Introduction

Re-recording and agenda

bit different to hello everybody this is a bit different to one of our usual recordings because this is a re-recording because the last one didn't quite go through today we're going to be going through notebook lm's mind map feature which is one of the coolest features they have to date before that i'm just going to show what we're going to be putting into the notebook so over the last seven weeks now i've been posting partial book reviews

Context: The six-book review project

of three books that claim that AI is overhyped, and then three books that claim that the hype is real, to try and match debates and paint a picture of how much AI could realistically be affecting our lives.

Books and themes being compared

So that's the four on End of Conflict by Simon Horton. This is the one on the four on AI41 by Kai-Fu Lee.

Setting up the Notebook

Creating a fresh workspace

And what I've done is I've created a little notebook here. You can see it all done fresh.

Get rid of this for sex.

Compiling posts into a single file

And I've compiled all those that all those posts into a dot text file, which can be make easier reading. You could do this in a PDF as well, but this is just a efficient way to do it.

Uploading and launching the Mind Map feature

And now I'm going to wait till it's uploaded and click on the mind map option here to a new feature notebook on. And what it's going to do is it's going to compile.

Exploring the Mind Map

Interactive nodes and speed

all of that information and then present it in an interactive mind map where you click on one little text cloud in the mind map and it will split into other related details we still have the audio overview which will create a podcast that you can interrupt and we also have the report look how fast that was you know i don't want to say that was 15 seconds i had to ad lib but let's have a look at this now

Top-level structure: Six books

So it's immediately splits into the six books that are in the document.

Drilling down: Co-Intelligence example

And let's have a look at Co-Intelligence by Ethan Mollock. It's once again split into the four reviews and it's now gonna split into the four parts of the reviews.

Controversial topic: Anthropomorphizing LMs

So let's look at, let's look at the Anthropomizing LMs, a particularly controversial take in AI ethics space.

If I wanted to,

Why think of LMs as human-like?

know why lms are best thought of as similar to a human i've got my nice list here and see how organized that is you can just imagine like if you've got a big pile of unstructured notes this is a really simple fast interactive way to then go and structure your notes which is just really, really cool.

Human-like traits highlighted

So if I want to talk about human-like traits of LMs, they can be, obviously creative is a controversial one, but certainly in brainstorming is the second most common use case of LMs now. Witty is a big one.

Persuasion and de-radicalization research

There's papers coming out now about the persuasive capability of them. I know Ethan Moore published a paper a while ago about how they can de-radicalize people who believe in conspiracy theories.

It's an interesting one. And yeah, they're responsive to lots of different things.

Why discuss anthropomorphism

Um, and there's a question of why discuss this simple reason narrative.

Chat integration with document context

1If you click on this long enough, it will automatically treat this as a prompt in the chats. And this chat bots is trained on this document as well.

And standard top notebook climb stuff.

Citations and source grounding

It sources its answers to the relevant part.

Narratives vs. objects: framing LMs

Um, the libertines, if they're conscious persons by acknowledging they're like conscious emotions or sense of all physical sensations.

It's narrative, it's easier for human beings to construct stories about beings rather than things. But there's a second reason.

Concept: Infinitely Fast Interns

Infinitely Fast Interns, it's a spin on Benedict Evans' Infinite Interns.

Quotations and Commentary

Distinguishing commentary from source quotes

So this bit, the simplest way to understand the vulnerabilities, this is actually part of my commentary on the post.

But then this is the quote from the post.

Vulnerabilities: Emotional manipulation

They respond to emotional manipulation. So if you attempt to bribe an LLM with $200 or threaten them with something, that can make a big difference too.

Conclusion

Takeaways and impressions

I think that's very, very cool.

Good content.