“AI #38: Let’s Make a Deal” by Zvi

Another busy week. GPT-5 starts, Biden and Xi meet and make somewhat of a deal, GPTs get explored, the EU AI Act on the verge of collapse by those trying to kill the part that might protect us, multiple very good podcasts. A highly interesting paper on potential deceptive alignment.

Despite things quieting down the last few days, it is still a lot. Hopefully things can remain quiet for a bit, perhaps I can even get in more work on that Jones Act post.

Table of Contents

1. Introduction.
2. Table of Contents.
3. Language Models Offer Mundane Utility. Structured prompts yay.
4. Language Models Don’t Offer Mundane Utility. Errors in error rates.
5. GPT-4 Real This Time. How do you protect against theft of your GPT?
6. Fun With Image Generation. Dalle-3 reluctant to let people have any fun.
7. Deepfaketown and Botpocalypse Soon. Terrorists ‘exploiting’ [...]

---

Outline:

(00:39) Language Models Offer Mundane Utility

(05:11) Language Models Don’t Offer Mundane Utility

(10:49) GPT-4 Real This Time

(14:11) Fun with Image Generation

(15:12) Deepfaketown and Botpocalypse Soon

(15:59) A Bad Guy With an AI

(22:46) They Took Our Jobs

(32:57) Get Involved

(34:29) Introducing

(35:18) In Other AI News

(42:00) Quiet Speculations

(43:32) Anti Anti Trust

(44:24) The Quest for Sane Regulations

(57:39) Bostrom Goes Unheard

(58:18) The Week in Audio

(59:37) Someone Picked Up the Phone

(01:01:15) Mission Impossible

(01:02:17) Rhetorical Innovation

(01:05:21) Open Source AI is Insafe and Nothing Can Fix This

(01:13:19) Aligning a Smarter Than Human Intelligence is Difficult

(01:36:02) People Are Worried About AI Killing Everyone

(01:37:37) Other People Are Not As Worried About AI Killing Everyone

(01:40:33) The Lighter Side

---

First published:
November 16th, 2023

Source:
https://www.lesswrong.com/posts/oCFX5xbhgCmpBFKnb/ai-38-let-s-make-a-deal

---

Narrated by TYPE III AUDIO.