NotebookLM is quite powerful and worth playing with

1. The speakers keep cutting each other off mid-sentence, completing each other's sentences. This is cute if it happens once in a while, but it's happening so often as to be extremely annoying.

2. It made a 20 minute file, the last 5 minutes of which were essentially a waste, a repetition.

3. There is excessive amounts of filler material. The speakers are not real people, and they don't need to act as such. The listener's time should be valued better.

Which means: LLMS are evolving faster than we are inventing ways of incorporating them into products. Thus all real-world impacts of LLMs lag behind their current state and capabilities by a margin that is possibly even getting larger with time.

NotebookLM is quite powerful and worth playing with https://notebooklm.google

It is a bit of a re-imagination of the UIUX of working with LLMs organized around a collection of sources you upload and then refer to with queries, seeing results alongside and with citations.

But the current most new/impressive feature (that is surprisingly hidden almost as an afterthought) is the ability to generate a 2-person podcast episode based on any content you upload. For example someone took my "bitcoin from scratch" post from a long time ago: https://karpathy.github.io/2021/06/21/blockchain/ and converted it to podcast, quite impressive: https://notebooklm.google.com/notebook/ba017fec-7068-4085-97...

You can podcastify anything. I give it train_gpt2.c (C code that trains GPT-2): https://github.com/karpathy/llm.c/blob/master/train_gpt2.c and made a podcast about that: https://notebooklm.google.com/notebook/2585c187-b059-475a-b4... I don't know if I'd exactly agree with the framing of the conversation and the emphasis or the descriptions of layernorm and matmul etc but there's hints of greatness here and in any case it's highly entertaining.

Imo LLM capability (IQ, but also memory (context length), multimodal, etc.) is getting way ahead of the UIUX of packaging it into products. Think Code Interpreter, Claude Artifacts, Cursor/Replit, NotebookLM, etc. I expect (and look forward to) a lot more and different paradigms of interaction than just chat.

That's what I think is ultimately so compelling about the 2-person podcast format as a UIUX exploration. It lifts two major "barriers to enjoyment" of LLMs. 1 Chat is hard. You don't know what to say or ask. In the 2-person podcast format, the question asking is also delegated to an AI so you get a lot more chill experience instead of being a synchronous constraint in the generating process. 2 Reading is hard and it's much easier to just lean back and listen.

Strictly speaking, two-voice conversations are slightly overrated. A single voice, akin to a lecture rather than a discussion, is no worse, and is quite sufficient for conveying information. Either approach does the job.

The downloaded audio file extension is wav, implying that it's not compressed, but this is a minor quirk.

Being able to talk to the books you're trying to learn, have them test you for knowledge, being able to combine many resources all at once, just so powerful.

It's like Perplexity for your personal docs.