akrymski

8 ms·

31.
by akrymski 2y ago
Why does Amazon even allow random resellers/distributors? Are there retailers that only resell original manufacturers?
32.
by akrymski 2y ago
https://archive.md/R0fKC
33.
Online Shopping Has Become a Giant Fake Product Machine(bloomberg.com)
4 points by akrymski 2y ago|4 comments
34.
by akrymski 2y ago
But GPT4 has been trained up to recent events, so you can't do rolling predictions using just historical data with it. If your LLM knows the future obviously it can predict well, even if the company name is hidden.
35.
by akrymski 3y ago
https://archive.md/djcGF
36.
Visa-Mastercard Deal(bloomberg.com)
9 points by akrymski 3y ago|1 comments
37.
by akrymski 3y ago
Very cool. If I ask to deduce the gender of my voice, can it do that? Training a projection layer makes sense, but ultimately you'd want to output audio conditioned on the input rather than text. Is there a way to train a reverse proj
38.
by akrymski 3y ago
Fantastic. I've always wondered why Postgres doesn't have this built in. Wish it was open source.
39.
by akrymski 3y ago
Imagine reading HN on this thing. Voice chat with your computer is possibly the worst input method. UIs exist for a reason. Clicking a few buttons is much faster than a conversation. The future of computing doesn't look like a bot t
40.
Bitcoin ETFs Approved by SEC(ft.com)
3 points by akrymski 3y ago|0 comments
41.
by akrymski 3y ago
> The other trick I've been planning to explore is using an LLM to turn the user's question into a small number of normal FTS search queries and then run those to try and get context data. I have also been working on this. I s
42.
by akrymski 3y ago
Pretty certain you can never stop humans from hallucinating either. Any human answer will have a probability/confidence level attached to it, it's never zero, because any statement relies on assumptions of the world encoded by the
43.
by akrymski 3y ago
Re candles - even longer term, hourly/daily? Are there actually strategies out there that deliver great sharpe over many years with just time series forecasting? Most hedge funds don't beat the index afaik
44.
by akrymski 3y ago
Does the PDF include illustrations generated by AI?
45.
by akrymski 3y ago
Why would banks adopt FedNow when it will ultimately eat into their revenue from card interchange?
46.
by akrymski 3y ago
I'd suspect so, but would love to see that confirmed.
47.
by akrymski 3y ago
Exactly. This looks like classic extortion: if you don't pay me more, I will stop fighting, or do a 180.
48.
by akrymski 3y ago
This is the only TV show I've rewatched 3 times. To me it captures the essence of our existence for a lot of older people in tech. In many ways it's the modern version of a western where a bunch of guys go out to the middle of n
49.
The Best Show No One Watched(imdb.com)
3 points by akrymski 3y ago|2 comments
50.
by akrymski 3y ago
These NanoGPT based models are great, thank you for contributing to OS. Would love to see this ported to CPUs ala llama.cpp. Any plans in that direction?
51.
by akrymski 3y ago
So is https://github.com/Lightning-AI/lit-llama
52.
Ponzi Scheme That Hooked Warren Buffet(theatlantic.com)
5 points by akrymski 3y ago|0 comments
53.
by akrymski 3y ago
When people say crypto is the answer to cash, I ask them what happens when the internet is down
54.
by akrymski 3y ago
by mass afaik
55.
by akrymski 3y ago
But lithium only makes up 11% of EV batteries https://www.statista.com/statistics/1203083/composition-of-l...
56.
by akrymski 3y ago
Nobody wants to run Google on their PCs, why should LLMs be different? I'd expect GPT models to be updated regularly fairly soon, and in much the same way that I wouldn't want to host a personal out-dated web index + search engin
57.
by akrymski 4y ago
> They aren't making it anymore. But aren't populations of most wealthy countries dropping already? EU population is already declining. We also aren't making efficient use of the land. We could have many more sky scraper
58.
by akrymski 4y ago
I actually still prefer SO as I get to see different solutions to the same problem and related discussion. Co-pilot is great for boilerplate stuff though, which is the majority of lines for most projects these days I'd guess.
59.
by akrymski 4y ago
What LLM is this using? Or did you train your own?
60.
by akrymski 4y ago
https://archive.md/7CmWl

More ›