schristoph.online

AI ยท 2 min read

๐—œ๐˜€ ๐—œ๐—ป๐˜๐—ฒ๐—น๐—น๐—ถ๐—ด๐—ฒ๐—ป๐—ฐ๐—ฒ ๐—๐˜‚๐˜€๐˜ ๐—–๐—ผ๐—บ๐—ฝ๐—ฟ๐—ฒ๐˜€๐˜€๐—ถ๐—ผ๐—ป? ๐—”๐—ป๐—ฑ ๐—ช๐—ต๐—ฎ๐˜ ๐—”๐—ฏ๐—ผ๐˜‚๐˜ ๐—˜๐˜…๐—ฝ๐—ฒ๐—ฟ๐—ถ๐—ฒ๐—ป๐—ฐ๐—ฒ?

๐—œ๐˜€ ๐—œ๐—ป๐˜๐—ฒ๐—น๐—น๐—ถ๐—ด๐—ฒ๐—ป๐—ฐ๐—ฒ ๐—๐˜‚๐˜€๐˜ ๐—–๐—ผ๐—บ๐—ฝ๐—ฟ๐—ฒ๐˜€๐˜€๐—ถ๐—ผ๐—ป? ๐—”๐—ป๐—ฑ ๐—ช๐—ต๐—ฎ๐˜ ๐—”๐—ฏ๐—ผ๐˜‚๐˜ ๐—˜๐˜…๐—ฝ๐—ฒ๐—ฟ๐—ถ๐—ฒ๐—ป๐—ฐ๐—ฒ?

I was listening to Machine Learning Street Talk (MLST) “The Real Reason Huge AI Models Actually Work” [1]. Many interesting take-aways, a lot to still be digested by me. But really this stuck to my mind.

In AI research, a fascinating concept has emerged: โ€œ๐—–๐—ผ๐—บ๐—ฝ๐—ฟ๐—ฒ๐˜€๐˜€๐—ถ๐—ผ๐—ป ๐—ถ๐˜€ ๐—œ๐—ป๐˜๐—ฒ๐—น๐—น๐—ถ๐—ด๐—ฒ๐—ป๐—ฐ๐—ฒ.โ€ The idea โ€” one theoretical lens, not a settled definition โ€” is that intelligence can be measured by how well a system compresses informationโ€”finding patterns and structures to represent data in fewer bits. On this view, the ability to simplify and predict effectively is taken as evidence of understanding and problem-solving skill.

On the other hand, Andy Jassy โ€” then CEO of AWS โ€” famously said [2], โ€œ๐—ง๐—ต๐—ฒ๐—ฟ๐—ฒ ๐—ถ๐˜€ ๐—ป๐—ผ ๐—ฐ๐—ผ๐—บ๐—ฝ๐—ฟ๐—ฒ๐˜€๐˜€๐—ถ๐—ผ๐—ป ๐—ฎ๐—น๐—ด๐—ผ๐—ฟ๐—ถ๐˜๐—ต๐—บ ๐—ณ๐—ผ๐—ฟ ๐—ฒ๐˜…๐—ฝ๐—ฒ๐—ฟ๐—ถ๐—ฒ๐—ป๐—ฐ๐—ฒ.โ€ This reminds us that while data and knowledge can be compressed into insights, acquiring real-world experience is a different thing: it takes time, trial, and the complex nuances of practice. You can compress the record of experience, not the living of it.

๐—ข๐—ป ๐˜๐—ต๐—ฒ ๐—ณ๐—ถ๐—ฟ๐˜€๐˜ ๐—ด๐—น๐—ฎ๐—ป๐—ฐ๐—ฒ, ๐˜๐—ต๐—ถ๐˜€ ๐˜€๐—ฒ๐—ฒ๐—บ๐˜€ ๐˜๐—ผ ๐—ฏ๐—ฒ ๐—ฎ ๐—ฐ๐—ผ๐—ป๐—ณ๐—น๐—ถ๐—ฐ๐˜ - ๐—ถ๐˜€๐—ป’๐˜ ๐—ถ๐˜?

These ideas sit intriguingly side by side: ย ย โ€ขย ย Compression reflects intelligence as measured in models: finding structure that represents data in fewer bits. ย ย โ€ขย ย Experience points to what that measure leaves out: the time and interaction it takes to acquire real-world skill.

๐—›๐—ผ๐˜„ ๐—ฑ๐—ผ ๐˜†๐—ผ๐˜‚ ๐˜€๐—ฒ๐—ฒ ๐˜๐—ต๐—ฒ๐˜€๐—ฒ ๐˜๐˜„๐—ผ ๐—ฐ๐—ผ๐—ป๐—ฐ๐—ฒ๐—ฝ๐˜๐˜€ ๐—ถ๐—ป๐˜๐—ฒ๐—ฟ๐—ฎ๐—ฐ๐˜๐—ถ๐—ป๐—ด ๐—ถ๐—ป ๐˜†๐—ผ๐˜‚๐—ฟ ๐˜„๐—ผ๐—ฟ๐—ธ ๐—ผ๐—ฟ ๐—น๐—ฒ๐—ฎ๐—ฟ๐—ป๐—ถ๐—ป๐—ด?ย  ๐—–๐—ฎ๐—ป ๐—ฒ๐˜…๐—ฝ๐—ฒ๐—ฟ๐—ถ๐—ฒ๐—ป๐—ฐ๐—ฒ ๐—ถ๐˜๐˜€๐—ฒ๐—น๐—ณ ๐—ฒ๐˜ƒ๐—ฒ๐—ฟ ๐—ฏ๐—ฒ ๐—บ๐—ฒ๐—ฎ๐—ป๐—ถ๐—ป๐—ด๐—ณ๐˜‚๐—น๐—น๐˜† ๐—ฐ๐—ผ๐—บ๐—ฝ๐—ฟ๐—ฒ๐˜€๐˜€๐—ฒ๐—ฑ ๐—ฏ๐˜† ๐—”๐—œ ๐—ผ๐—ฟ ๐—ฝ๐—ฟ๐—ผ๐—ฐ๐—ฒ๐˜€๐˜€๐—ฒ๐˜€?ย  ๐—ข๐—ฟ ๐—ถ๐˜€ ๐—ถ๐˜ ๐—ณ๐—ผ๐—ฟ๐—ฒ๐˜ƒ๐—ฒ๐—ฟ ๐—ฑ๐—ถ๐˜€๐˜๐—ถ๐—ป๐—ฐ๐˜ ๐—ณ๐—ฟ๐—ผ๐—บ ๐—ฝ๐˜‚๐—ฟ๐—ฒ ๐—ถ๐—ป๐˜๐—ฒ๐—น๐—น๐—ถ๐—ด๐—ฒ๐—ป๐—ฐ๐—ฒ?

๐—ช๐—ผ๐˜‚๐—น๐—ฑ ๐—น๐—ผ๐˜ƒ๐—ฒ ๐˜๐—ผ ๐—ต๐—ฒ๐—ฎ๐—ฟ ๐˜†๐—ผ๐˜‚๐—ฟ ๐˜๐—ต๐—ผ๐˜‚๐—ด๐—ต๐˜๐˜€!

#AI #Intelligence #Experience #Learning #MachineLearning #Leadership

[1] Machine Learning Street Talk, “The Real Reason Huge AI Models Actually Work” (with Prof. Andrew Wilson), Sep 19, 2025 - https://www.youtube.com/watch?v=M-jTeBCEGHc

[2] CNBC, “AWS CEO Andy Jassy: No compression algorithm for experience,” Nov 29, 2017 - https://www.cnbc.com/video/2017/11/29/aws-ceo-andy-jassy-no-compression-algorithm-for-experience.html


๐Ÿ“ Last updated: August 13, 2026 โ€” Technical corrections from a quality audit

AIInnovation