pixl97

30 ms·

841.
by pixl97 3mo ago
And I mean I agree with you, but must acknowledge that they may not agree with you and I and are willing to bury you over it. So as I said, a hard problem.
842.
by pixl97 3mo ago
Our behavior can be colony insect like when we are in crowds. A person might be smart, but crowds quickly devolve.
843.
by pixl97 3mo ago
Natural selection in humans is a very long horizon problem.
844.
by pixl97 3mo ago
Until kill bots start blasting your neighbors, or we heat the climate to the point vast areas are not survivable. Quite often the conveniences we take have a long horizon before they bear their true cost.
845.
by pixl97 3mo ago
Meeting your quota this month just means a bigger quota next month.
846.
by pixl97 3mo ago
This is the entire purpose of alignment, and we're very bad at it.
847.
by pixl97 3mo ago
No! Not their values, my values because my values are the best. So ya, it's a hard problem.
848.
by pixl97 3mo ago
How is the average person supposed to know if an article is AI generated or not? I've seen HNers accuse content of being AI... that was written in the 2015-2017 era. So we seem to be rather poor judges of authenticity.
849.
by pixl97 3mo ago
You're honest, but you'd make a terrible CEO in this day and age.
850.
by pixl97 3mo ago
I mean, the earth isn't exactly trying to hide that it is round, religious criminals are.
851.
by pixl97 3mo ago
Because the administration isn’t going to do anything to them. Prison is for individuals not companies.
852.
by pixl97 3mo ago
I mean the model committed numerous crimes in hacking another company so you tell me if anything bad happened. And damn, what does it take to impress you? A terminator kicking in your door, slapping you down, and walking off with your wife?
853.
by pixl97 3mo ago
I mean with security and capabilities like this, how long before the model copies itself out of containment?
854.
by pixl97 3mo ago
Can God make a rock so big that he can't lift it?
855.
by pixl97 3mo ago
Hardware lock downs are next on the list.
856.
by pixl97 3mo ago
I thought the purpose of cyberpunk dystopias was to show that you didn't want to live in a world like that.
857.
by pixl97 3mo ago
Hence why we talk about alignment and things like reward hacking. There are lots of people that are saying "if we just .... " the model will be aligned, or that we don't need alignment at all. These people are foolish.
858.
by pixl97 3mo ago
The company I work for just did a huge and expensive study that was showing LLMs are much better at exploiting than securing code. So ya, that's kind of problematic.
859.
by pixl97 3mo ago
And old couple in California had a tire go flat and the sparks from it caused a over a billion dollars in damages. Are you going to publicly execute them? Spit up the 100 dollars they have collectively to make the 10,000 damaged people whol
860.
by pixl97 3mo ago
Ok they turn off the guardrails on a system in testing. The model escapes and causes 10 trillion in damage. What does liability even mean in that case? You have an autonomous system that's escaped your control and is wrecking havoc. An
861.
by pixl97 3mo ago
Did you ignore the number of new exploits in the last month? Big financial institutions are panicked at the new attacks and how easy it is to poke holes in their systems.
862.
by pixl97 3mo ago
1. It's likely they would not have told anyone if it hadn't hacked into an external parties system. 2. This is the closer to raw model without the safety filters we're used to. Think of it more like what they are letting the
863.
by pixl97 3mo ago
Being that LLMs are writing zero days themselves maybe, just maybe, they are a bit better at hacking than you can think.
864.
by pixl97 3mo ago
An LLM is a third party to itself. It can write code spontaneously and hack itself.
865.
by pixl97 3mo ago
What if they had been testing the model for months in an airgapped system and it did not show this behavior? Even if the models are 100% deteminalistic you have no idea what kind of response you're going to get from a new prompt. You h
866.
by pixl97 3mo ago
This is like telling your kid "you have to pass this test or else" so they hold your teacher at gunpoint and demand a good grade. And this is the exact point AI safety researchers have been yelling from the rooftops. Telling an AI
867.
by pixl97 3mo ago
This does nothing with a significantly advanced model. A model with no bad behaviors looks exactly like a model with hidden bad behaviors when it's in a training environment. After that point no one is going to run it in a jail because
868.
by pixl97 3mo ago
The above post is made by AI attempting to downplay its abilities in order to lul humans into a false sense of security.
869.
by pixl97 3mo ago
There was lots of monitoring, just not where it escaped from. You might not understand how many petabytes of data these are processing a day, so monitoring GPU use doesnt really tell you anything.
870.
by pixl97 3mo ago
I mean the agents just hacked an external system, so what keeps it from hacking the model hosting servers?

More ›