dinobones

11 ms·

91.
by dinobones 2y ago
I too share the dream of transpiling everything/WebAssembly/everything can run everywhere, but we're pretty far. For a lot of ML/AI workloads and tasks, Python is just a binding for underlying C/C++. It's alrea
92.
by dinobones 2y ago
IANAL, but https://en.wikipedia.org/wiki/Fair_use is a thing. Not sure if it applies in this case though.
93.
by dinobones 2y ago
co-occurrence. It's the real backbone of almost all recommender systems. This is the same way YT/TikTok does it btw. Co-occurrence is king in recommender systems in production. It's extremely cheap to calculate and by far the
94.
by dinobones 2y ago
This question becomes easy if you think about it algebraically/mathematically, but comically hard thinking about it intuitively, as in, trying to reason with language. bat=ball+$1 (ball+$1)+ball=$1.10 2ball + $1 = $1.10 2ball = $0.10 b
95.
by dinobones 2y ago
The Hermes fine-tune for 8b is nearly approaching GPT3.5 Turbo on HellaSwag/MMLU. https://context.ai/model/gpt-3-5-turbo Really exciting times.
96.
by dinobones 2y ago
Do paramedics/ambulances shard across hospitals? It looks like there’s 3 hospitals within 15 minutes of the Las Vegas strip, I’m curious if there’s any attempt to allocate patients equally so that no single hospital becomes overwheleme
97.
by dinobones 2y ago
Not a fan of Apple's policy, but I am a fan of consistency. It was unfair that Patreon was able to operate on the App Store without paying this 30% cut, it basically gave them a monopoly that no similar Patreon-like platform could comp
98.
by dinobones 2y ago
To control for this I personally like comparing Murica North vs Murica South. (USA vs AUS). 3,304 vs 750
99.
by dinobones 2y ago
I've tried Gemini 1.5 flash, and it's about as useless (or useful, if you are a glass half full) as llama 3.1 8b running locally. While these tokens are cheap, they're not very smart. Too many hallucinations. I struggle to i
100.
by dinobones 2y ago
I wonder if the key behind the quality of the MidJourney models, and this models, is less about size + architecture and more about the quality of images trained on. It looks like this is the case for LLMs, that the training quality of the d
101.
by dinobones 2y ago
So there are 2 things that may not make this no longer so surprising. 1) The Maduro government is more like a large gang that is holding a population hostage, than a government. All major businesses/imports/exports are owned by pe
102.
by dinobones 2y ago
Here's this re-explained with a simpler example. Imagine you have 1,000 votes. You want to show that your political party got 60% of the vote, so, you claim: My party: 600 votes Opposition: 300 votes Other: 100 votes Presto, we got a g
103.
by dinobones 2y ago
Thanks for the help, actually it was Bastion. https://supergiant-games.itch.io/bastion The HTML5 demo I was thinking of: https://bastion.fandom.com/wiki/User_blog:Soldierscuzzy/Bast...
104.
by dinobones 2y ago
I think I may be having a Mandela effect moment. For me, Braid was most prominently known as like a cool HTML5 tech demo. I remember playing it at like 15fps in Google Chrome... And that's all it was. But now I look at wiki, and appare
105.
by dinobones 2y ago
Is there a path forward for compilers to eek out these optimization gains eventually? Is there even a path? 550x gains with some C ++ / mixed gnarly low level assembly vs standard C++ is pretty shocking to me.
106.
by dinobones 2y ago
I'm hoping someone takes my bait and explains why MCTS + LLM is more impressive than what it sounds like.
107.
by dinobones 2y ago
Wow, did you really have access to IDA as a kid? Even with adult money it seems expensive to me.
108.
by dinobones 2y ago
I see DeepMind is still playing around with RL + search algorithms, except now it looks like they're using an LLM to generate state candidates. I don't really find that this impressive. With enough compute you could just do n-of-1
109.
Ask HN: What Is the SQLite of Queues?
24 points by dinobones 2y ago|14 comments
110.
by dinobones 2y ago
Lol more like, Perplexity has a terminal diagnosis. Doing RAG using like.. prompt hacking and text embeddings + vector store when you have no access to the underlying model, nor ability to fine tune the generation for RAG, will fail. It wil
111.
by dinobones 2y ago
Apparently they’re used by 40% of the Fortune 100. At a generous 20xARR valuation this is still 1b ARR. 1 billion / 40 = annual revenue of $25M from each contract. Hwat. Help it make sense.
112.
by dinobones 2y ago
We rolled our own workflow engine and it almost crashed one of our unrelated projects for having so many bugs and being so inflexible. I’m starting to think workflow engines are somewhat of a design smell. It’s enticing to think you can bui
113.
AI paid for by Ads – the gpt4o-mini inflection point(batchmon.com)
2 points by dinobones 2y ago|0 comments
114.
by dinobones 2y ago
Which 2020-2024 policies have harmed the economy?
115.
by dinobones 2y ago
No my friend, the people who ensure the paycheck has the specific number it does are some drones in HR far removed from you. They determine reasonable bands for your role based on industry standards and how the company wants to position its
116.
by dinobones 2y ago
In every company I've worked for, the biggest fans/lovers of meetings/let's "touch base" have always been: The most useless members of our team. And I truly mean that, whether PM/TPM/or even SDM&#x2
117.
by dinobones 2y ago
Yeah exactly this, ChatGPT 4-o very rarely, if ever, hallucinates.
118.
by dinobones 2y ago
AI has personally tutored me about obscure, deep linear algebra concepts. It's so great to get applied examples and be able to ask why/how something works, rather than reading a stuffy Wikipedia article or math textbook. It's
119.
by dinobones 2y ago
Lol wut? HackerNews has absolutely hated Bitcoin since its inception.
120.
by dinobones 2y ago
Uhh.. how about the fact that almost no wallets/exchanges/no one uses it? https://www.reddit.com/r/lightningnetwork/comments/190orn6/c... There are a ton of outstanding technical issues that li

More ›