diwank

10 ms·

151.
by diwank 2y ago
It does on a lot of editors and I hate it. On my vim the highlights only get added/removed when I switch back to normal mode, during editing nothing changes except for the actual keystrokes.
152.
by diwank 2y ago
Yep but to me this feels like a lot of poor ux decisions too. I would def appreciate the visual indicators of italics and bold text but ffsake please dont toggle it every time I add a bloody keystroke. It's so jarring when editors do t
153.
by diwank 2y ago
Yes! I sometimes prefer to use `====` h1s and `----` h2s especially when I am writing slides or taking notes on a live call for others and I hate that prettier and other linters would automatically turn them into `# ` and `## ` ... Please s
154.
Tree Attention: Topology-Aware Decoding for Long-Context(arxiv.org)
79 points by diwank 2y ago|22 comments
155.
Robotaxis can finally take you out of San Francisco(sfstandard.com)
5 points by diwank 2y ago|1 comments
156.
by diwank 2y ago
More on the Omega monad from its author: http://web.archive.org/web/20210616105135/http://lukepalmer....
157.
Transistors: An Unfulfilled Promise(nytimes.com)
6 points by diwank 2y ago|2 comments
158.
by diwank 2y ago
> These included, for instance, a run of voters whose second names were identical except that the first letter was changed, in alphabetical order; and a run of voters whose names were translations of consecutive numbers. This is just hil
159.
by diwank 2y ago
Inverse: “CryptGPT: Privacy-Preserving Language Models Using Vigenere Cipher” https://huggingface.co/blog/diwank/cryptgpt-part1
160.
by diwank 2y ago
The number of tokens per image are actually fairly small, ranging from 85 to ~500.
161.
by diwank 2y ago
Yeah I think their findings are def interesting but the title and the strong claims are a tad hyperbolic.
162.
by diwank 2y ago
I think while the findings are interesting, the title "Vision language models are blind" is hyperbolic and misleading of the findings. It is true that VLMs have a very different way of sensing and working with image inputs and for
163.
'Her' Interface Components(scifiinterfaces.com)
2 points by diwank 2y ago|0 comments
164.
Hottest Place on Earth Lies Somewhere Between Folklore, the Desert and the City(earthobservatory.nasa.gov)
1 points by diwank 2y ago|0 comments
165.
by diwank 2y ago
While I agree that AI infra startups are hard to build, I strongly disagree with the idea that they are harder than foundational or application layer startups. I think it boils down to what you know and what resources you can muster .
166.
by diwank 2y ago
Just to add- not easily defeated. I am even hosting a bounty for whoever can break this one (the model weights and tokenizer are available on huggingface) :D
167.
by diwank 2y ago
Yes of course but this was just to validate the hypothesis that FMs and their tokenizers can generalize on a fully encrypted dataset. Now that the approach is clearly viable, I am going to extend it to use a modern cipher like XChacha20 whi
168.
by diwank 2y ago
Also to clarify- privacy preserving here means protecting from the model inference provider. Sort of like, if openai trained a GPT-4 using this scheme and gave the government the key. Then the gov can use it safely even while it’s hosted on
169.
by diwank 2y ago
It is possible to reconstruct the text from an embedding but also more importantly, I believe that by embedding, the OP means sending the computed embedding matrix instead of the input. Which is a simple matrix inversion problem. Computing
170.
by diwank 2y ago
Yep and embeddings can be decoded back into meaningful text if you have the model weights and the tokenizer.
171.
by diwank 2y ago
It can only be decrypted using the encryption key. The inputs are encrypted and look like a gibberish and so are the outputs. It preserves privacy by essentially making the inputs and outputs unreadable.
172.
CryptGPT: A Simple Approach to Privacy-Preserving LLMs Using Vigenere Cipher(huggingface.co)
10 points by diwank 2y ago|13 comments
173.
by diwank 2y ago
The Pagani Utopia is fairly new (launched September 2022) and is completely bereft of screens and gizmos. A complete work of art costing a mere $2.5 million
174.
FBI's List of Famous Cases and Criminals(fbi.gov)
1 points by diwank 2y ago|0 comments
175.
by diwank 2y ago
I would love to do research in Foundation Models and Philosophy of Mind.
176.
Google is about to change the whole Internet – again(nymag.com)
5 points by diwank 2y ago|1 comments
177.
by diwank 2y ago
Lolll. It’s 1024 but only for documents and not the tools (we changed the embedding model for RAG)
178.
by diwank 2y ago
Not as a “junior dev” that sweeps markets itself as but it is useful in its own ways. For example, one really nifty way I found to effectively use it is to: - git diff - gh issue create “sweep: update docs for this file change” for every fi
179.
by diwank 2y ago
Right, I stand corrected, I think I confused the branding of other competing products. I remember really liking the fact that ellipsis does _not_ sell itself as a developer. I’ll edit my comment to reflect that. :)
180.
by diwank 2y ago
We are using ellipsis and sweep both for our open source project and they are quite helpful in their own ways. I think selling them as an automated engineer is a little over the top at the moment but once you get the hang of it they can spo

More ›