Comments URL: https://news.ycombinator.com/item?id=42560601

Points: 1

# Comments: 0

I was curious how this was measured since benchmarking accuracy for LLMs is tough. Found this in the paper: "This classification accuracy was benchmarked by manually analyzing over 400 papers across a range of representative searches, and comparing the human evaluation to the language model’s judgment"

I'm skeptical that their dataset of 400 papers with 3 classification labels (highly relevant, closely related, or ignorable) is large enough to represent the diversity of queries they're going to get from users. To be clear, I don't think this undermine's (haha) the value of what they've built, still very cool.

Ok, fair enough. I wouldn't pin any of these issues on "technology". I lay them at the feet of "businesses" specifically operating within a capitalist landscape. Unfortunately "technology" is tightly coupled to "businesses" in the Western world and I don't think that removing constraints on the latter will benefit anyone except, you know... rich people.

"We believe the techno-capital machine of markets and innovation never ends, but instead spirals continuously upward. Comparative advantage increases specialization and trade. Prices fall, freeing up purchasing power, creating demand. Falling prices benefit everyone who buys goods and services, which is to say everyone. Human wants and needs are endless, and entrepreneurs continuously create new goods and services to satisfy those wants and needs, deploying unlimited numbers of people and machines in the process."

Calls for further doubling down on Capitalism given the dire inequality and our inability to meet even basic needs for so many people in our already rich society strikes me as the utmost expression of societal tone deaf.

Comments URL: https://news.ycombinator.com/item?id=35657969

Points: 7

# Comments: 0

I built this example using semantic search + GPT-3 to make files conversational regardless of whether they fit within GPT-3's token limit. Enter a website url or upload a document then just ask for a summarization or any other questions about the document. Behind the scenes the document is parsed and I'm using embeddings to pull out context for GPT-3 when asking your question. Summarization seems to work pretty well but I've found it to really shine when asking specifics about a document that ChatGPT wouldn't be answer on it's own. I'd love any feedback or thoughts on other things that would be interesting to do with this app (Reddit bot?, Chrome extension?). Thanks!

Example Articles:

How Our Reality May Be a Sum of Possible Realities https://app.conifer.chat/threads/68087fd1-a9d8-43ad-a7bb-f41...

Building a smart mirror - Raspberry Pi https://app.conifer.chat/threads/02a0af22-03f0-4fd5-8724-b1d...

President Biden State of the Union Address - CNN https://app.conifer.chat/threads/b24cb5fb-f973-4ba9-b92f-391...

Comments URL: https://news.ycombinator.com/item?id=34885744

Points: 2

# Comments: 6

If you would like to ask it more things about this article I have it searchable here: https://app.conifer.chat/threads/37321fdc-6a04-4c56-a857-21b...

Given the equations that describe quantum mechanics (ie Schrodinger equation) MWI is essentially the "null hypothesis". No equations that describe collapse have ever passed the rigor of experiment and all collapse theories require modifications to the mathematics of QM. The burden of proof here is on theorists that support collapse theories not proponents of MWI.

According to MWI, all possible results of an experiment are manifest and real. You're asking "which particular entangled state we end up in?". You end up in all of them. For a 2-state superposition system, there are 2 versions of "you" that exist after the experiment, both of which are equally real. Subsequent measurements of the quantum system will appear to be "collapsed" for both versions of you but each will see different and opposing values.

The pull towards a more spiritual understanding of my condition has been heavily motivated by the mystery of my own consciousness. By consciousness I refer only to pure awareness. This moment in time with this collection of protons and electrons and complex field fluctuations is being experienced. And I find that so incredibly truly strange an inexplicable by any physical description. To carry out this dance of wave function evolution it does not seem that experience would be required at all. Yet it's there. Furthermore, when I'm quiet and my mind settles I can become acutely in touch with this pureness of experience and it appears to transcend mental processing, thoughts, memories and all of the things that are typically used to describe "I".

My story for what the universe is, is something along the lines of a Maxwell Tegmark level 4 multiverse populated with all possible mathematical structures and infused with a panpsychism quality of fundamental experience. Yes, I suspect that even electrons are experienced and I know that's weird lol. In the end though I know this description is just a story that my ego mind needs to make sense of all this stuff and I try not to place too much stock in it.

Practices that I find helpful and meaningful:

-Meditation (I usually aim for about 30 minutes a day)

-Listening to and reading Alan Watts

-Listening to and reading Ram Dass

-Semi-regular psychedelic experience (psilocybin primarily)

-Listening to East Forest's music