HNHacker News
TopNewBestAskShowJobs

sean_pedersen

368 karma · joined August 8, 2019

https://seanpedersen.github.io/
submissionscomments
sean_pedersen··on Ask HN: Why isn't email spam solved using Proof of Work yet?
If they add PoW to their malware, it will slow them down or even alert users / admins because of the high CPU usage it will cause.
sean_pedersen··on Ask HN: Why isn't email spam solved using Proof of Work yet?
Yes pass it down to the user by computing eMail PoW client side if that is a concern for your business...
sean_pedersen··on Ask HN: Why isn't email spam solved using Proof of Work yet?
Not every server needs to understand it. Only the ones wanting to participate. If you choose to filter by PoW and a server does not provide PoW, you will still receive all of its emails just flagged as spam. So email PoW roll out can work iteratively.
sean_pedersen··on Show HN: HyperTag 0.5.0 – Semantic Search for Images Using Text Queries
Just released 0.5.0 which adds semantic search for images using text queries. This is powered by OpenAI's CLIP model (https://openai.com/blog/clip/) released on Jan. 5, 2021. Works pretty good on my end - awesome stuff!
sean_pedersen··on HyperTag 0.4.3 – Introducing Semantic Search for Text Documents (yes, even PDF)
Github Repo: https://github.com/SeanPedersen/HyperTag
sean_pedersen··on Show HN: HyperTag – File management made easy using tags
Hey, I have been hacking on HyperTag over the last few days. I am keen on feedback and potential contributors.

From the README: What is the point of HyperTag's existence? HyperTag offers some unique features such as the import function that make it very convenient to use. Also HyperTag's code base is written in Python and thus extremely small (<500 LOC) compared to TMSU (>10,000 LOC) and SuperTag (>25,000 LOC), making it very easy to modify / extend / fix it yourself.

sean_pedersen··on The Future of Programming
In this blog post I lay down why I think existing programming eco systems could greatly benefit from content and behavior addressable code.
sean_pedersen··on A Recipe for Training Neural Networks by Andrej Karpathy
IMO a great collection of advice. A must read for anyone planning to or already using neural networks in research or business applications.
sean_pedersen··on The Erlang Rationale (2008) [pdf]
I had the pleasure of using Erlang in Data Structures, Graph Theory and Distributed Systems classes in university, had been lucky with the professor.
sean_pedersen··on Print RAM Usage of process including children
Yup usage is simple:

$ ram firefox

Will print total firefox RAM usage.

On my machine (Arch Linux) it works without root...

sean_pedersen··on Setting up your own Nextcloud instance to get off of Google
I concur. I am happy with my lean setup:

- SyncThing: decentralized, fast & unix aligned

- CryptPad: end-to-end encrypted, real-time collaborative editing

sean_pedersen··on A Nuclear Powered Nuclear Waste Dispenser
Another possible propulsion method: Build a giant Gauss cannon [0]. Using electricity though would add another energy transformation step reducing overall efficiency. Water steam would be directly produced off the nuclear reactors heat.

[0]: https://en.wikipedia.org/wiki/Coilgun

sean_pedersen··on A Nuclear Powered Nuclear Waste Dispenser
Found an image that visualizes the multistage concept of the cannon I image: https://en.wikipedia.org/wiki/File:Haskell_USP484011_1892_01...
sean_pedersen··on A Nuclear Powered Nuclear Waste Dispenser
Interesting, thanks. The second link is kinda related to my idea but would produce fallout for every launch, so not really practical.
sean_pedersen··on A Nuclear Powered Nuclear Waste Dispenser
Just an idea that occurred to me. Curious if this could work out.
sean_pedersen··on Show HN: Self-hosted offline Internet from your browsing history
This would be nice to have in combination with IPFS. https://ipfs.io/
sean_pedersen··on On generalization problems of ML models in real-world applications
Abstract: ML models often exhibit unexpectedly poor behavior when they are deployed in real-world domains. We identify underspecification as a key reason for these failures. An ML pipeline is underspecified when it can return many predictors with equivalently strong held-out performance in the training domain. Underspecification is common in modern ML pipelines, such as those based on deep learning. Predictors returned by underspecified pipelines are often treated as equivalent based on their training domain performance, but we show here that such predictors can behave very differently in deployment domains. This ambiguity can lead to instability and poor model behavior in practice, and is a distinct failure mode from previously identified issues arising from structural mismatch between training and deployment domains. We show that this problem appears in a wide variety of practical ML pipelines, using examples from computer vision, medical imaging, natural language processing, clinical risk prediction based on electronic health records, and medical genomics. Our results show the need to explicitly account for underspecification in modeling pipelines that are intended for real-world deployment in any domain.
sean_pedersen··on How is search so bad? A case study
Cuz it is goddamn hard!?
sean_pedersen··on German cartel office to take Facebook case to High Court
"BERLIN (Reuters) - Germany’s cartel office said on Monday it would appeal a regional court decision to suspend restrictions it had placed on Facebook’s data collection practices to the country’s highest court. "

Am I understanding this correctly: They are removing restrictions, so this a win for FB?

← PreviousPage 3 of 3