---
date: '2024-01-07'
description: intelligence as tool for thought built atop human intelligence rather than replacing it, questioning whether next-token prediction constitutes true intelligence.
id: Intelligence amplification
modified: 2026-06-05 15:08:27 GMT-04:00
tags:
  - seed
  - ml
title: Intelligence amplification
created: '2024-01-07'
published: '2024-01-07'
pageLayout: default
slug: thoughts/Intelligence-amplification
permalink: https://aarnphm.xyz/thoughts/Intelligence-amplification.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
<blockquote class="twitter-tweet" data-lang="fr" data-dnt="true"><p lang="en" dir="ltr">I’m playing around with calling our tech, as it is today, IA (intelligence amplification) instead of AI. IA have the vibe of tools for thought, needing human interaction, and resemble a lot more what we actually have today. AI feels more like independent long-running agents. <a href="https://t.co/vVDH77JI3U">https://t.co/vVDH77JI3U</a></p>&mdash; Andrej Karpathy (@karpathy) <a href="https://x.com/karpathy/status/1744062845426532473?ref_src=twsrc%5Etfw">7 janvier 2024</a></blockquote>



Intelligence should be thought as a tool for thought, not an independent agent

These systems should be built on top of human intelligence, not replace it.

Next-token prediction is primitive to call a system intelligent.

Can a [[thoughts/Transformers|transformers]] ever be [[thoughts/Turing-complete Transformers|Turing-complete]]?

## research area.

A lot of alpha in mechanistic analysis of the [[thoughts/representations|representations]] these models exhibit., or “virtual brain analysis”.