Submit Podcast or Channel

The Three Most Important Words We're Taught Not to Say

October 24, 20250 comments♥ 0
00:00Good morning. As a podcast host, there is one answer that I love to hear when I ask my guests a question, but I rarely ever do. To me, this answer is a signal of maturity and nuance and honesty. And it's not trying to give an answer to all the world's problems. That answer is, I don't know. Why is hearing I don't know so refreshing to me? And I'm guessing to many of you, which is why I'm doing this.this frankly. We are all members of a social species in a modern culture that's turbochargedby energy surplus and social technology. But in this modern setting, we still seek statusand respect a product of our evolutionary wiring. And because of this, in most publicsettings in today's society, including and especially the media, we value, and I would argueovervalue confidence, bravado, and certainty. So saying, I don't know, is now seen as a sign ofweakness, not of wisdom. So when you hear it in a public setting, it's like an antidote to ourcultural consensus trance. And I'm beginning to think that the reluctance to express, I don't know,
01:20or its equivalent, out loud, is a fatal flaw in our culture as we've been.begin to discuss our vastly complex and risky and rapidly approaching future.And so today I'd like to unpack this a bit.So back in the day, 25 years ago, no longer ago than that, 25 years ago is when I left.When I started at Solomon Brothers in 1992, so over 30 years ago, Solomon Brothers was one ofthe coolest places on Wall Street.And in the training program, which was much revered and respected, they kicked our asses.And a lot of times they would ask you a question about simple bond math, like what's the duration of a 30-year note?
02:19And then they would ask you a little bit harder question.What's the ticker symbol of some stock?And then they would ask you a really hard question.And after you answered the first two questions, like, yeah, I know that.Yahoo, Y-H-O-O or whatever, then they would ask you a question that you weren't supposed to beable to ask.And what ended up happening, especially as we wanted to impress our bosses, we would make somethingup, our guess.
02:43And they would come down on us so hard, like, no, what you were supposed to say as eventualsalespeople who would be talking to billionaires and institutional managers is, I don't know,but I will find out and get back to you.So that concept was drilled into me in my early 20s.But in the intervening 30 years, I don't know is not spoken in our culture.And I want to try to understand why and what it means.So there's many levels of this.And our culture today doesn't merely tolerate and accept overconfidence, but I think we actually
03:27prioritize it and basically pay a premium for overconfidence in our culture. Social media feedsboost the really bold claim that then goes viral as opposed to the nuanced one that is moreaccurate. And I feel that dynamic in spades hosting the content on TGS the past few years.TV and news producers, they booked the guests with the crisp and sharp take, not the carefulqualifying one. And in classrooms, the quick hand beats the methodological thinker. And I know this because even inthird grade, I had the fast hand. In boardrooms, in C-sweets, it's the fast, confident answer ranks a lothigher than the humble hypothesis. So our modern status economy runs on conviction and nuance and cautionslow us down. And there are three, at least three, intertwining reasons why confidence and
04:30lack of humility rise to the top in our current culture. First, at the physiological level,within an individual, uncertainty itself feels bad. Because uncertainty is more than justan abstract concept. It's actually a bodily state within us. Human brains are prediction engines,It's always guessing what's going to come next and then checking that guess against the realityin front of us.And when the world becomes more chaotic, these prediction errors spike and kick in our sympatheticnervous system, aka our alarm network.And so when our stressed system releases cortisol and things happen like a tight chest or a fluttery
05:18stomach and increased heart rate, these inform our gut that something is off here, even beforewe can explain why.And furthermore, being uncertain, by definition, occurs when we build multiple mentalpossibilities and hold them all at once.And doing this requires more energy in the form of glucose.But when we're now running parallel mental scenarios and models, inhibiting quick,answers and context flip-flopping, our bodies don't like all this energy use and inefficiency.And as this stress increases, our bodies push us to pick some story, sometimes to pick any
06:03story to shut the alarms off.So evolution wired us as this way on the plains of Tanzania back in the day.It was safer to assume that the Russell in the bushes was a lion than it was to assumeit's the wind because it was better a false alarm than a fatal pause or maybe fatal clause.Clause isn't clause.And now today, a modern society reinforces this group's reward decisive signals over caveats.So I don't know can professionally feel like a status risk.You put that all together and at the level of the individual human physiology, uncertainty feels
06:53both uncomfortable and potentially costly.So it's no wonder that fantasy and doom often become our defaults rather than sittingwith the unknown open to learning.Okay, that's the first level.So building on that at a higher level, beliefs once they're formed develop antibodies to change.We all carry what's been referred to as an ideological immune system, which is once a worldview works for us, we defend it.And importantly, especially the smart, well-read among us who get better at arguing their own side rather than testing it.And there's lots of modern research to back this up on highly charged and divisive topics.
07:43smart people aren't necessarily right more of the time, but they are better at rationalizingtheir own positions. This is called motivated reasoning, and it's very effective at preservingour personal blind spots. Then a layer above that, there are the cultural-wide humancognitive phenomenon like authority bias. Authority bias and many other related biases arebuilt-in tendency to trust and comply with confident signals of what's perceived to be highcultural status, titles and uniform forms and expert credentials over any actual evidence to thecontrary. It shows up over and over. Famously in the Milgram studies, ordinary people kept deliveringwhat they thought were painful and nearly fatal shocks because some guy in a lab coat
08:41told them, please continue. There was another classic hospital field study where 95% of nursesprepared an excessive, unauthorized medical dose based solely on a stranger's phone order claimingto be a doctor, even though the rules forbade this. There's many other studies who've shown thathumans are more likely to obey requests from persons in authority or perceived authority.this all doesn't seem surprising, but it does help explain why a lot of us tend to go on autopilot at times when we probably most need to be thinking critically.And these examples are all from different settings, but they show the same pattern.Signals of authority lower our skepticism and they raise our compliance.And this also partially explains why misinformation in today's world works so well, throw some money.
09:38at a confident spokesperson and you distill some non-science-based message good for your business or whatever.This is a high leverage return proposition.So lastly, we can see why in our culture people in authority rarely say, I don't know, because if they did,they'd be replaced by some of their leader, some of their person who's more perceived to know what they're doing.So basically confident guesses rise to the top.And almost as if by compulsion from what I refer to as the global economic superorganism,we have learned to speak past uncertainty and to round off any error bars and ultimatelyculture-wide, and this is a global phenomenon, to confidently act first and check later.
10:36And these incentives shape the outcomes we see around us.Projects start with rosy baselines and then end up with invariably with cost overruns.Politicians will campaign on certain guarantees and then walk them back when they're elected and actually have to govern.Even science gets pulled towards headline ready narratives.even in choosing guests for this show,I lean towards the articulate, confident, charismatic spokesperson for topics over the best scientist.So overconfidence in our culture is rewarded at the front end and punished, if at all, only in hindsight,which means the person who benefits is rarely the person or culture who pays.
11:30the public and the environment pay in trust and money and time and obviously to those following this showin ecosystem stability. What about AI? Well, AI adds an interesting wrinkle here. AI also rewardsconfident responses, turbocharging what is already a human tendency naturally because humans created AI.So in competitions, large language models are trained on what wins, not on what's true.So you've heard of the concept of hallucinations.This is when a model, a large language model, confidently generates an answer that isn't true.Chat GPT5 main software hallucinates around 10% of the time with Internet access and about half of the time without the Internet.So why do chatbots make stuff up?
12:28It's because they were trained to answer, not to pass.It's a fundamental logic difference.So if you think of a quiz show where guesses get points but saying, I don't know, gets younothing, and you play that a million times.Over time, such a system with lots of iterations, that system would learn to speaksmoothly and confidently, even when it's unsure. In a recent paper, OpenAI, who makes Chad GPT,even stopped referring to hallucinations as bugs and referred to them as statistical destiny.So you combine humans and AI and you get even more overconfidence and less humility. There was a new Stanford paper
13:24that just coined the term Mollocks bargain for what happens when large language models start competingfor attention, sales, or votes.And the results were striking, though perhaps if you followed this far, not unsurprising.Every gain in the model performance came with a bigger loss in honesty.In effect, more deceptive marketing, more disinformation in political campaigns, and more fakeand a harmful social media post. And as an aside and not one I enjoy thinking or talking about,I worry a lot about the merger of AI with military capacity. I've been informed by people who are in aposition to know that we've actually avoided perhaps a dozen or even more potential nuclear
14:19wars in the last 50 years. And mostly because at the time a single human wasunsure. And they chose to, something's not right here. Let's wait for more information. And therational speed bumps of uncertainty and waiting might disappear if large language models trained onworst case possibilities eventually replace the human interface and the human decision making andjudgment. This is one of the few things that wakes me up at night.Of course, you're aware this is not just a machine AI problem, but really a mirror intoour cultural values and behaviors because we train ourselves the same way.As I mentioned before, media cuts to the confident clip and companies promote decisive,
15:18confident talkers and politicians prioritize simplicity.and easy to understand things over truth and complexity.Economists tell stories of infinite growthand that technology will be able to solve any challengesalong our way.The result is human hallucination.Stories we tell ourselves and act onthat wildly outpace our energy, ecosystem, and time constraints.
15:50And then we're shocked, shocked when reality shows us.to stare us in the face and remind us.So what are the costs of overconfidence?People following along this platform are well aware of the modern cost of overconfidenceand certainty on the environmental side.But beyond the ecological impacts that we're learning more and more about, think of exampleslike the Challenger space shuttle launch.So managers waved off the engineers caveats about the O-Ring,
16:24and low temps, and it resulted in seven lives lost.I watched that live, by the way, at Cronzidge Hall at University of Wisconsin when I wasat college.Strange how the amygdala has these little emotional memories during the lifespan of a human brain.What else?The housing bubble, ratings and models said, this is safe, leverage and optimism.said, let's party. And that all again resulted in another global crisis. The Deepwater Horizonshortcuts looked efficient until they weren't. Governments do this all the time. Big projects are
17:11sold as on time and on budget and important and they morph into years of overruns and then bailouts.So it was the optimistic baseline. That was the mistake from the beginning.This is a repeating pattern in our culture.Certainty beats caution.And the costs come later.And the costs come later.And the costs come later might actually currently be a good mnemonic tagline for homo sapiensand we're starting to see the costs that are coming later.
17:52So what do we do with this?I almost said I don't know, but I have some directional ideas.One could argue that overconfidence and lack of humility and caution is one of the core underlying drivers of the maximum power principle, which itself is underpinning global human ecological overshoot and the impending great simplification, which I suppose is itself maybe.its own frankly in the future. But we have a Trump card, which is self-introspection,learning, and the fast pace of culture change. So injecting, I don't know, as a culturally acceptedphrase could be impactful, in my opinion. I don't know shifts the dynamic back to curiosityand the possibility for change. It re-engages our prefrontalcortex, it widens our time horizons, and it makes cooperation with others easier because we're
19:02no longer defending an identity whose scaffolding was certainty.If you can't admit that you don't know, then you can't really hear anything else.Admitting uncertainty lets us run experiments instead of arguments, and it creates room to developscenarios and more importantly extend conversations with people who we mightotherwise disagree with and not talk to. So how do we put this into practice? In someways this human authority bias phenomenon dovetails, I suspect, with thepsychopathy dark triad story that I've been talking about recently.Now, I'm going to think on that more.
19:54I think there's an important overlap there that I need to muse on.But what do we do in practice to shift this dynamic?I don't know.But here are a few directional suggestions.So AI itself is evolving in emergent ways.And Open AI in their YAI models hallucinate paper, they proposed a fix to hallucinate.Not more data or the size of compute or the model sides, but merely a change in the rules on how models are trained.And they're considering rewiring the benchmarks. So I don't know. Specifically, I don't know is rewarded instead of punished.
20:43And if this works and confidence thresholds are added to calibrate behavior, the thinking is that models will stop bluffing.when they're uncertain.And this is really encouraging, I think.But of course, a big caveat is that downstream from what Open AI is doing are all theircustomers, the corporations and the confident individuals.And convincing them to embrace new models that favor I don't know over accuracy pornis perhaps a harder challenge than just patching the chatbots themselves.But what they're suggesting is to change the scoring.
21:23So humility wins, or at least has more equal footing.In effect, to penalize confidently spoken errors more than they penalize uncertainty and alsoto give partial credit for appropriate expressions of uncertainty.So Open AI is doing this in training their next versions of AI.What if our society did this?So what else?What about us as individuals?Again, these are somewhat confident guesses on my part.
22:02I'm into the, I don't want to highlight a problem and then just say goodbye.So I'm putting some things I'm uncertain about that might directionally apply to you, the viewers.So maybe pay attention to yourself and notice when you're giving an answer that you don't fully understand.or know to be true and recognize this and call it out to your version of little Nate sitting onyour shoulder. Of course, it would be little Kathleen or little Joe or little Feng Li or whoeveris watching this. And then fact check it. Get in tune with your gut, your gut feeling. And you mightfind by listening more and more to the little part of you that says something is off.off with this guy. Something is off with this story. Something is off with what I just heard. It may
22:59become more and more accurate and sensitive to when others are trying to sell you something that'sonly half baked. Actively, consciously calibrate. Put probabilities on your own claims.I think this is going to 60% happen this week, 30% next week, 10% the week after oror something like that and then score yourself when reality arrives.I think this could be super helpful,but of course this gets into the collective action problem.If you're the only one or one of the only ones that are doing that,then it might not be that helpful.
23:42If a lot of people do this, I think it would be helpful.And then the concept of red teams make dissent and uncertainty a formal role.in your work as opposed to a risk and then rotate the whoever's tasked with being the skeptic.Rotate that role and thank these people publicly for playing that role on your team,whether that's at your community board meeting or at your organization or even your family.If it works for machines rewarding honest uncertainty, why couldn't it work for us?So in conclusion today, in a society that is assailed from all angles with social and environmental problems and too much information in addition to gambling, pornography, and shopping and new AI videos of some Sasquatch-looking creatures selling a new tech gadget and all of this is available 24-7 on the internet to increasingly overly full.minds, we are moving further and further away from the cultural ability to express, I don't know.
24:59Because such an answer in today's world implies weakness rather than wisdom.And someone on TV, someone testifying to Congress, or someone publicly asked for answers to ourfinancial ecological problems, replying, I don't know, but I can find out and get back toyou in our culture would quickly be.replaced by someone with a pithy, witty, or confident answer, and with all three, they'll bebranded an expert and invited back.So yes, modern human culture rewards overconfidence and overrewards overconfidence.What's the answer?
25:37I don't know.But I suspect if we're able to change how we keep score.At home, at work in media, in decisions about the future, so that truthfulness and humanhumility get equal time as performance and and posturing, I expect our systems would naturallyget smarter and kinder. So what if those three magic words aren't the end of the sentencebut the start of learning? Thank you. Talk to you next week.