this new gadget makes it easy to take Windows system events, and feed them into Copilot, looking for potentially malicious activity. And while it’s not perfect, it did manage to detect about 40% of the malicious tests that Windows Defender missed. It seems like LLMs are going to stick around, and this might be one of the places they actually make sense.

Yes, the pattern recognition engine is good at pattern recognition.

In all seriousness, it really would be great if we’d focused development of transformer models on stuff like this instead of everyone getting caught up in the fact that they can kinda sorta pass the Turing test and deciding that the singularity had arrived and they could be the ones to sell tickets to it.

Pro-tip: If you marry said friend, they have to seat you together anyway.

Nvidia Shield. The regular version is $150 US and from what I understand it gives flawless playback. I have the pro version which is more powerful, but that’s specifically for running games.

It’s Android TV OS, so app selection is great. You can load Smart Tube Next on there to get YouTube without ads, and there’s a very solid Jellyfin app. You can also use Kodi for local direct playback. Remote is perfectly functional, and you can use an app to rebind most of the keys.

Not even remotely. LLMs have failed to find any viable market fit.

The problem continues to be hallucinations and limited utility. This is compounded by the fact that LLMs are very expensive to run. The latter problem wouldn’t really be a problem if LLMs were truly capable of replacing a human employee, but they’re not. They’re just too unreliable for any serious enterprise grade application, and they’re too expensive for any low severity application.

For example, as a coding assistant, a lot of people quite like them. But as a replacement for a human coder, they’re a disaster. That means you still have to employ the expensive human, and you also have to pay an exorbitant monthly fee for what amounts to a very cool search engine.

There are tonnes of frivolous applications where they work really well. The AI girlfriend stuff, for example. A chatbot that sexts you is a very sellable product, regardless of how icky it might seem to some people. But no one is going to pay over $200 / month for it (as an example, ChatGPT still doesn’t make a profit at their $200/month tier).

LLMs are too unreliable to make anything better than toys, but too expensive to sell as toys.

Insanely bad idea. You should not be honebrewing anything with the capacity to kill people.

Not that Tesla’s solution is any safer, but that’s a separate discussion.

This is the selfhosted community; Who are you training? In most cases there’s literally only one person who would ever need SSH access to this server. Maybe two or three in a tiny handful of cases, but anyone who can’t figure out Netbird in 30 seconds absolutely should not be accessing anything via SSH.

And you’ve clearly never used Netbird, Tailscale, or any similar service, if you think that update, maintenance and config constitute any kind of meaningful burden, especially for something as simple as remote access to a VPS.

We’re in selfhosted. If you have to bring up use cases that are in no way relevant to 99% of self hosters to justify your argument, you don’t have an argument.

A lot of things are “fine”, but the cost of adding Netbird to your VPS is effectively zero, whether counted in dollars, time, or convenience.

Given the massive security benefits of using a VPN, and the lack of any meaningful downside to doing so, you’d be an idiot not to.

And mine. Clearly one of us is better at it.

You seem like a fan of the “pull out” method.

This is the correct answer. Never expose your SSH port on the public web, always use a VPN. Tailscale, Netmaker or Netbird make it piss easy to connect to your VPS securely, and because they all use NAT traversal you don’t have to open any ports in your firewall.

Combine this with configuring UFW on the server (in addition to the firewall from the VPS provider - layered defence is king) and Fail2Ban. SSH keys are also a good idea. And of course disable root SSH just in case.

With a multi-layered defence like this you will be functionally impervious to brute force attacks. And while each layer of protection may have an undiscovered exploit, it will be unlikely that there are exploits to bypass every layer simultaneously (Note for the pendants; I said “unlikely”, not “impossible”. No defence is perfect).

It really doesn’t. You’re just describing the “fancy” part of “fancy autocomplete.” No one was ever really suggesting that they only predict the next word. If that was the case they would just be autocomplete, nothing fancy about it.

What’s being conveyed by “fancy autocomplete” is that these models ultimately operate by combining the most statistically likely elements of their dataset, with some application of random noise. More noise creates more “creative” (meaning more random, less probable) outputs. They do not actually “think” as we understand thought. This can clearly be seen in the examples given in the article, especially to do with math. The model is throwing together elements that are statistically proximate to the prompt. It’s not actually applying a structured, logical method the way humans can be taught to.

I’m not sure what would make you think the “customers” for an enormous DNA database were the people providing the DNA.

Those people were just paying to be the product.

So yes, I did, and yes, their docs suck (better documentation is on their roadmap).

There’s a really good guide here on Lemmy that I recommend instead. https://lemmy.ml/post/25006407

Following this I had it up and running in no time. Check the comments as well, I added some notes on getting attachments working. If you’re still having issues shoot me a message and I’ll try to help.

If you can, take a moment to upvote Drawing Support in their suggested features section; https://notesnook.com/roadmap/

I did find that with a very large OneNote account the importer struggled, specifically because OneNote was timing out and rejecting the requests after a while.

My solution was to backup (in onenote) and then delete the notebooks that had been moved already and then run the importer again.

Not even once. The syncing has been incredibly robust for me. It also has a really nice flow for handling conflicts.

Of course, it’s worth keeping in mind that it can new self-hosted, so experiences will vary.

I’m using the self hosted version. Take from that what you will.

Yes, it’s all open source and can be self-hosted. They run a paid plan, but if you self host then you get all the paid features free.

Yeah, the potential for real hazard to life and limb is very high here. This isn’t like fucking around with your IOT lightbulbs. This could kill somebody.

This is a good time to switch to Notesnook, which has a OneNote importer.

Why am I about to shill so hard for this particular app? Simple, because after Evernote enshittified over a decade ago, I switched to OneNote as the least terrible alternative, and then spent the next ten years trying to find an actually good, open source notes app.

Call me Ahab because this motherfucker has been my white whale for a not-insignificant portion of my life.

Notesnook, finally, hit everything I wanted;

• You can self host it (but you don’t have to)
• Self hosters get everything on the paid plan for free
• It has a web app, a desktop app, and a healthy ecosystem of phone apps, with - very importantly - 1:1 feature parity. Everything you want to do you can do from any of the interfaces and for the most part they’re even laid out identically.
• It has a proper rich text WYSIWYG editor. It does not demand you learn FUCKING MARKDOWN. JESUS H CHRIST I DO NOT WANT TO LEARN A FUCKING SYNTAX TO MAKE NOTES, WHAT IS WRONG WITH YOU?
• But for those who care about that stuff, it is built on markdown, and all your notes can be exported in markdown, so there’s no lock in. And you can use markdown in the editor (without even having to switch modes like a lot of other editors).
• Everything is encrypted by default. Notes can also be individually password protected.
• You can share copies of notes with optional password protection and self-destruction.
• It has a really slick UI. Everything works, everything is intuitive, there are tonnes of keyboard shortcuts. I find I actually have an easier time writing long form text content (such as a novella I’m working on) in Notesnook than I did in Word or LibreOffice.
• It builds a TOC for notes automatically. You can link notes to each other, and links are bidirectional so you can track which notes link to a particular note.
• You have sorting by both tags, and notebooks. Notebooks are infinitely nestable, and - this is really cool - notes can exist in multiple notebooks simultaneously.
• It has robust web clipper for Firefox and Chrome.
• Very robust attachment support.
• God so much more, I’m having to deliberately stop here.

What it’s currently lacking is drawing support. If that’s a must have for you, check out Joplin instead (at least for now, I’ve seen some talk about Notesnook integrating Excalibur for digital canvas, which would be a superb solution).

Anyway, please check out Notesnook. It’s excellent, and I like sharing excellent things. https://notesnook.com/downloads/