hackertube.io

Sparks of AGI: early experiments with GPT-4

Sebastien Bubeck · lecture · Watch on YouTube

LLMs & generative AIAI safety, AGI & impactUniversity lectures & courses

What it is

Sebastien Bubeck presents the Sparks of AGI paper at MIT, showing early experiments with GPT-4 across many domains and asking whether artificial intelligence has arrived. It includes the TikZ unicorn example.

Why Hacker News recommends it

HN recommends it as a thought-provoking talk from a researcher with early GPT-4 access, especially for the unicorn degrading under safety tuning.

  • 7 people recommended it
  • 4 different months
  • 4 strong recommendations
  • first 2023, last 2026

What HN commenters said

  1. “I'd highly recommend giving some of these a look”

    Describes what GPT-4 can do that earlier LLMs could not.

    lhl, 2023-04 · Strong recommendation · in “The LLama Effect: Leak Sparked a Series of Open Source Alternatives to ChatGPT”

  2. “There’s a very good video that is worth watching as well that may not change your mind, but will certainly make you think”

    Makes you think beyond next-word prediction about LLMs.

    iamflimflam1, 2023-04 · Strong recommendation · in “What GPT-4 Does Is Less Like “Figuring Out” and More Like “Already Knowing””

  3. “Highly highly highly recommend listen to this lecture by Sebastien Bubeck”

    Lecture by the paper's author, extremely interesting.

    neom, 2023-05 · Strong recommendation · in “ChatGPT Egg Balancing Task Convinced Microsoft That AGI Is Closer”

  4. “Can recommend the talk, one that has stayed in the back of my mind since I first saw it two years ago.”

    Talk that has stayed with the commenter for two years.

    Topfi, 2025-10 · Strong recommendation · in “Gemini 3.0 spotted in the wild through A/B testing”

  5. “he explains the lobotomization process in a particularly poignant way”

    Explains how tuning progressively deadened GPT-4's unicorn drawing.

    ftxbro, 2023-05 · Recommendation · in “The AI revolution is powered by these contractors making $15 an hour”

  6. “are also well worth your time”

    The Sparks of AGI talk is well worth your time.

    pmoriarty, 2023-05 · Recommendation · in “LLMs are not greater than the sum of their parts: Stanford researchers”

  7. “You should take a look at this video.”

    Researcher with private access shows GPT-4 evolution over time.

    nazka, 2023-06 · Recommendation · in “GPT4 Can’t Ace MIT”

Excerpts from Hacker News comments, linked to the originals. Summaries and labels were written by a language model from those comments.

Topics: gpt-4, agi, llm, alignment tax, microsoft research

Also recommended by the same people