I create proper libraries. I don’t do snippets because they make code dirty, redundant and difficult to read on the long run.
I actively discourage people in my team to use snippets copy and pasted everywhere themselves. If it’s reusable code, it should be usable by everyone and well tested
This was my immediate reaction as well.
For those who like living a messy life, there’s always Visual Studio (the original beast, not VSCode)
VSCode supports it also for other shells. This repo is not about vscode, it’s about actual shells. You are the one incorrect in this case
When someone brings points in the discussion, you react like a fan boy student that just bought the new gaming laptop.
Could you please reply about the discussion or go back to school? I am too old for your “no, you shit, you stoopid”
I wonder myself why I keep answering to your comments and that’s why this is my last comment
Unfortunately… That is why I commented above…
Powershell is a shell that pretends to be an OO language and that fails dramatically at both.
It was a design mistake. Much better having a real separation, a real shell and a real OO language. As even Microsoft has recently understood. As you can see also in this case, where Powershell is the last entry
What is a real advantage compared to plain autocomplete? Was it trained to know command flags?
It is anyway nice to see that even in Microsoft the abomination of Powershell is the lowest in the list of shells. It is probably time for them to drop it completely
I know, I shouldn’t play payday 2… Friends forgive me!
ML/AI is a huge field. If you don’t like 1 side of it, there are millions others.
ML/AI. Everything else has been eaten by agile/product owners/MBAs/micromanagement. Luckily those people still don’t understand AI… And AI is still stochastic, pi planning and burnout charts don’t work.
If agile bites your ass… Run away, you are too young for wasting your life
All scientific computing is built on top of fortran. Even cutting edge AI runs on top of high performance libraries written in fortran and c. Simply there is less need for fortran developers because high performance subroutines are wrapped to be called by higher level languages, such as python
First time I see “cool” and “c#” on the same sentence. I’ve always thought stereotype of c# is that it is the language for corporate, extremely uncool projects.
Just a comment. Cobol nowadays is heavily outsourced. There are jobs but not so lucrative as in the past. Fortran is still strong in scientific computing, but nowadays it is wrapped in python. All people I know who were strong in fortran (me included) are nowadays mostly working with python or scala, most of us on ML/AI related stuff.
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
What is a linux apologist?
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 requires user profiling, something most lemmy users are vocally against.
It also requires a data science team and a bunch of additional legal stuff.
Lemmy user here, I don't see it either… The guy has bad settings
In the next issue of “non interesting tech news”,
Prepare yourself for the non hype