python expert, wannabe hacker, dungeon master, he/him
France & Europe
Joined August 2010
- Tweets1.8K
- Following483
- Followers790
- Likes5
wapiflapi retweeted
I made an easy decision tree for anyone considering using git submodules or a monorepo.
wapiflapi retweeted
An unexpected result: Cutting emissions by ships has increased global warming.
science.org/content/article/…
wapiflapi retweeted
Ah yes, the well known "except you, FAANG" clause that's so common in *open source* licenses like GPL, MIT, BSD, Apache2, ...
Here I go again, this can't be for real lol
This is huge: Llama-v2 is open source, with a license that authorizes commercial use!
This is going to change the landscape of the LLM market.
Llama-v2 is available on Microsoft Azure and will be available on AWS, Hugging Face and other providers
Pretrained and fine-tuned models are available with 7B, 13B and 70B parameters.
Llama-2 website: ai.meta.com/llama/
Llama-2 paper: ai.meta.com/research/publica…
A number of personalities from industry and academia have endorsed our open source approach: about.fb.com/news/2023/07/ll…
wapiflapi retweeted
It's so stupid to try to teach teenagers "entrepreneurship" by having them pitch made-up startup ideas. You make them focus on the one thing they shouldn't be focusing on, and then you rate them by their appeal to investors, instead of users.
wapiflapi retweeted
It is if you want to start a startup. It might be good if you wanted to start a small business.
wapiflapi retweeted
this is wild — kNN using a gzip-based distance metric outperforms BERT and other neural methods for OOD sentence classification
intuition: 2 texts similar if cat-ing one to the other barely increases gzip size
no training, no tuning, no params — this is the entire algorithm:
![for (x1, _) in test_set:
Cx1 = len(gzip.compress(x1.encode()))
distance_from_x1 = []
for (x2, _) in training_set:
Cx2 = len(gzip.compress(x2. encode())
x1x2 = " ".join([x1, x2])
Cx1x2 = len(gzip.compress(x1x2. encode())
ncd = (Cx1x2 - min(Cx1,Cx2)) / max(Cx1, Cx2)
distance_from_x1.append(ncd)
sorted_idx = np.argsort(np.array(distance_from_x1))
top_k_class = training_set[sorted_idx[:k], 1]
predict_class = max(set(top_k_class), key=top_k_class.count)](https://pbs.twimg.com/media/F05HxqsWwAAHRCI.jpg?name=orig)
ALT for (x1, _) in test_set: Cx1 = len(gzip.compress(x1.encode())) distance_from_x1 = [] for (x2, _) in training_set: Cx2 = len(gzip.compress(x2. encode()) x1x2 = " ".join([x1, x2]) Cx1x2 = len(gzip.compress(x1x2. encode()) ncd = (Cx1x2 - min(Cx1,Cx2)) / max(Cx1, Cx2) distance_from_x1.append(ncd) sorted_idx = np.argsort(np.array(distance_from_x1)) top_k_class = training_set[sorted_idx[:k], 1] predict_class = max(set(top_k_class), key=top_k_class.count)
this paper's nuts. for sentence classification on out-of-domain datasets, all neural (Transformer or not) approaches lose to good old kNN on representations generated by.... gzip aclanthology.org/2023.findin…
wapiflapi retweeted
A 14-line Python script using gzip outperforming a 345m parameter transformer model is probably the most hilarious result I've seen all year.
aclanthology.org/2023.findin…
wapiflapi retweeted
In addition to maintaining pwntools and pwndbg which lots of folks know and love, and on top of being a great CTF player, @ebeip90 was also just a super nice person.
This has definitely not been a good year for the infosec community :(
I learnt that Zach Riggle, the author of pwndbg, passed a few days ago.
We weren't close friends, but acquaintances, and discussed stuff in Twitter DMs.
I am sad. My Twitter DMs have lost multiple good folks in the last years :(
smithfamilyfuneralhomes.com/…
wapiflapi retweeted
As a specialist in evaluating language models, I declare that this is the best way of evaluating language models:
wapiflapi retweeted
Archive of all versions of libc for a lot of architectures since 2014 (over 5k versions, it's still being updated). We used it for exploit. Thanks @bestpig for formatting and publishing github.com/BestPig/libc-db
wapiflapi retweeted
If more than 4 billion ppl in fact tune in to the Queen's funeral today, that's more than half the world. And it would have been the *entire* world around 1980. mirror.co.uk/news/royals/que…
wapiflapi retweeted
On April 18, 1930 the BBC reported "There is no news, so we're offering this collection of poems".
wapiflapi retweeted
🚨 ALERT 🚨
Python's ctx library and a fork of PHP's phpass have been compromised. 3 million users combined.
The malicious code sends all the environment variables to a heroku app, likely to mine AWS credentials.
wapiflapi retweeted
A year ago, our team set out to create structured documentation (aligned with @MITREattack) for each and every Android malware technique we have observed in the wild. We just passed 750 techniques and our growth rate is so fast that we're clearly nowhere near the end.
wapiflapi retweeted
This is a maddening review of the U.S. pandemic response, by economist @ATabarrok. Every American should watch this; certainly every politician should. So many unforced errors. youtu.be/IIzT-IlJsts
wapiflapi retweeted
We just hacked a billionaire!
Got consent 1st then got to work hacking Jeffrey Katzenberg. @Evantobac & I stole his pics, emails, and contacts then turned on his mic (without an indicator light) & listened to his phone calls.
Here's the video on how we hacked a billionaire:
