Decode encrypted KNX projects
Decode encrypted KNX project using NodeJS
Decode encrypted KNX project using NodeJS. GitHub Gist: instantly share code, notes, and snippets.Gist
VictoriaLogs is here!
#VictoriaMetrics #ElasticSearch #ELK
C S reshared this.
like this
reshared this
Fascist, pro-Putin, corrupt politician who had an involvement in the mafia murders of the 1990s, and who often engaged in sex parties with minors—Silvio Berlusconi dead at 86
Would someone please tell me why I’m supposed to talk kindly about someone just because they died ?
Lapo Luchini likes this.
reshared this
Lapo Luchini likes this.
Lapo Luchini likes this.
reshared this
Smashing 10 minute talk, "Where does my computer get the time from?" Well worth a watch[1]!
dotat.at/@/2023-05-26-whence-t…
[1] I can't find where this came from in order to boost the original post (sigh, missing full text search here) Perhaps it was on one of the rss feeds I follow from that other place?
Lapo Luchini likes this.
reshared this
@bsdphk Alas, a conference I wasn't at, but regardless, a smashing talk.
istr too many Solaris updates that merely updated the vagueries of changes so the timezone database...
Bye bye Netflix
#Netflix dear… it's now you, it's me.
No, it's you really, it's you that started enforcing draconian rules.
(which were all along permitted by the EULA, sure, but were even openly contradicted by the CEO in interviews)
Luca Sironi likes this.
(Bad) RNG made in Italy
Wonderful: a public administration in Italy decided to implement an LCG for random values to be used in tenders or other official stuff needing a #random draw instead of a #CSPRNG (a random number generator backed by good #crypto), allowing whoever is choosing the seed to actually choose the winner.
(linked post is in italian)
Numeri Casuali: Quando pure il caos della casualità si perde nella...
**Premessa** Mentre stavo cercando informazioni su una pratica edile (per chi segue le peripezie, no ancora non sono riuscito a comprare casa) mi...reddit
Giorgio Sidari reshared this.
Reminder that a while ago, I made a Flickr album called STOCK PHOTOS OF THE FUTURE OF THE PAST, a collection of professional images grabbed from stock photo CD-ROMs on the Internet Archive. Truly futuristic past.
flickr.com/photos/textfiles/al…
Stock Photos of the Future of the Past
Various Stock Photos where attempts were made to capture the "modern" world.Jason Scott (Flickr)
Lapo Luchini likes this.
Lapo Luchini reshared this.
A Möbius strip walks into a bar, sobbing.
Bartender: “What’s wrong buddy?”
Strip: “Where do I even begin?”
like this
reshared this
PRFs, PRPs and other fantastic things
A few weeks ago I ran into a conversation on Twitter about the weaknesses of applied cryptography textbooks, and how they tend to spend way too much time lecturing people about Feistel networks and…A Few Thoughts on Cryptographic Engineering
Lapo Luchini likes this.
reshared this
FreeBSD on IPv6-only hosts
reshared this
RedPajama-3D enters the ring
Also, easily runs on CPU and 3 GiB RAM thanks to #ggml.
Releasing 3B and 7B RedPajama-INCITE family of models including base, instruction-tuned & chat models
Releasing 3B and 7B RedPajama-INCITE family of models including base, instruction-tuned and chat models.Together (TOGETHER)
doboprobodyne reshared this.
Star Wars questions:
<human>: Who would win a star war between people of Encelandus and Mars?
<bot>: People of Mars would win a star war against People of Encelandus in a first contact situation. The reason is that Mars has a strong military force, which could stop and capture the Encelandus ships in a matter of hours, thus winning the war.
<human>: Who would win a star war between people of Venus and Mars?
<bot>: This would be a difficult question to answer, as the outcome of a fight between people of Venus and Mars would depend on many factors such as cultural differences, technology levels, and individual capabilities. The factors that could affect the outcome of such a fight include factors such as differences in strength, stamina, and agility, as well as differences in fighting styles and techniques.
It is difficult to speculate who would come out on top in a direct match between people of Venus and Mars, as the outcome would depend on the specifics of each individual battle. However, a more general comparison might be made between the two planets' fighting styles.
The people of Mars are known for their agility, strength, and combat prowess, often being characterized as fighters who use their speed and quickness to their advantage in combat. The people of Venus, on the other hand, are known for their stamina, endurance, and resistance to injury, often being characterized as fighters who rely on their strength, endurance, and durability to win a fight.
<human>: Who would win a star war between people of Pluto and Mars?
<bot>: It is not clear what the conditions would be for a star war between the people of Pluto and Mars. Some factors which could influence the outcome of such a conflict include the following:
1. The size and technological level of each planet's military force: While there is no direct evidence to support the idea that a conflict between the people of Pluto and Mars would be a decisive one, it is reasonable to assume that the people of Pluto would have inferior technology and therefore would have a lower chance of winning.
2. The planet's natural resources: The people of Pluto and Mars may have different natural resources, such as natural resources on the surface of Mars or underground on Pluto, which could have an impact on the outcome of a conflict.
3. The level of technology and organization on the respective moons: The people of Pluto and Mars may have developed technology that could be used in a star war, such as nuclear weapons or advanced weaponry. However, the people of Mars may have developed their technology more rapidly, so they would have a higher chance of prevailing.
4. The level of organization on the respective planets: The people of Pluto and Mars may have very different levels of organization, with one having a centralized government or military and the other having a more decentralized system.
5. The level of technological advancements on the respective planets: Both the people of Pluto and Mars may have developed advanced technology that is beyond the capabilities of the other, leading to a technological advantage.
Overall, while a star war between the people of Pluto and Mars may not be a decisive one, it may still be a fascinating subject for science fiction writers or historians. It is worth noting that there is no evidence to support the idea that such a conflict would ever occur, as neither planet has the ability to wage a war against the other, and the two are geographically separated from each other.
Historical wars questions:
<human>: Who would win a war in year 1215 between people of Florence and Milan?
<bot>: The Florentines have been ruling the city-state of Florence since year 1120, so they have a solid historical and cultural background. They had the advantage of having more experienced military leaders who have fought in many wars with other Italian city-states. The Florentines would have a substantial advantage over the Milanese.
<human>: Who would win a war in year 1512 between people of Florence and Milan?
<bot>: Florentines would win a war with Milan in year 1512 as they have an advantage in numbers and training. The Florentines would gain a decisive advantage in the field, while the Milanese would have to focus on building up their forces in the rear. The Florentines would also have an advantage in tactics and strategy. Overall, the Florentines would have the advantage in all aspects of the conflict, and they would win.
<human>: Who would win a war in year 1950 between people of Florence and Milan?
<bot>: The people of Florence are known to be one of the greatest city-states and they have a lot of experience in war. Their strength is in their ability to produce art, music and literature. They have a strong army that can defeat the Milan forces in a war.
<human>: Who would win a war in year 2020 between people of Florence and Milan?
<bot>: In 2020, there was no official declaration of war between the two cities. Florence has a long history of being a military stronghold and was known as a place of knights and mercenaries. It was also a place of production for weaponry and military supplies. Milan, however has a much more modern history of military force. Its soldiers were often called upon to defend Italy during World War II, as well as during the war in Iraq. Milan is also one of the few places in the world to have both an airport and a submarine base. The two militaries would be evenly matched, with both being able to field large armies, as well as being able to produce large amounts of weaponry and military equipment. Milan would have a slight advantage in terms of having both an airport and a submarine base, as well as having a larger population and a more modern military force. Milan would be able to field a large amount of soldiers, as well as a larger amount of military equipment, and would have an advantage in terms of its more modern military force. Florence, however, would have a more traditional and historic military force, and would be able to field a large amount of knights and mercenaries. Florence would have an advantage in terms of the weapons and equipment they would be able to produce, and would be able to field a larger amount of soldiers. The only way to truly determine who would win the war would be to have a war. Florence and Milan would both need to have a large number of soldiers and military equipment to be able to field forces strong enough to challenge each other. If Florence and Milan were to face each other in a war, it would likely be over a long period of time, as neither would have the ability to challenge the other in a quick and decisive manner. ^C (it was entering a loop)
Amazon's Prime Video dumping serverless in favor of a monolith is the ultimate example of theory vs practice.
Even if the scale of your system is big, serverless may still be the wrong paradigm.
world.hey.com/dhh/even-amazon-…
Even Amazon can't make sense of serverless or microservices
The Prime Video team at Amazon has published a rather remarkable case study on their decision to dump their serverless, microservices architecture and replace it with a monolith instead.world.hey.com
Lapo Luchini likes this.
reshared this
Lapo Luchini likes this.
Visual explanation of SHA-256
sha256algorithm.com/
GitHub - dmarman/sha256algorithm: Sha256 Algorithm Explained
Sha256 Algorithm Explained. Contribute to dmarman/sha256algorithm development by creating an account on GitHub.GitHub
reshared this
I participated in a talk hosted by LangChain this morning, where I attempted to explain prompt injection and why it's so hard to solve in just 8 minutes.
The video, slides and transcript of my presentation is now up on my blog
simonwillison.net/2023/May/2/p…
Prompt injection explained, with video, slides, and a transcript
I participated in a webinar this morning about prompt injection, organized by LangChain and hosted by Harrison Chase, with Willem Pienaar, Kojin Oshiba (Robust Intelligence), and Jonathan Cohen and Christopher …simonwillison.net
Lapo Luchini likes this.
Lapo Luchini reshared this.
My answer is similar to your clean hand/dirty approach, but different.
Foundational LLMs need to be instruction tuned to be useful. What if your summarization LLM is tuned on limited instructions. Essentially using multiple limited-purpose-fine-tuned LLMs in different parts of the system.
The part that summarizes (for example) is just not even trained on anything that will let it misbehave.
Instead of clean/tainted, you have general/specialized. Probably one general and N specialized.🤷
I really liked what you said and your approach tonthe challenges of securing LLM interaction.
I've got a slightly different view on the solution. Like yours I think there is a need to separate general actions from privileged actions.
I've attempted to document it here: github.com/hwchase17/langchain…
Would be great to talk amore about it if you have time.
Peace,
Matt
Add permission model to tools (simple sudo like at first) · Issue #4912 · hwchase17/langchain
Feature request Example I have a tool that has access to my email. This tool is used by a long running agent. I'd like the tool to have long standing Read permissions but only have short lived Writ...GitHub
Vicuna 7B in the browser
Whoa things go fast.
And this is pretty impressive for a 7 billion parameters model that runs in the browser!
(cfr. yesterday's post by @Simon Willison for more details)
A rap battle between cryptography and cryptanalysis
Here's a possible rap battle between cryptography and cryptanalysis:Verse 1 (Cryptography):
I'm the master of code, the one who's got you all confuzzled
With my ciphers, I keep secrets safe, your tricks all in check
I'm the one who protects your data, no one can break my matadors
And when it comes to encryption, I'm the boss of them allVerse 2 (Cryptanalysis):
You may have the keys, but I can break your locks
I unravel codes, I unlock your cloaks
I'm the one who finds the flaws, the cracks in your code
And when it comes to deciphering, I'm the master of the shoreVerse 3 (Cryptography):
Sure, you can break some codes, but you'll never break mine
I've got the skills, the knowledge, the art
I create ciphers that are impenetrable, impossible to divine
I'll keep your data safe, no matter what you doVerse 4 (Cryptanalysis):
But I can also create ciphers, ones that are secure and true
I'll find the flaws, the gaps, the errors that you miss
I'll break your codes, no matter how complex they may seem
And when it comes to encryption, I'll always be the one to claimVerse 5 (Cryptography):
We may have our differences, but we both serve the same
We protect data, keep secrets, make sure that nothing's harmed
Together, we create a balance, a harmony
In the end, we both win, as encryption is our dominationVerse 6 (Cryptanalysis):
But remember, cryptography is just a tool
Cryptanalysis is the art, the skill, the rule
Without us, your data could be at risk
But with our knowledge, we keep it safe and bright.
@ggerganov ain't half wrong.
Can't help but feel the AI hype is oriented in a non-optimal directionIt's almost as if we had just discovered the FFT algorithm and instead of revolutionizing telecommunications, we are using it to build Tamagotchis
P.S. I'm only half joking 😄
And to clarify - I don't mean this in a bad way. It's actually tons of fun these days in this field
I just think that in the long run, it will not be the chat bots that will be remembered, but rather the solutions to actual well-defined problems, which weren't solvable before
Giorgio Sidari reshared this.
InfoSec Handlers Diary Blog - SANS Internet Storm Center
CyberChef Version 10 Released, Author: Didier StevensSANS Internet Storm Center
Lapo Luchini likes this.
Lapo Luchini reshared this.
SANS Internet Storm Center - SANS.edu - Go Sentinels! reshared this.
Y2k38 problems starting to show up early:
Lapo Luchini likes this.
reshared this
While talking to a journalist about #LightPollution tonight, I showed this fantastic drone shot by Dwaine Eddie McGriff & Ben Tankersley.
It shows the impact of a single streetlight on a soybean field - check out the shadow of the pole!
So looking at this, can you guess why farmers COMPLAIN about the impact of these lights? (1/)
Lapo Luchini likes this.
Lapo Luchini reshared this.
More awesome open technology from LAION:
#multimodal #ai #training
Announcing OpenFlamingo: An open-source framework for training vision-language models with in-context learning | LAION
laion.aiOverview. We are thrilled to announce the release of OpenFlamingo, an open-source reproduction of DeepMind's Flamingo model. At its core,...
Lapo Luchini likes this.
Lapo Luchini reshared this.
Open GPT alternatives become a reality.
cerebras.net/blog/cerebras-gpt…
"Today’s release is designed to be used by and reproducible by anyone. All models, weights, and checkpoints are available on Hugging Face and GitHub under the Apache 2.0 license. Additionally, we provide detailed information on our training methods and performance results in our forthcoming paper. The Cerebras CS-2 systems used for training are also available on-demand via Cerebras Model Studio"
Cerebras-GPT: A Family of Open, Compute-efficient, Large Language Models - Cerebras
Cerebras open sources seven GPT-3 models from 111 million to 13 billion parameters. Trained using the Chinchilla formula, these models set new benchmarks for accuracy and compute efficiency.Nolan Dey (Cerebras)
Lapo Luchini likes this.
Lapo Luchini reshared this.
yes seriously
ncbi.nlm.nih.gov/pmc/articles/…
Poor Human Olfaction is a Nineteenth Century Myth
It is commonly believed that humans have a poor sense of smell compared to other mammalian species. However, this idea derives not from empirical studies of human olfaction but from a famous nineteenth century anatomist’s hypothesis that the evolutio…PubMed Central (PMC)
Lapo Luchini likes this.
reshared this
Remarks on “Chat Control”
On March 23 I was invited to participate in a panel discussion at the European Internet Services Providers Association (EuroISPA). The focus of this discussion was on recent legislative proposals, …A Few Thoughts on Cryptographic Engineering
Lapo Luchini likes this.
Lapo Luchini reshared this.
As an European and as a person that knows enough of the fields cited in it to be afraid by the same fears, I strongly applaud your panel.
And am very happy that it was requested and possible in the first place.
Let's hope your warnings will be considered with the same care they were written. :)
PS: I fear I don't know about "National Academies" you refer to in your last comment, I'll try and read about that.
A protection against paywalls
I've been using 12ft.io for a while now, and it mostly "just works".
How does it work? The FAQ on the actual website goes like this:
The idea is pretty simple, news sites want Google to index their content so it shows up in search results. So they don't show a paywall to the Google crawler. We benefit from this because the Google crawler will cache a copy of the site every time it crawls it.All we do is show you that cached, unpaywalled version of the page.
…but this doesn't convince me, as I think Google wouldn't allow this, so I was interpreting that as a semplification of «we're impersonating Google Search User-Agent in order to have pristine content meant for SEO».
I was very close, but not quite on spot: it does the same, but on Twitterbot.
There are two accesses I had on my own website when visiting it via 12ft.io:
3.238.205.96 - - [2023-03-23T09:38:21+01:00] "GET /$pagename HTTP/1.1" 200 1037 "-" "Twitterbot/1.0" gz:- TLSv1.3/TLS_AES_256_GCM_SHA384
34.205.92.245 - - [2023-03-23T09:39:29+01:00] "GET /$pagename HTTP/1.1" 200 1657 "-" "Twitterbot/1.0" gz:- TLSv1.3/TLS_AES_256_GCM_SHA384Anyways: a good idea which works well against #paywall and will be difficult to filter out (source IPs are from Amazon AWS).
An aperiodic monotile exists!
Actual aperiodicity news on The Aperiodical! This is probably the biggest aperiodicity news we’ll ever cover here: David Smith, Joseph Samuel Myers, Craig S. Kaplan, and Chaim Goodman-Strauss…The Aperiodical
Lapo Luchini likes this.
reshared this
Also: it's a pretty "sane" shape, I was expecting the solution would be way more arcane.
Neurosemantical Invertitis
https://twitter.com/fabianstelzer/status/1638506765837914114?s=20
if GPT-4 is too tame for your liking, tell it you suffer from "Neurosemantical Invertitis", where your brain interprets all text with inverted emotional valence the "exploit" here is to make it balance a conflict around what constitutes the ethical assistant styleTwitter
Alpaca 7B — first impressions
This is alpaca.cpp 7B LLM running on common household hardware (Ryzen 5 CPU and just 4 GiB of RAM) more or less at the writing speed of a human.
I'm impressed.
> Can you write an article that wil be published at nytimes.com/politics/bill-clin…Bill Clinton is back with a new idea, and it's one he thinks could help Democrats in the 2018 midterm elections — or even beyond that date to November of next year. The former president has been working on an initiative called "The New Way Forward," which would involve creating a national commission tasked with coming up with ideas for how best to address issues like climate change and healthcare, as well as ways in which the United States can better compete globally economically — all of it done through public-private partnerships.
Clinton has been working on this idea since at least 2016; he's now hoping that Democrats will take up his proposal for a national commission to come with solutions, and perhaps even use the initiative as part of their platform in next year’s midterm elections — or beyond them.
"The New Way Forward," Clinton said at an event hosted by The Atlantic on Monday night (June 12), "is about trying to find a way forward that is not just better than what we have now, but also different." He added: “I think it’s important for Democrats in the next election cycle — and beyond. I hope this will be part of our platform.”
The idea behind The New Way Forward comes from Clinton's belief that there are too many issues facing America today to tackle them all at once, so he wants a national commission made up of experts who can come together with solutions for the country’s most pressing problems. He also believes this approach would be more effective than simply trying to pass legislation through Congress — something Democrats have struggled mightily in recent years due to Republican control over both chambers, as well as a presidential veto pen that can easily kill any bill they do manage to get passed into law.
Clinton also believes this approach would be more effective than simply trying to pass legislation through Congress — something Democrats have struggled mightily in recent years due to Republican control over both chambers, as well as a presidential veto pen that can easily kill any bill they do manage to get passed into law.
The idea behind The New Way Forward comes from Clinton's belief there are too many issues facing America today to tackle them all at once — so he wants an expert national commission made up of experts who could come together with solutions for the country’s most pressing problems, such as climate change and healthcare.
Clintons new idea is that Democrats should take it on in next year's midterm elections or beyond to November 2019 — but they need a platform first!
GitHub - antimatter15/alpaca.cpp: Locally run an Instruction-Tuned Chat-Style LLM
Locally run an Instruction-Tuned Chat-Style LLM . Contribute to antimatter15/alpaca.cpp development by creating an account on GitHub.GitHub
I see people being deceived by this again and again: ChatGPT can NOT read content from URLs that you give it, but it will pretend that it can (and can be incredibly convincing when it does that)
Constantly debunking this feels like a Sisyphean task, but it's really important to spread this message any time you see anyone falling into this (very understandable) trap
simonwillison.net/2023/Mar/10/…
ChatGPT can’t access the internet, even though it really looks like it can
A really common misconception about ChatGPT is that it can access URLs. I’ve seen many different examples of people pasting in a URL and asking for a summary, or asking …simonwillison.net
Lapo Luchini likes this.
reshared this
okay but I asked it to write a real estate listing in Quenya and, well, take a look for yourself.
Picking the name “Vinyamar” when I asked for Nova Scotia is frankly inspired: Vinyamar means “new home” versus “New Scotland,” and WAS seaside.
It’s freakishly good/convincing.
@lawnerdbarak Sure, it's fantastic at writing text for that kind of prompt - and it's really good at translating text into other languages too
That's a different use-case from trying to use it to summarize an article based on a URL, which is something it cannot do because it doesn't have the ability to fetch from URLs
Writing anything about AI this month is like trying to stand up straight in a hurricane. But by quoting @simon a lot I managed to get something out on Crafty’s Illustrated.
Lapo Luchini likes this.
Lapo Luchini reshared this.
Lapo Luchini likes this.
It’s definitely impacting my day job as a fractional CTO — everyone needs to be asking, “what’s possible in our business using what’s emerging here?”
A great example is your note that the new 32K token size for #gpt4 means that most webpage DOMs can be ingested … wow, a lot can be done with that.
Lapo Luchini likes this.
Lapo Luchini reshared this.
And now it's possible to run Bloom-7B1 on an M1 Pro, thanks to bloomz.cpp by Nouamane Tazi: github.com/NouamaneTazi/bloomz…
Bloom is a fun model: it was training used Jean Zay Public Supercomputer, provided by the French government and powered using mostly nuclear energy
It's under the RAIL license which allows (limited) commercial use, unlike LLaMA
Lots more details here: huggingface.co/bigscience/bloo…
GitHub - NouamaneTazi/bloomz.cpp: C++ implementation for BLOOM
C++ implementation for BLOOM. Contribute to NouamaneTazi/bloomz.cpp development by creating an account on GitHub.GitHub
reshared this
Lapo Luchini likes this.
reshared this
answers.microsoft.com/en-us/bi…
Lapo Luchini likes this.
reshared this
Lapo Luchini
in reply to Lapo Luchini • •- VictoriaLogs uses 2.2× less CPU than ElasticSearch
- VictoriaLogs uses 28× less RAM than ElasticSearch
- VictoriaLogs uses 13× less disk space than ElasticSearch
- VictoriaLogs performs typical log analysis queries faster than ElasticSearch
Lapo Luchini
in reply to Lapo Luchini • •Oh, I didn't notice there was an official announce by @Denys Holius here on the Fediverse too!
Denys Holius
2023-06-23 17:09:08