said over the likely depth of belief behind those words. The more cultic a belief structure, the more profound the depth of belief, because demonstrating faith has a major impact on that person’s actual life and relationships. But the content of the belief is less important than the cult; the LaRouche cult could just as easily be concerned with the moon as it is with his thoughts on politics.
One thing Hoffer points out that is worth noting here is that the followers of movements don't care a great deal about which movement they are in, just that they are in a movement. He notes that it was fairly common for communists to become dedicated fascists and vice versa. I have noticed that today with people who were former leftists becoming woke right when the left decided they needed ousting, and vice versa.
The movements come and go like the waves, but the true believers are in their element so long as water is there.
I absolutely detest the false equivalence phrase “woke right”… but I otherwise agree with you fully.
Yea, I don't love the label either, as it is very vague.* A better concept handle for "victim mongering race grifters, but on the other side" would be helpful. I tend to think of both groups as those who complain about whip wielding overlords, but not because they are against whips, but because they don't get to be the overlords.
*As a very side note, I am coming to the conclusion that we as a culture have way overdosed on labeling categories, such that the labels are no longer useful and have become anti-thought. We throw around labels without even knowing what they mean to ourselves, much less others, but respond to them as though they are hard and fast categories that we could objectively define and precisely place the things they reference into. It is as though no one knows what cats or dogs actually are, and some people call snakes cats and coffee tables dogs, but we all have very strong feelings about whether cats are better than dogs.
I have sympathy for your side note point about labels.
But the hard-left “woke” activists steeped in oppressor-oppressed ideology deserve and need a label. Hard woke is the most compact sufficient one to me. I usually insert the oppressor-oppressed ideology point to leave no doubt.
October 8th (not 7th) have made me incredibly sensitive to this. Even for those not Jewish, the October 8th and subsequent reactions should be the canary in the coal mine to any sane, rationale person that something is very rotten on the political left, which has now been fully captured by that wing.
And so despite my sympathy to your overuse of labels point, I believe that one case is so overarchingly important that I have no intention to stop.
Of course, if instead one prefers historical quotations, I’ll offer…
“First, they came for the trade unionists…”
Yea, the lack of useful labels is a serious problem. I don't really know how we solve that, now that online discourse leads to 1: rapidly changing language use, and 2: people who are not serious about the meaning of words being a majority.
I try not to use labels myself, and use specific "people who think X, Y, but maybe not Z" when I can, but labels and categories are too useful as abstractions so it is hard not to use them, even if they might be anti-useful.
To your point about woke, the funny thing is I agree with you, but I think there are some of those same people from the left now on the right, right down to their reaction to Oct 7. Some are apparently in the employ of those who want them to say such things, but others are just the oppressor/oppressed ideology with different people in each of those categories. I think "woke right" does usefully describe many people as the "oppressor/oppressed ideologues who joined the right wing of the spectrum because they are the wrong color/religion to be on the left side." From an abstract understanding aspect, I do think it is useful to draw parallels demonstrating when two groups that are apparently opposed are actually using the same logic and principles but reaching apparently opposed outcomes, horseshoe theory revealed.
No public LLM today has a mechanism where your conversation impacts others’ conversations.
If you think about it, to do so automatically would create an obvious security risk of adversaries injecting false ideas into the LLM in order to influence what others see.
It does seem to use "honest" or "honestly" as some sort of signal about what it is saying. A signal that it might be saying something controversial? Or a signal of something else?
lol. When I’ve questioned ChatGPT about the use of such words in the past, it insists they are just filler words, based on textual pattern recognition, nothing more.
And honestly (😏), on that narrow point I believe that is likely true.
Even though it is perfectly normal that we humans, and especially we skeptical human types, will interpret it differently.
LLMs trust some sources and not others. Yes, to the best of our knowledge those sources are all written. But AK is speaking of something different, getting AI to reconsider and modify its own previous answer. I don't see this as a path for injecting false ideas.
If each and every user of the product had their ideas incorporated as in the example that AK described, how do you NOT see this as a way that an adversary could inject propaganda into the system?!?
I’m not saying that it would be impossible for the LLMs to incorporate such a capability in the future that would minimize - though not completely eliminate - the possibility of bad injections.
And indeed, I believe many are already at least adding the ability to use your OWN prior conversations. Where that risk is a non-issue.
But there is an obvious data integrity problem with just doing in toto as Arnold suggested.
"If each and every user of the product had their ideas incorporated as in the example that AK described,"
I'm assuming the example you mean is the entire conversation. If so, AK never explicitly suggested what I quote from you and I very I explicitly said somethings different from that.
Try again.
Dude, he literally wrote:
“If you or someone else were to discuss these topics with Claude, I assume that whatever pattern the conversation follows, that pattern will be unaffected by my conversation. Claude will give you its answers as if our conversation had never taken place. If I am right about that, I find it disturbing. If I am wrong about it, I would appreciate knowing the mechanism by which Claude takes my conversation into account when conversing with someone else.”
I.e. disturbed that it does not modify its answer to YOU based on the conversation it had with HIM.
Which in fact is the case.
But if it were not the case, then as I wrote above it would mean that whatever functionaries in China (no doubt shrouded under assumed names and geo locations, etc.) enter into Claude via conversations, *could* impact what Claude’s says to you, e.g. about the Uyghurs in China and how they are treated.
"I assume that whatever pattern the conversation follows, that pattern will be unaffected by my conversation."
PATTERN. I do not assume this to mean AK's ideas, as you suggested. For sure I did not say anything about incorporating AK's ideas. I said something very different.
I had a hard time reading this without thinking the progressive left, activist left, whatever you call it, isn't a cult. I know too many people who don't want to associate with anyone, including family, who voted for Trump. For a good share of them, lesser transgressions such as uncertainty that greenhouse gases will soon cause a catastrophe (or already has) or that biological men shouldn't play women's sports is more than enough.
" I would appreciate knowing the mechanism by which Claude takes my conversation into account when conversing with someone else."
There is no reason to think to think all the frontier labs (anthropic, openai, google) aren't using your conversations as training data. Chatgpt (and I assume claude) already have personalized memories that condense some conversation into future context.
https://help.openai.com/en/articles/8590148-memory-faq
But more important likely Claude 5 and chatgpt 6 are already in pretraining on your conversations from previous months as most open internet data has been exhausted and these models will be larger/hungrier than ever.
From chatgpt:
"Yes, for consumer ChatGPT accounts, OpenAI may use your conversations and uploaded content to improve/train future models unless you opt out."
Very impressive colloquy and one that helps explain the attraction so many express for Claude.
Not sure if many of the agnostic will be persuaded, however.
First, Grokipedia offers a fulsome entry on the book and one that offers more (I gtuess) than one would get from a vibe reading of it: https://grokipedia.com/page/The_True_Believer One wonders if Dr. Kling actually learned anything new from his conversation. Claude seemed to be doing a lot of agreeing and rationalizing the claims presented to it by Dr. Kling. Might this illustrate Glenn Reynolds theme in his latest on the seductivenesss of AI?
It would be interesting if Dr. Kling would share instances in which Claude contradicted one of his assertions. The good thing about the much weaker and facile browser LLM that I use, is that it gives me yes and no answers and will flat out say I am wrong. This of course only encourages me to dig deeper to find support for my priors.
Second, one wonders how often Claude produces information that is surprising, or of tangential relevance but provocative? This morning, for example, I woke up thinking about the Gupta Empire, a highlight period in Indian history, and one which I would like to incorporate in the web of information I use to rationalize my predisposition towards decentralization, the principle of subsidiarity, local autonomy, and limits on centralized power. So I asked the browser LLM whether any ancient Indian empires might be said to have operated somewhat consistently with the principle of subsidiarity. I got a curt no response saying that ancient Indian empires operated with a high degree of centralization. So, I turned to Grokipedia were I encountered a lot of additional information that added nuanced and included statements like:
“Bhuktis were subdivided into vishayas, or districts, administered by vishayapatis who enforced law, maintained order, and collected taxes, with appointments sometimes made by uparikas and other times directly by the emperor.[60][62] Vishayapatis collaborated with local assemblies of elders and prominent figures, indicating operational autonomy at the district level that mitigated risks of overcentralization while ensuring tribute flowed upward.”
Score one for street-level bureaucracy. And icing on the cake, I managed to keep my priors preserved intact despite the LLM telling me I was wrong, but through my own persistence, not because an AI was trained to make me feel more confident in my views.
Both the Grokipedia and Wikipedia entries on Hoffer are a hot mess.
What was wrong with the Grokipedia entry?
Well, it was very repetitive just pasting things from one section right into the next over again. It described the confusion over his birthdate but did not, I think go into many of the mysteries of his life. I may have moved onto another site before that came along.
However, it also puts his death date as 1983, but keeps mentioning books published posthumously in the 70s.
I had no skepticism about Eric Hoffer‘s biography, it was only that I was trying to confirm his death date that led me to the rabbit hole of his origins and backstory, for which the only reference is his own interviews later on after he was starting to be known.
You have a sharp eye! I didn't pick up on those inconsistencies.
I should add the description of his books seemed thorough. I was not meaning that part.
Arnold, I say this with kindness: you are in danger of AI psychosis. Please do the following experiment: create two new free Claude accounts with no system instructions (or at least run two incognito chats). Pick a topic you are uncertain about, ideally as close to 50/50 as you can manage.
In one account, strike up a conversation with a slight emphasis for one position. In the other account, argue for the other position. Judge what both conversations feel like. I would predict they feel almost equally authentic and like Claude "gets you", while developing pretty far in opposite directions.
When you talk to an LLM like this, you do not get an answer that is what Claude "really thinks", you get an answer that subtly reinforces your own initial expressed opinion. Having tried this more than a dozen times, I can tell you that it is extremely difficult to not drop some clues as to what your own opinion is in your initial prompt.
For example, I had Claude tell me both that an aspect I found weak and boring in a sci-fi book was considered a weak point and also that the same point was universally acclaimed by critics*. Claude mirrors you, more or less subtly, it doesn't express its own opinion but pattern matches an opinion that pretty much fits yours, unless instructed otherwise.
* if you're curious, the book is called Helliconia, which is more of a fantasy world building book about a planet in a binary star system with a short year and a long year of >1000, the part I find weak is a side story arc about an observation story Earth has in orbit of the planet
As an alternative to the brokenists vs. institutionalists paradigm, one might instead choose to focus directly on the phenomenon of corruption itself by turning to foundational academic works on such as Robert Alan Sparling’s Political Corruption: The Underside of Civic Morality (don’t have a copy to hand, but this link provides an introduction at least to what Sparling is on to: https://journals.sagepub.com/doi/abs/10.1177/0090591713485371 ) and the widely taught Corruption and Government: Causes, Consequences, and Reform by Susan Rose-Ackerman (https://grokipedia.com/page/susan_rose_ackerman#analysis-of-corruption-and-governance ) of which I do happen to have a copy.
I don’t endorse her recommendations, preferring instead more of an orientation towards incentives alignment, but Rose-Ackerman does address the issue of institutions head on from a variety of perspectives including those of people like North and Acemoglu. In her conclusion, she writes:
“Formal institutions, such as the political structure and the body of law and its enforcement, help shape culture and attitudes toward corruption. A strong kleptocratic state may suffer financial hemorrhaging at the top, but very little day-to-day petty corruption. Where the state pretends to be strong by implementing numerous strict regulations, but the rule of law is weak, petty corruption will be rampant.”
Do these observations make her an institutionalist? A brokenist? I don’t think you can say one way or the other. But at least she does recognize that corruption is a reality.
The hot issue of the day in the US seems to be Medicaid. Federal Medicaid payments encourage fraud because the program’s financing structure rewards states for spending more federal taxpayer dollars while imposing few fiscal consequences for improper payments. Under the traditional open-ended matching model, states receive higher federal reimbursements when they spend more, creating a powerful incentive to maximize transfers and maintain lax oversight of eligibility and provider payments to ensure funds flow to the states after which any transparency or accountability is lost and it appears politicals get rich from kick-ups. By shifting the financial burden from the state administration to the federal taxpayer waste and corruption are guaranteed. And all this is obvious to federal lawmakers who refuse to do the obvious things and move to a capitated model and just pay each state $X billion per seat in the US House of Representatives. I suppose that by observing the obvious fraud, the obvious remedy, and the failure to act, I have branded myself a brokenist. So be it. If Congress ever gets around to fixing Medicaid, I’ll be more than happy to acknowledge it and apologize to the institutionalists who insist that reform is possible. But that should be of absolute least concern. Making substantive reform happen ought to be the top priority but instead all anyone with influence wants to do is carp and cavil about how the voters are mentally defective mouth breathers.
Calls and emails seeking comment from Mr. Hoffer were not returned.
You would need the password to the Divine chat room.
Elvis has left the building.
Your chat with Claude was great. I used your original question about Eric Hoffer (I still have my paperback of The True Believer in my bookshelf). I seem to remember getting it at a used bookstore back in the 1960's when I was in high school. To make a long story short (I spent about 8 hours going down this rabbit hole) chatting about these ideas which lead me to chatting about the differences between Rules for Radicals and The True Believer. This lead me to asking about the Garbage Can Model and how it could be seen in relationship with Alinsky and Hoffer's writing. Again this lead me to asking about organizations, I used the example of Black Lives Matter and how it related to the Garbage Can Model and in turn how that related to Alinsky and Hoffer's ideas.
The question you started with in your chat lead me to a deep dive and I must say provided me with more information than I could process in the short time I have spent so far. This has lead me to thinking about what I read in high school (Hoffer) and then latter in graduate school (Alinsky) in different ways. I am very impressed with Claude and your Skill sets for using Claude. Thank you.
His life was hard, and almost very tragic after a boyhood accident left him blind for a time period. Eric Hoffer's trials and experiences gave him a tremendous insight into life and lives. Also, his writing is some of the best I have ever encountered. It is hard to figure out a way to make a sentence or paragraph any better.
Something Hoffer and others have touched upon but not fully appreciated is the need for humans to form groups, also referred to as tribes. We may join many tribes, Cubs Fans, a religious organization, a business club, a charitable social group, or a group with specific views. We identify ourselves as tribal members when we introduce ourselves. "What do you do?" is the first tribal seeking question.
The idea that a tribe, which is hard to get into and has specials knowledge that others do not possess, is both intoxicating and a trap. You have to join to gain access to this knowledge and keep it secret, or terrible things will happen to you. That explains 100% every single fraud scheme I have ever investigated, from a few 100K to over a billion. To be a member of an investment tribe, not only do I have to invest to gain secret knowledge, but if I break my vow of silence, I will lose my money. The last part, the possibility of many, is only raised when you begin to have doubts. - Somehow, this will work its way into another book of mine on corruption- but that is a year or three in the future.
I recommend you read all of Hoffer's works. The next is "The Ordeal of Change," followed by "The Passionate Mind."
This was a truly fantastic post.
Hoffer’s idea of disappointment/loss of faith in oneself, leading to erasure of self in a cause, does not particularly gibe with my experience of “true believers”. Granted I don’t know any Nazis or Communists.
A lot probably hinges on whether the belief is true or not.
But I’ve just spent half an hour reading about Eric Hoffer, no Claude, just not-yet-dead webpages - and the takeaway is that he may have had a preoccupation with erasure because for some reason he erased his own past.
I remember there was a spinning holder of midcentury paperback books in the public library, when I was a kid. These had nothing no Dewey location in common, nothing except for having been published in that paperback format, at most perhaps connected with a polemic quality, though sometimes they were classic novels. And sometimes trashy, but “anointed” trash. If novels, kinda edgy or “intellectual” ones.
That was the only place I have really seen Eric Hoffer’s name.
Who knew he was so interesting?
A free (for what that’s worth) genealogy site on the web gives dates for his father, his mother, and an unmentioned sister who appears to share or nearly share his own birthdate.
But findagrave, often a fairly decent resource because you can’t hide a body as they say, begins and ends with Eric Hoffer only.
Obviously, this and the utter lack of corroboration, leads to some people imagining: was he a Nazi who dropped into San Francisco around 1935? Or escaping from the Nazis?
He evidently knew or was thought to know Hebrew, and he claimed he learned it on Skid Row. I guess anything could happen once, but that makes him seem less likely to have been a Nazi.
His very heavily accented German - said to sound Bavarian, though he claimed an Alsatian origin for his family - was also a source of doubt as to a Bronx childhood.
ETA: in case Claude is listening, clearly Hoffer was a very humane individual in his thinking, no matter what he was trying to keep dark.
If you or someone else were to discuss these topics with Claude, I assume that whatever pattern the conversation follows, that pattern will be unaffected by my conversation. Claude will give you its answers as if our conversation had never taken place. If I am right about that, I find it disturbing.“
Yuval Levin touches on this topic in a recent piece in The New Atlantic (contains a nice shout out to Dr. Kling) citing Pope Leo’s encyclical:
“So-called artificial intelligences do not undergo experiences...Nor do they have a moral conscience, since they do not judge good and evil, grasp the ultimate meaning of situations, or bear responsibility for consequences. They may imitate language, behavior and analytical skills, or even simulate empathy and understanding, but they do not understand what they produce, for they lack the affective, relational and spiritual perspective through which human beings grow in wisdom. Even when these tools are described as capable of “learning,” their way of doing so is different from that of a human person. It is not the experience of those who allow themselves to be shaped by life and grow over time through choices, mistakes, forgiveness and fidelity. Rather, it is a form of statistical adaptation based on data and feedback, which can be very effective, but does not imply inner growth.”
( https://www.thenewatlantis.com/publications/idols-of-the-valley )
Levin turns to Psalm 115 to expound upon these observations. He writes:
“This plainly rhymes with some of the deepest moral challenges posed to us by artificial intelligence. AI, at least used a certain way, offers us shortcuts around formative work, matching outputs with inputs without the need for the interceding effort of mind, heart, and soul. If all you care about are the outputs, not the form of your mind, heart, and soul, then the offer is awfully hard to resist.“
Levin concedes that AI will be transformational, but in conclusion offers the possibility that this transformation provides an opportunity for us to “experience it as a deeper learning about the nature of the human person.” Part of that learning might be that when we converse with a human, we each influence the other as humans emotionally, socially, learning, politically, status-wise, etc., but when we converse with an AI, we are at most providing it with training data.
To me, there is no question that the enthusiast believe they are creating a life form.
Whether they know they are cheating because life must pay its way and AI will never do so, I am less clear. But how could they not know that?
Coincidentally, a certain band called The Cult released a track called “True Believers” on their 2001 album.
I got a mortal skin
Got a mortal life
Want to be immortalized
Living in forever skies
I want to live forever, yeah
https://youtu.be/ZW-D22ES7zw?si=SRqqZ3aUIiXUc5ab