Live data from Hacker News

Google Illuminate: Books and papers turned into audio

illuminate.google.com

191–200 of 251 posts

Re: Google Illuminate: Books and papers turned into audio

#191

Earlier quoted context omitted.

Sure, but now now I – an idiot – can publish a podcast on... "Bayesian Multilevel Models," and fool almost everyone into thinking I know anything about it. I've seen YouTubers provide tutorials on auto-creating YouTube videos and podcast episodes on niche scientific subjects, on how to build seemingly-reputable brands with zero ongoing effort. That is all totally novel. Being able to lie or be wrong before is orthogo…

All the more reason to empower people to review, rate, comment on, block, downvote, and otherwise signal when something is incorrect.

You realize it's a feedback loop, don't you?

If the people interacting are not reliable, then it means the system is not reliable. Karma points, youtube views, thumbs ups, likes... none of those things have any significant value as an indicator of correctedness.

Re: Google Illuminate: Books and papers turned into audio

#192
By now, we can find thousands of hours of discussions online about popular papers such as "Attention is All You Need". It should be possible to generate something similar without using the paper as a source -- and I suspect that's what the AI does.

In other words: I suspect that the output is heavily derivative from online discussions, and not based on the papers.

Of course, the real proof would be to see the output for entirely new papers.

Re: Google Illuminate: Books and papers turned into audio

#193
post #12

One problem I see with this is legitimizing LLM-extracted content as canon. The realistic human speech masks the fact that the LLM might be hallucinating or highlighting the wrong parts of a book/paper as important.

We can find thousands of hours of discussions about popular papers such as "Attention is All You Need". It should be possible to generate something similar without using the paper as a source -- and I suspect that's what the AI is doing here.

In other words: it's not summarising the paper in a clever way, it is summarising all the discussions that have been made about it.

Re: Google Illuminate: Books and papers turned into audio

#194

By now, we can find thousands of hours of discussions online about popular papers such as "Attention is All You Need". It should be possible to generate something similar without using the paper as a source -- and I suspect that's what the AI does. In other words: I suspect that the output is heavily derivative from online discussions, and not based on the papers. Of course, the real proof would be to see the output…

There are much newer papers shown than "Attention is All You Need" (all of them?) and much less talked about (probably all of them, too).

It shouldn't be surprising that a LLM is able to understand a paper, just upload one to Claude 3.5 Sonnet.

Re: Google Illuminate: Books and papers turned into audio

#195
post #111
post #73

Earlier quoted context omitted.

I would be interested in seeing an AI developed to listen to auto-generated podcasts, removing humans from the equation altogether.

Of course the whole point would be in adding an acoustic side channel imperceptible to humans but affecting the listening AI in interesting ways.

dead internet theory kicks in

Re: Google Illuminate: Books and papers turned into audio

#196

Earlier quoted context omitted.

Uncanny, maybe ;-) It should be easy to ingest an RSS feed into your personal RAG system's vector database. Then set up speech to text and text to speech for your PrivateGPT so you can then ask it to create a podcast just for you

There is no truly great open source TTS system available, as far as I know. Most of them are rather scrappy.

How do you define great?

Re: Google Illuminate: Books and papers turned into audio

#197
post #12

One problem I see with this is legitimizing LLM-extracted content as canon. The realistic human speech masks the fact that the LLM might be hallucinating or highlighting the wrong parts of a book/paper as important.

Happens in the very first example: [Attention is All You Need - 1:07] > Voice A: How did the "Attention is All You Need" paper address this sequential processing bottleneck of RNNs? > Voice B: So, instead of going step-by-step like RNNs, they introduced a model called the Transformer - hence the title. What title? The paper is entitled "Attention is All You Need". People are fooling themselves. These are stochastic p…

You left this out

"The transformer processes the entire sequence all at once by using something called self attention"

Re: Google Illuminate: Books and papers turned into audio

#198
post #12

One problem I see with this is legitimizing LLM-extracted content as canon. The realistic human speech masks the fact that the LLM might be hallucinating or highlighting the wrong parts of a book/paper as important.

Happens in the very first example: [Attention is All You Need - 1:07] > Voice A: How did the "Attention is All You Need" paper address this sequential processing bottleneck of RNNs? > Voice B: So, instead of going step-by-step like RNNs, they introduced a model called the Transformer - hence the title. What title? The paper is entitled "Attention is All You Need". People are fooling themselves. These are stochastic p…

Noticed this as well. But on second thought: That's how humans talk - far from perfect. :)

Re: Google Illuminate: Books and papers turned into audio

#199
While this is very nice what I need is my computer to take voice commands, read content in various formats and structure, and take dictation for all of my apps. I need this in my phone too. I can do this now but I have to use a bunch of different tools that don't work seamless together. I need the Voice and Conversational User Interface that is built into the operating system.
Post reply on HN