The Exchanges

Every argument clarity score on this site is built from rows on this page. Each question and answer was assessed with names hidden, the host's own answers included, on four things from 1 to 5: directness (does it answer the question asked), coherence (do the ideas follow), precision (concrete details and clear references), compression (says a lot per word). The weighted mix (30/30/25/15) is the exchange score. A person's published score averages their exchange scores on raw tape only, at least 8 of them, shrunk toward the cohort mean. Full method →

Gary Marcus no published score: only 1 usable exchange on raw tape, and a fair score needs 8+ record → ← everyone

Every exchange below was scored with names hidden, four dimensions each from 1 to 5. An exchange's score is 0.30·directness + 0.30·coherence + 0.25·precision + 0.15·compression. The published score averages the raw tape exchange scores and shrinks small samples toward the cohort mean, so five great answers can't beat twenty good ones. Produced feed rows count only toward coarse estimates, never toward a full score.

clear all ✕
1exchanges match
1on raw tape
0redirected or not addressed
Answered raw tape D 5 · C 5 · P 4 · Cm 4 4.60

Q as long as it achieves its end goal? I mean, you know, I think the argument is, um, you know, planes for air travel are prime example of just brute force. There's nothing that they do that's close to what a bird innately knows how to do, yet it works amazingly well, in fact, you know, much better than a bird. So does it, does it matter how it's done?

A It only depends, I mean, it matters depending on what your problem is. So, If your problem is a closed problem, like Go or chess, where the rules haven't changed, there's a limited number of options, then you can use brute force. I mean, you can either use the kind of brute force that, um, you know, Deep Blue used and just searched through a lot of positions, or there's a more nuanced version of, of brute force in AlphaGo, um, which combines Monte Carlo Tree Search, which is kind of classical AI technique, um, with Deep Learning to do the pattern recognition, um, but still uses an awful lot of data there, and so it's still, as you say, it's an extension. Um, of brute force. But it's an empirical question in each domain about whether that will get you far enough, and it mostly depends on how open-ended the problem is and how many outlier cases there are. So you already, I think, get into trouble with driving. So, you know, people thought that by now we would have driverless cars. If you go back to, like, 2012, um, I wrote a piece in The New Yorker, and I thought that it would, you know, they were legal in three states, and I wrote this thing about, um, what would happen if a school bus went out of control, and I thought, you know, um, You know, do you save the, the children in the school bus, or do you save yourself? I, I thought that, you know, by 2020 maybe we would have drive…

AI assessment note: “it matters depending on what your problem is.”

page 1
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 400 conversations transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.