qumpis

12 ms·

91.
by qumpis 3y ago
Regarding nutrition: maybe meal-replacement shakes, Huel et al.
92.
by qumpis 3y ago
> I'm am fairly convinced that in the majority of the cases it isn't the work load that burns someone out, it is when one put in lots of efforts and realizes it it doesn't matter at all Absolutely. And that's why havi
93.
by qumpis 3y ago
How do you know if you're truly happy? That you're working on something not because you feel the need to prove something, battle your insecurities, but rather because you want to do something meaningful? I'm genuinely curious
94.
by qumpis 3y ago
On what level of topics do you query LLMs to help you learn? As you become a domain expert, the usefulness of LLMs diminishes, I imagine? Regarding the learning of learning from first principles vs from 'working interfaces', could
95.
by qumpis 3y ago
I guess this isn't applicable for technical books, i.e. you don't read them on a kindle?
96.
by qumpis 3y ago
I think it's a great advice. I'd also add that reading a book from cover to cover without a clear application of the knowledge can be wasteful in terms of time and retention, and might lead to feeling motivated to finish the book
97.
by qumpis 3y ago
Maybe you need to keep objects centered when you take pictures :)
98.
by qumpis 3y ago
This looks super cool, but why not use this tool as a non-visual tool to show similar books given a title? As far as I know there aren't many tools for this Also it would be super cool if we could import out goodreads reading lists and
99.
by qumpis 3y ago
I actually meant exactly this. People apply ML to their domains and say thay that this is similar to having a chemist with a PhD.
100.
by qumpis 3y ago
I remember the CEO of google saying that more capable models will be releases a few months (a month?) ago already. Either they delayed or the model is just as bad (maybe they released after the announcement of coding in 20 languages?)
101.
by qumpis 3y ago
In regards to 2), it sounds a bit directionless to be proposing stuff people don't need? Isn't that more of a problem of selecting relevant problems to solve, and getting supervision on your ideas?
102.
by qumpis 3y ago
Yes, uplifting minorities is great, but anyone should be accountable equally when it comes to workplace conduct
103.
by qumpis 3y ago
Wiki doesn't seem to give detail into the situation, nor the paper in question
104.
by qumpis 3y ago
I'm yet to see an ML PhD be required to learn chemistry to a similar extent that chemists would need to doing ML (especially at research level)
105.
by qumpis 3y ago
All of this conditions on having a single goal: career success, or any other singular goal. Feeling the need to hustle non-stop might not be the 'best', whatever that means :)
106.
by qumpis 3y ago
If humans were grown in farms for the purpose of using their meat, then yes, we shouldn't waste them.
107.
by qumpis 3y ago
Nice. Is there a similar UI that allows to go back in chat history and branch from there? Re-entering initial prompts is quite inconvenient
108.
by qumpis 3y ago
That novel reasoning is due to the model transitioning to a generalization regime, not merely parroting training data?
109.
by qumpis 3y ago
Well, then it's a great mystery of up to 10^18: "T. Oliveira e Silva ran a distributed computer search that has verified the conjecture for n ≤ 4 × 10^18"
110.
by qumpis 3y ago
Is this a joke in Russia, i.e. people referring to him as black in order to lessen the view of Russia's outlook to foreigners? I've never heard people refer to Pushkin as being black (and his parents indeed aren't), besides h
111.
by qumpis 3y ago
In my experience 3.5 is so far away from 4 that it makes it makes 3.5 useless. I'd give another shot to gpt 4 if 3.5 was almost working for you (what did you use it for, anything for specific?)
112.
by qumpis 3y ago
Rlhf wasn't introduced by openai. And GPT is a pretty standard transformer, no? Yes they did it at scale and it speaks volumes on their production skills, but OP was asking about research
113.
by qumpis 3y ago
Not sure, maybe there are some savants with photographic memory. But likely people remember references to certain art and can look them up and then 'copy paste' stylistic elements (but with a lot of effort!)
114.
by qumpis 3y ago
Youtubers who rise to the top seem to satisfy their viewers even if there's hard monetary incentive
115.
by qumpis 3y ago
Are there artists who create in isolation? I.e. the ones who somehow can prove that their art is not based on what they've seen?
116.
by qumpis 3y ago
Train only on data that is hard to predict/complete?
117.
by qumpis 3y ago
Oh this explains a lot, thank you for the information!
118.
by qumpis 3y ago
Interesting. I remember when the speedup of chat-gpt happened, the API prices dropped by around 10x, so I'd imagine there were some tricks of making them run faster. If they still haven't implemented these, it would be positively
119.
by qumpis 3y ago
That sounds like having a mixture of experts model (at high scale popularly developed by Google): train multiple specialised models (say embedders from text to a representation) that could be fed into a single model at the end. Each expert
120.
by qumpis 3y ago
(Why) Is that technically correct? I'm really curious since I too thought they meant that capping effect would increase (fewer messages allowed), and not decrease (more messages allowed), as was my intuitive understanding.

More ›