Fumfer Physics 35: Cognitive Limits, Big Data, and AI’s Role in Human Reasoning

In this exchange, Scott Douglas Jacobsen asks what human consciousness cannot process adequately. Rick Rosner argues that people hit hard limits with big data, large parameter spaces, and even simple mental representations like number grids. Computers can find correlations, but humans struggle to hold enough information at once to test whether patterns are causal. Rosner suggests AI could surface correlations and generate wide-ranging analogies across culture at superhuman scale, while humans remain responsible for interpretation and meaning. He extends the point to scientific imagination—alternative cosmologies and modified-gravity ideas—and notes AI may help break cognitive ruts, even if it is not yet a top-tier theoretical mathematician.

“We cannot hold all the data in our minds, identify multiple correlations, and then analyze whether any of them are causal. AI—if it can hold large data sets in something like a Bayesian network—might help. Then it becomes our job to interpret them: meaningful or happenstance, trivial or real.”

Scott Douglas Jacobsen: What do I think human consciousness ultimately cannot process adequately?

Rick Rosner: Big data sets, for one—and large parameter spaces more generally. We simply do not have the capacity to manipulate or meaningfully explore extremely large data sets in our heads. We can defer to computers, which we already do. We do not even need AI to find correlations; conventional computing can do that. But the correlations you find in huge data sets do not appear to us naturally without extensive external manipulation. We cannot hold all the data in our minds, identify multiple correlations, and then analyze whether any of them are causal.

There is a reason we say correlation is not causation. If we cannot handle big data cognitively, we cannot adequately wade through it. AI—if it can hold large data sets in something like a Bayesian network—might help. Not because it “thinks” like a human, but because it can extract massive numbers of correlations from large data sets and present them to us.

Then it becomes our job to interpret them: are they meaningful or just happenstance? Are they trivial, or do they reflect something real? That is where the human work still has to happen.

In addition to not being able to process huge data sets, we also struggle with very basic representational limits. Take something simple, like imagining a grid of numbers. Most people—unless they deliberately train this skill—cannot hold even a four-by-four grid in their mind all at once. Sixteen numbers, each in its own square, is already too much for most people.

Sure, if you make it your specialty, you can get very good at it. You can flash a 4×4 grid for a second or two and memorize the whole thing. But most people cannot do that. Many would struggle even with a 2×2 grid.

AI systems, by contrast, can hold all of that in what you might call an analytic workspace and pull enormous numbers of correlations out of it. We simply lack the capacity—“bandwidth” is not quite the right word; “active working storage” is closer. We cannot keep a full grid of data in mind, let alone a multi-dimensional data set, and then intuitively extract all the correlations an AI system can find.

You can extend that limitation beyond mathematics and statistics. Think about what Daniel Kahneman described as associative thinking—how ideas trigger other ideas automatically, and how the range of associations available to you shapes your understanding. He did not use the phrase “cognitive horizon” formally, but the concept is implicit in his work: the breadth of concepts, experiences, and analogies you can draw on when making sense of something.

I was pleased with an analogy I used this morning on one of my People Yell at People shows. People were discussing what happens when Trump is no longer president—how the world will treat the United States once there is a non-asshole in the Oval Office again. I said the world will be ecstatic, much like when Obama replaced Bush and was handed a Nobel Prize for essentially nothing.

But there will still be a lingering loss of trust, because America can go bad depending on who wins elections. In that sense, America is like an alcoholic: a great person when sober, but with a known tendency to fall off the wagon. That history creates permanent caution and trepidation. I thought it was a decent analogy.

AI, once it really gets its legs under it, will be able to generate far more extreme and wide-ranging analogies, pulling from all over culture. Bill Simmons figured this out early. He realized you could use analogies from outside sports to enrich sports commentary, because sports fans are not interested only in sports. He built an empire on that insight. AI will be able to do that at scale, across domains, at a speed and breadth no human can match.

Jacobsen: What else do you think humans cannot do?

Rosner: I do not know. An AI tool, if used properly, might help people avoid cognitive ruts—the habitual ways of framing the world that we grow too attached to. Take the Big Bang theory, for example. You could ask an AI to generate a dozen alternative models and outline what kinds of experimental, observational, or mathematical evidence would be required to support them.

I do not think AI is currently capable of doing deep, original theoretical mathematics at a very high level, but it will get better.

There are theories about anomalous galactic motion that most scientists currently explain using dark-matter halos around galaxies. There are also alternative approaches—most famously modified gravity models—that propose departures from the inverse-square law under certain conditions. Some of these ideas adjust how gravity scales with distance, though not typically with a clean exponent like 1/r¹·⁹⁷; that number is illustrative, not canonical.

I have a half-formed idea I have never adequately explored: that space itself might “count for less” where there is less matter. As you move toward the outskirts of a galaxy, it is as if the effective geometry or scaling of space changes slightly. I am sure there are many problems with that idea, including orbital dynamics.

Still, you could give hints like that to a future AI, and it could explore the space seriously—generate models, test implications, and suggest refinements that incorporate that intuition.

We know the large-scale structure of the universe contains enormous filaments—hundreds of millions to billions of light-years long—along which galaxies are strung together. If space or gravity behaved differently outside those filaments, that might help account for some observed phenomena. Anyway, I was going to talk about figures for Kitten Kicked Off soon anyway. Best material. I will get you to subscribe to it. I also need to work on the audio clips and get those scheduled.

I think I got the wrong Sinclair earlier. I said Sinclair Lewis, but it was actually Upton Sinclair who said, “It is difficult to get a man to understand something when his salary depends on his not understanding it.” Yes. I meant to correct that—I had the wrong Sinclair.

Jacobsen: All right, let us call it a day. I will see you tomorrow. Thank you.

Rosner: Thank you. Bye.

Rick Rosner is an accomplished television writer with credits on shows like Jimmy Kimmel Live!, Crank Yankers, and The Man Show. Over his career, he has earned multiple Writers Guild Award nominations—winning one—and an Emmy nomination. Rosner holds a broad academic background, graduating with the equivalent of eight majors. Based in Los Angeles, he continues to write and develop ideas while spending time with his wife, daughter, and two dogs.

Scott Douglas Jacobsen is the publisher of In-Sight Publishing (ISBN: 978-1-0692343) and Editor-in-Chief of In-Sight: Interviews (ISSN: 2369-6885). He writes for The Good Men Project, International Policy Digest (ISSN: 2332–9416), The Humanist (Print: ISSN 0018-7399; Online: ISSN 2163-3576), Basic Income Earth Network (UK Registered Charity 1177066), A Further Inquiry, and other media. He is a member in good standing of numerous media organizations.