ppsreejith

8 ms·

31.
Is HN under some kind of spam attack?(imgur.com)
7 points by ppsreejith 2y ago|3 comments
32.
by ppsreejith 2y ago
Seems there's a lot of links submitted from new accounts which are dead on submission. This doesn't seem normal. Screenshot above^
33.
Why Meta can be the most valuable company in the world(stratechery.com)
2 points by ppsreejith 2y ago|1 comments
34.
by ppsreejith 2y ago
This is pretty incredible, kudos to the team! I wonder if there's still an option to call native modules asynchronously (since I'd guess the synchronous native calls block JS execution?) Also, I remember transferring lots of data
35.
Demis Hassabis Wins Nobel Prize in Chemistry(techcrunch.com)
10 points by ppsreejith 2y ago|1 comments
36.
by ppsreejith 2y ago
To be pedantic, the comic doesn't mention _targeted_ lightning strikes.
37.
by ppsreejith 2y ago
> If every major APM vendor and dozens of startups release agents in the next year, it will be difficult for customers to tell what’s snake oil or what’s actually useful. One approach, also seen in the financial space, is having open ben
38.
LLM output tokens are upto 200x slower than input tokens(minusx.ai)
3 points by ppsreejith 2y ago|0 comments
39.
Motivations on building a FOSS browser extension(minusx.ai)
3 points by ppsreejith 2y ago|0 comments
40.
by ppsreejith 2y ago
As a Firefox user myself, yes! We plan to launch for other browsers after open sourcing MinusX (in ~1-2 weeks).
41.
by ppsreejith 2y ago
We're currently exploring the tools displayed on our website (Tableau, Grafana, Colab, & Google Sheets). But if you have a specific tool in mind, please do tell us at https://minusx.ai/tool-request
42.
by ppsreejith 2y ago
> How’s the performance on doing the same analysis on a dataset with 1 billion rows for instance? This really depends on whether your tool can handle the scale. We only use a sample of the outputs when constructing the context for your i
43.
by ppsreejith 2y ago
Yep! MinusX uses Metabase APIs to pull relevant tables, schema, & dashboards to construct the context for your instruction. > Anecdotally, my hardest problems w/ nl2sql are finding the right tables and adding the right filters.
44.
by ppsreejith 2y ago
We're building an assistant that works across all your analytics apps. This means MinusX can use context from multiple apps to better fulfil your instructions. You can imagine a future version of MinusX reading data from a spreadsheet,
45.
by ppsreejith 2y ago
Currently, we're using GPT-4o. We've tested it with Claude as well and plan to roll out support soon!
46.
by ppsreejith 2y ago
We've done a bunch of work to strip down the context and minimise the output tokens (which tends to be 100x as slow as input tokens). GPT-4o is pretty fast too :)
47.
Comma.ai: Refactoring for Growth(blog.comma.ai)
133 points by ppsreejith 2y ago|104 comments
48.
by ppsreejith 2y ago
Possibly related to Elon's new lawsuit? https://www.theverge.com/2024/8/5/24213557/elon-musk-openai-... > The new lawsuit filed in federal court in Northern California on Monday says that Altman
49.
by ppsreejith 2y ago
I think the unit is off. Starting from 2410 GWh & a compound increase of 59% per year gives us: 61,915 GWh (2410 * 1.59^7) which is about 61.915 TWh. So perhaps the author meant 61.915 TWh instead of GWh. No way is this in anyway close
50.
by ppsreejith 2y ago
Very interesting prompt regarding the boxer Eupolus of Thessaly. Gemini Advanced gets this wrong as well as Llama 3 70b (as run on groq.com). However, if I start with: "What is the earliest record of cheating in the olympic games?&quot
51.
by ppsreejith 3y ago
From https://www.reuters.com/article/idUSL1N2XV1HA/ > According to the U.S. Geological Survey, published scientific estimates of the global CO2 emissions for all on land and submarine volcanos “lie in a range f
52.
by ppsreejith 3y ago
IIRC Dashtoon studio allows you to create comics with consistent characters using stable diffusion: https://dashtoon.com/create
53.
by ppsreejith 3y ago
Thanks again! Hope I'm not overwhelming but one more question: Are you decoding with batch size = 1 or is it more?
54.
by ppsreejith 3y ago
Thank you, that demo was insane! Follow up (noob) question: Are you using a KV cache? That would significantly increase your memory requirements. Or are you forwarding the whole prompt for each auto-regressive pass?
55.
by ppsreejith 3y ago
Thank you for doing this AMA 1. How many GroqCards are you using to run the Demo? 2. Is there a newer version you're using which has more SRAM (since the one I see online only has 230MB)? Since this seems to be the number that will dri
56.
by ppsreejith 3y ago
Relevant thread from 5 months ago: https://news.ycombinator.com/item?id=37469434 I'm achieving consistent 450+ tokens/sec for Mixtral 8x7b 32k and ~200 tps for Llama 2 70B-4k. As an aside, seeing that this is buil
57.
Zuckerberg reviews the Apple Vision Pro(theverge.com)
3 points by ppsreejith 3y ago|3 comments
58.
Meta's $197B Surge Is Biggest in Stock-Market History(bloomberg.com)
18 points by ppsreejith 3y ago|3 comments
59.
by ppsreejith 3y ago
If you go by what Zuck says, he calls this out in previous earnings reports and interviews[1]. It mainly boils down to 2 things: 1. Similar to other initiatives (mainly opencompute but also PyTorch, React etc), community improvements help t
60.
by ppsreejith 3y ago
> The energy stored in the oil is analogous to the "outside heat" for heat pumps. This isn't correct. Here's a thought experiment to clarify: If we burn X MJ of oil, we transfer X MJ of energy as heat to our house. Ot

More ›