Mine was a comment to say that llms are not just fancy auto complete. Although technically an evolution, it is a bit like saying humans are fancy worms because evolved from worms
Mine was a comment to say that llms are not just fancy auto complete. Although technically an evolution, it is a bit like saying humans are fancy worms because evolved from worms
Common Reinforcement learning methods definitely are.
LLMs are an evolution of a markov chain as any method that is not a markov chain… I would say not directly. Clearly they share concepts as any method to simulate stochastic processes, and LLMs definitely are more recent than markov processes. Then anyone can decide the inspirations.
What I wanted to say is that, really, we are discussing about a unique new method for LLMs, that is not just “old stuff, more data”.
This is my main point.
A markov chain models a process as a transition between states were transition probabilities depends only on the current state.
A LLM is ideally less a markov chain, more similar to a discrete langevin dynamics as both have a memory (attention mechanism for LLMs, inertia for LD) and both a noise defined by a parameter (temperature in both cases, the name temperature in LLM context is exactly derived from thermodynamics).
As far as I remember the original attention paper doesn’t reference markov processes.
I am not saying one cannot explain it starting from a markov chain, it is just that saying that we could do it decades ago but we didn’t have the horse power and the data is wrong. We didn’t have the method to simulate writing. We now have a decent one, and the horse power to train on a lot of data
We do. I pay to work with it, I want it to do what I want, even if wrong. I am leading.
Same for all professionals and companies paying for these models
It’s a bit like saying a human being is a fancy worm. Technically it is true, we evolved from worms, still we are pretty special compared to worms
LLMs are not markovian, as the new word doesn’t depend only on the previous one, but it depends on the previous n words, where n is the context length. I.e. LLMs have a memory that makes the generation process non markovian.
You are probably thinking about reinforcement learning, which is most often modeled as a markov decision process
That’s borderline criminal. Nothing more to add
As far as I understand, these are posix requirements https://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18
Powershell is not compliant with that document even now in the interactive part. Wsl2 is, as one can istall a standard Linux shell
Thanks, this explains:
The Windows NT POSIX subsystem did not provide the interactive user environment parts
So the interactive part, the shell itself, is not compliant. That is why I was confused
How can it be? It’s oo. Not saying you’re wrong. Honestly curious
Upvoted for your courage!
It targets router firmwares though… These bot farms do not usually target real gnu/Linux os, because it is easier and more effective to attack router firmwares that are not well configured by producers and telcoms, and are practically never upgraded.
Therefore they are not a real threat for standard mint or popOS user… Let alone gentoo users
I use python professionaly. Never seen a real successful supply chain attack on libraries used by “normal” people. There was recently a supply chain attack to pytorch, that I remember, but it was solved within few hours.
It is not a real risk for non developers. It is a risk, but veeery low, miles lower than pdf.exe.
Just check this stat for ransomwares taken as an example of viruses: https://www.statista.com/statistics/701020/major-operating-systems-targeted-by-ransomware/
Windows server is ~20% of server market. Still it is there second, with in practice no GNU/linux (80% of server market). This is why people do not really worry much, the risk exists, but it is minimal for well configured system compared to competition, even where competitors are a niche and Linux machines are the main target.
On windows, an antivirus is not a bad idea… On Linux, a firewall and basic care are usually sufficient
I agree with you, but, it is also true that the overwhelming majority of ransomwares affect windows https://www.statista.com/statistics/701020/major-operating-systems-targeted-by-ransomware/
Linux is not a significant target despite being so diffused
Edit. For those downvoting, windows server is ~20% of the server market and it is second in that stat. GNU/Linux distros such as rhel, debian and so on are almost 80% of server market and still there are no sufficient attacks reported to end up in that stat
Ok, than the experiment you are doing is just to check how many attacks you can get over a certain time… It is not really representative of a common use case. And again, this is not a virus. It is a successful attack from a bot on a purposely misconfigured service exposed to the internet. An antivirus is not needed. What is needed is basic configuration. An antivirus cannot help there
And disable password authentication as first step
Does the attack succeed? Never happened to me. You see bot trying, but really never seen succeeding irl. How is it configured?
Do you have also a rdp honeypot by chance? Do you see different rates of attack? Honestly curious.
I don’t have any windows licenses around, otherwise, it would have been an interesting test
Not at all. You leave a ssh port open, you don’t necessarily get a virus. Try it. Set up a raspberry pi, install ssh and leave the port open in your firewall. It is much less risky than exposing rdp (the most comparable windows protocol) on windows for instance.
It is a security risk, but absolutely not comparable of installing pdf.exe. Not even in the same league of risk.
As said, try it now and tell me how it goes.
There is a lot of misinformation around security on Linux
I have been using linux for almost 2 decades, never seen a virus. And I never heard of a colleague or friend who got one on Linux. That’s why no one has ever installed an antivirus, because, till now, the risk has been practically zero.
On windows, on the other hand, I saw so many viruses on friends and relatives computers…
People install antiviruses depending on the experience.
To be fair, we all know on Linux viruses exist, but is objectively pretty difficult to get one. It is not worth installing an antivirus if one doesn’t actively install garbage from untrusted sources
Spaghetti is 1 type of long noodles. You have spaghetti, tagliatelle, bucatini, linguine, vermicelli…