Why Warren Buffett’s superpower is an Achilles heel for AI
The great investor instinctively knew that humans are much smarter than computers in volatile environments. So he bet on common sense.
By Angus Fletcher
Trillions of dollars are now at risk because global investors are making the same mistake that caused the dot-com bubble, Black Friday, and the 1929 Wall Street crash. The mistake is: To believe that numbers can predict the future.
The mistake originates in logic. Logic reduces life to statistics. And statistics convert real-world business into spreadsheets of digits: total income, net profit, employee productivity. . . Those mathematical values are then scoured by logic for patterns and trends. Which is to say: The spreadsheets are fed into AI machine learning systems or analyzed by traders who (by engaging Daniel Kahneman’s system 2) have rigorously purged their minds of bias. Until — with probabilistic precision — the cost of future commodities is computed, prompting a set of rational epiphanies: This stock will rise! That stock will fall! Go hard, with maximum leverage! Go wide, to capture the entire index!
Years ago, Warren Buffet realized that this logical method was bullshit. There’s no way for math to predict asset prices because markets aren’t platonic algorithms. They’re biological ecosystems. And biological ecosystems, as a guy called Charles Darwin once observed, are sites of incessant competition where victory comes from innovation.
Innovation obsolesces yesterday. In business, the data you gathered from last quarter can therefore be used to calculate … last quarter. The only way that it can foretell next quarter is if life has become a TV sitcom where tomorrow’s episode takes place in exactly the same world as yesterday’s.
When Buffett grasped this, he abandoned logic. And turned instead to common sense.
Common sense is typically viewed today as a logic function. But this is wrong. (Just as it’s wrong that inference, abduction, and other biological mechanisms of practical intelligence are logic functions.) You can see this by looking at computers. Computers think in pure logic. Their brain, the Arithmetic Logic Unit, is composed entirely of logic gates. Yet AI is legendarily incapable of common sense. It’s the ultimate example of book smarts not cashing out in street smarts. AI can memorize a million libraries but will hallucinate wildly when you ask it questions like: Why did my banana eat the color blue? Any human will tell you: Uh, I’m pretty sure that didn’t happen. Every AI will give you an elaborately bonkers explanation.
That’s because common sense requires a cognitive process unachievable for an infinite number of Arithmetic Logic Units. The process is: Detecting “unknown unknowns.”
Unknown unknowns are the inevitable result of innovation. (They differ from known unknowns, which can be hand-coded into computers, allowing for newfangled neurosymbolic AI and old-fashioned decision trees.) We humans can spot unknown unknowns because our heads contain neurons. Neurons are far more mechanically sophisticated than electronic transistors, enabling the brain to think not just in equations but in actions. Thanks to that action-thinking, you (and all animals) are born with a healthy anxiety, aka “spidey-sense”: Out in the murk, concealed in places I can’t see, other lifeforms are evolving novel strategies to ambush me. . .
Innovation obsolesces yesterday. In business, the data you gathered from last quarter can therefore be used to calculate … last quarter.
Such novel strategies are the doom of AI, and more generally, of logic. Logic exists in the mathematical present, an eternal state of equation that prompts AI to think that its current knowledge is what it has always known — and all there ever is to know. This logical but nonsensical feature of computer cognition is why large language models like ChatGPT lie. They’re not trying to deceive. Quite the opposite: They aim to tell the truth. But when queried about something they don’t know, they don’t know that they don’t know. So they fabricate guilelessly, filling the gap in their knowledge by extrapolating from past trends. Why did your banana eat the color blue? Well, eating is a sign of hunger, so your banana probably wanted more primary pigment in its diet. . .
Young Warren Buffett was blissfully unaware of AI and its endemic foolishness. But he was chock full of common sense. And common sense told him that the human brain worked the inverse of a computer. A computer uses equations to seek verification from data; the human brain uses actions to seek falsification from the unexpected.
Computers, in other words, are passive learners. They amass facts and act only when prompted. Humans, meanwhile, learn dynamically. We take the initiative — and base our behavior not just on what we’ve previously memorized but also on the newness of our current environment. In familiar circumstances, we repeat what worked before. In novel circumstances (i.e. when our neurons sense unknown unknowns), we perceive: I don’t know exactly what to do. So, we improvise, experiment, and venture original behaviors.
That’s why human experts are able to act almost as smart as computers in regular environments — and why human experts are able to act much, much smarter than computers in volatile environments where newness (and therefore, not-knowing) is high.
Buffett realized all this. So, he pivoted away from computer think. Ditching statistical prediction, he put his own commonsense twist on a financial method known as value investing.
Logic exists in the mathematical present, an eternal state of equation that prompts AI to think that its current knowledge is what it has always known — and all there ever is to know.
Value investing was pioneered by Buffett’s mentor, Benjamin Graham. A polymath who invented calculators and wrote Broadway plays, Graham worked on Wall Street during the 1920s. The instability of those boom-bust years taught Graham the futility of trying to predict the future, so he focused on the present; rather than speculating on which prices would increase later, he hunted down good bargains now. To identify good bargains, Graham looked for gaps between a company’s valuation and its earnings. He looked, that is, at spreadsheets. Although Graham distanced himself from market forecasters, he therefore held onto part of their logic.
Buffett didn’t. Instead of fixating on numbers, he placed weight on people, assessing each company’s human value. That value could be gauged in many ways, but Buffett’s preferred measure was: common sense. If a company’s leadership possessed common sense, Buffett knew: His investment was safe. Regardless of what the future brought, the company would turn a profit. In stable times, its leadership would aggressively accelerate past-proven strategies. In volatility, they would innovate rapidly.
But how can you measure a leader’s common sense? For a comprehensive guide, see my book Primal Intelligence, but as a quick start, do this: Ask a company’s leader to tell you their plan for the coming year. Then ask the leader how they would change their plan in response to a shift in market fundamentals: Maybe oil prices quadruple, war breaks out in the Pacific, or half the world’s telecom satellites go down.
If the leader pivots to a brand-new plan, they have common sense. Invest immediately. If the leader freezes, denies that the shift is possible, lightly tweaks their original strategy, or spits out a pre-made contingency plan — then they are thinking like a computer. Withhold your money.
By employing this commonsense test, you can cultivate your stock portfolio wisely. That’s because you’ll be doing what Buffett did and no numbers can: Determining an organization’s real value. The value of its people to act smart, whatever tomorrow’s unknowns.
Check out more of Big Think’s content:
Big Think | Mini Philosophy | Starts With A Bang | Big Think Books
I’ve been a student of Buffet for many years. This is the, ‘Most Common Sense’ explanation of his philosophy…and how ai works!
Thank you
Logic does not reduce everything to statistics. How people think about logic and have been taught logic for a century and a half or more is what has produced this distorted view of logic that everybody accepts and runs with. And then they end up throwing the baby out with the bath water.