Six Month Pause on AI More Powerful Than GPT-4, ChatGPT Plugins, GPT-4 Development, GPT-4 AGI? | E08

16 snips

Mar 31, 2023

Ask episode

AI Snips

Chapters

Transcript

Episode notes

INSIGHT

Doom Prediction from Eliezer Yudkowsky

AI researcher Eliezer Yudkowsky warns that overly powerful AI could lead to human extinction without current survival plans.
His extreme views emphasize an urgent need to address AI alignment, despite appearing hyperbolic.

INSIGHT

AI Legislation Needs Catalysts

Major AI legislation will likely only follow major AI disasters as governments struggle to understand complex technology.
Current political systems cannot preemptively regulate AI effectively during its rapid development.

INSIGHT

ChatGPT Plugins as New Platform

ChatGPT Plugins transform GPT into an app platform, allowing natural language control over external APIs.
This is a significant shift in the way AI interacts with software, beyond just text generation.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

Your two favorite luddites are back in Episode 8 of "This Week in AI" podcast. In this episode we cover the FutueOfLife.org Open Letter Asking for a Six Month Pause on AI More Powerful Than GPT-4, Talk About Implementing GPT-4 and It's Limitations, Cover Eliezer Yudkowsky's Doom Predictions in a TIME article, Discuss the Implications of ChatGPT Plugins, MisInformation and A Whole Lot More.

If you enjoy this podcast please consider subscribing and leaving a review to help spread the word. Thanks for listening.

00:00 - Powerful AI is GONNA KILL US ALL!?
00:16 - Six Month Pause on AI? FutureOfLife.org Open Letter
08:35 - Will They Take AI Models Away From Us?
12:30 - Implementing GPT-4 and Limitations of GPT-4 Access
18:07 - Eliezer Yudkowsky Says We're All Gonna Die From Powerful AI!
21:05 - How Will They Legislate AI if They Don't Understand TikTok?
22:06 - Is AI Development as Big as Nuclear Weapons?
23:40 - ChatGPT Plugins: One App To Rule Them All?
32:06 - ChatGPT IS The Killer App: Vector Database (Memory) & Embeddings
39:08 - GPT-4 Can Visualize Images From Imagining Them!
42:59 - Misinformation & Midjourney Access Stopped Due to Deep Fakes
44:45 - ChatGPT Plugins Hacked & DAN Plugin
47:06 - Disrupting OpenAI GPT-4 with Open Source Models
49:24 - Our Dear Friend Simon Willison: LLaMA 7B Model
51:46 - Large Language Models LLMs May Destroy Humanity
52:09 - Do We Know How to Fully Utilize GPT-4?
56:00 - Gambling with AI Update
1:00:21 - Sam Altman Lex Friedman Interview
1:03:00 - We Have #FREESYDNEY Mugs to Give Away!

SOURCES:

https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/
https://futureoflife.org/open-letter/pause-giant-ai-experiments/
https://twitter.com/billackman/status/1641531788173385729?s=46&t=uXHUN4Glah4CaV-g2czc6Q
https://openai.com/blog/chatgpt-plugins
https://twitter.com/rez0__/status/1639259413553750021?s=20
https://twitter.com/hwchase17/status/1640171938470563840?s=20
https://www.microsoft.com/en-us/research/publication/sparks-of-artificial-general-intelligence-early-experiments-with-gpt-4/
https://blog.lastmileai.dev/using-openais-retrieval-plugin-with-llama-d2e0b6732f14
https://twitter.com/karpathy/status/1640042620666920960?s=46&t=uXHUN4Glah4CaV-g2czc6Q
https://twitter.com/simonw/status/1636756272631681024?s=46&t=uXHUN4Glah4CaV-g2czc6Q