Sveriges mest populära poddar
Start / LessWrong (30+ Karma) / Gpt 5s are alive synthesis by zvi

LessWrong (30+ Karma)

“GPT-5s Are Alive: Synthesis” by Zvi

66 min • 13 augusti 2025

What do I ultimately make of all the new versions of GPT-5?

The practical offerings and how they interact continues to change by the day. I expect more to come. It will take a while for things to settle down.

I’ll start with the central takeaways and how I select models right now, then go through the type and various questions in detail.

Table of Contents

  1. Central Takeaways.
  2. Choose Your Fighter.
  3. Official Hype.
  4. Chart Crime.
  5. Model Crime.
  6. Future Plans For OpenAI's Compute.
  7. Rate Limitations.
  8. The Routing Options Expand.
  9. System Prompt.
  10. On Writing.
  11. Leading The Witness.
  12. Hallucinations Are Down.
  13. Best Of All Possible Worlds?.
  14. Timelines.
  15. Sycophancy Will Continue Because It Improves Morale.
  16. Gaslighting Will Continue.
  17. Going Pro.
  18. Going Forward.

Central Takeaways

My central takes [...]

---

Outline:

(00:33) Central Takeaways

(02:55) Choose Your Fighter

(05:43) Official Hype

(21:25) Chart Crime

(27:19) Model Crime

(28:09) Future Plans For OpenAI's Compute

(30:49) Rate Limitations

(32:19) The Routing Options Expand

(33:57) System Prompt

(35:14) On Writing

(41:01) Leading The Witness

(41:59) Hallucinations Are Down

(43:17) Best Of All Possible Worlds?

(47:53) Timelines

(58:25) Sycophancy Will Continue Because It Improves Morale

(01:00:22) Gaslighting Will Continue

(01:01:01) Going Pro

(01:04:09) Going Forward

---

First published:
August 13th, 2025

Source:
https://www.lesswrong.com/posts/4wYKkbkHooeQ4xznf/gpt-5s-are-alive-synthesis

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Screenshot of a ChatGPT conversation about declining bread quality in England.
Text explanation of Bernoulli Effect and airplane wing shape principles.
Flow diagram showing GPT model evolution from GPT-2 through GPT-5-thinking, with magnitude indicators.
Bar graph titled
NASA diagram showing incorrect
ChatGPT-5 model selection menu showing different response speed options and legacy models.
A diverse team in a conference room with laptops and screens behind them.
ChatGPT conversation about bread quality improvements in England, with follow-up question.
Flow diagram comparing two approaches to handling AI developments and their outcomes. The top path (in red) shows a simple, reactive adjustment. The bottom path (in green) shows a more thoughtful approach with multiple learning outcomes, including refined perspectives on intelligence, data importance, and AI timelines.
Zvi Mowshowitz tweets:
Steve Ruiz tweets:
Mathias Kirk Bonde tweets:
Sam Altman tweets:
Simple cartoon eye icon with blue iris and white highlight.

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Senaste avsnitt

[Linkpost] “In defense of the amyloid hypothesis” by dsj

15 augusti | 1 min

“Training a Reward Hacker Despite Perfect Labels” by ariana_azarbal, vgillioz, TurnTrout

15 augusti | 13 min

“Somebody invented a better bookmark” by Alex_Altair

14 augusti | 4 min

[Linkpost] “METR Research Update: Algorithmic vs. Holistic Evaluation” by David Rein

14 augusti | 1 min

“Should you make stone tools?” by Alex_Altair

14 augusti | 6 min
Vad är en podd?

En liten tjänst av I'm With Friends. Finns även på engelska.

Podcastbild

00:00 -00:00
00:00 -00:00