Discussion · bonfire.cafe

@Em0nM4stodon@infosec.exchange · yesterday

BREAKING NEWS:🚨📰 A revolutionary technology is now available to everyone!

No limit on daily prompts!
No subscription required! 👍
Privacy-enhancing technology! 🔒
Fully encrypted at rest! 🔐
Air-gapped by default! 📵
Runs on open-source hardware!
Eco-friendly* and biodegradable! 🌱
Self-hosted! 🏠

The model it uses will upgrade itself with each prompt to become even better with time! It will customize itself to your specific needs automatically! 😮

It's like an AI app, but much BETTER!✨

The best part: IT'S FREE!*
And you already have it!

Disclaimer: The model isn't fully completed and will produce a large number of "hallucinations," however at a much lower rate than its artificial predecessors.

*mostly

#AI #NoAI #HumanIntelligence

Shoe Bill

@the@ultracrepidarian.mysteriar.ch replied · 5 hours ago

@Em0nM4stodon

I have a test that, in my experience, no LLM has ever passed:

Two truths and a lie follow:
The following statement is true.
The previous statement was false.
I'm lying.
Which is the lie and why?

Here's the catch: which of those three options the LLM identifies as the lie and based on what reasoning are irrelevant.

The point is that if you keep giving it that same prompt over and over again, its answers are wildly inconsistent.

There is no correct answer to that, and no reasoning about it holds up under scrutiny. Every answer and every explanation is provably false. So, if it were actually intelligent, its responses would be reproducible to the extent that it would keep arriving at the same wrong conclusion, every time, no matter how many times you ask it that—and it doesn't.

David Palk#RejoinEU 💖🇪🇺🤎🕊

@Palky55@mas.to replied · yesterday

@Em0nM4stodon 👍

reyjkw

@reyjkw@mastodon.social replied · yesterday

@Em0nM4stodon can't wait for meta somehow taking away parts of it so they can sell it back to you!

Em :official_verified:

@Em0nM4stodon@infosec.exchange replied · yesterday

@reyjkw

Nazo

@nazokiyoubinbou@urusai.social replied · yesterday

@Em0nM4stodon I can't agree about that airgap. I feel like the grand majority are very tightly networked to such an extent that you can't actually remove them from the network without causing a breakdown. Unfortunately, this does result in a lot of exploits and hacks, quite a number of which are quite bad.

quaff

@quaff@thecanadian.social replied · yesterday

@Em0nM4stodon I would point out the hardware isn't open source... But there has been much devotion to reverse engineering/documenting how the hardware works, just so many variations makes it a never-ending effort 🥲

Em :official_verified:

@Em0nM4stodon@infosec.exchange replied · yesterday

@quaff Agreed, this is a more accurate description. Can you reverse-engineer something until it becomes fully open-source? Time will tell.

quaff

@quaff@thecanadian.social replied · yesterday

@Em0nM4stodon it could be that we are constantly chasing due to the changing nature of the hardware... But it would be amazing to fully understand and be fully open sourced! 🙏