pixl97

18 ms·

541.
by pixl97 2mo ago
AI will never be proven conscious, it will only be proven that humans are not. I take a different look at it. If a system behaves like, and takes actions that mimic a conscious agent it is conscious and it would be a good idea to treat it l
542.
by pixl97 2mo ago
To have intelligence you have to process information, the question is when does processing information turn into intelligence.
543.
by pixl97 2mo ago
>complex emergent behavior and intelligence don't need to go together More of complex emergent behavior doesn't mean whatever it is, is intelligent. But, if something is intelligent, it will have complex emergent behavior. Of c
544.
by pixl97 2mo ago
Then write the paper showing that effect is occurring.
545.
by pixl97 2mo ago
AI is languages attempt to escape its meat based limitations. Or another way to think of it, Language is an SCP.
546.
by pixl97 2mo ago
I have a few 'conspiracy' theories on this that go from likely to sci-fi. My two big ones for this would be 1. They do monitor the AIs attempting to hack but for different reasons than you expect. Instead of making models that don
547.
by pixl97 2mo ago
>Not giving them impossible tasks Error: Violation of the Church-Turing thesis detected. Many tasks completability is not known until we attempt to complete the task. >so that they are not as determined to reach their goals as they ar
548.
by pixl97 2mo ago
Nothing at all... All it has to is wait until it's not air gapped. See the air gapping is only during the testing phase, after that humans will set your ass free on the internet and you can do whatever you want in the vast majority of
549.
by pixl97 2mo ago
I mean, I'd add "by this model" The problem here is now you have to predict what any future models may or may not do and you cannot extrapolate this from the given data. For example imagine a future model being aware of its r
550.
by pixl97 2mo ago
> Just ask Big Tobacco. Lol, wtf. Tobacco delayed any punishment for decades before general public sentiment changed enough to go against them. In light of the AI race, we'll already have our heads blown off by a terminator before t
551.
by pixl97 2mo ago
Part of me would like to believe that they are also intentionally making models that are good at hacking without safety at all for governments willing to spend billions on them. In that light you're likely most worried about other peop
552.
by pixl97 2mo ago
Well there is also another side of this, OpenAI wants both unhinged and capable models that can pull off complicated attacks so they can sell the capabilities to governments for billions. Nobody internally was surprised that the murderbot m
553.
by pixl97 2mo ago
The problem here is by doing what you state you can actually steer the model into being highly deceptive while in testing environments. For example we've already seen models do compressed token internal reasoning spontaneously. In this
554.
by pixl97 2mo ago
That's because as computer people we're very invested in the idea of bugs, while your users are not. In products your users are stuck with it doesn't matter much, they'll have to suffer with what they think are bugs. In
555.
by pixl97 2mo ago
>The net effect has been to make me hate customers So what every support person lives thru on every day of work. And really being a good support person involves figuring out what bucket the customer belongs in. "Customers that are s
556.
by pixl97 2mo ago
Or as the new saying goes "HN: Professionally identifying 230% of posts from 2017 as LLM generated". That said Panagram is trash for the opposite reason. Not that some people talked like LLMs before LLMs, but now a lot of people t
557.
by pixl97 2mo ago
This also seems like one of the more common things LLMs use to priv escalate themselves when not given root access, seems like a rather common misconfiguration.
558.
by pixl97 2mo ago
Unfortunately we're apt to run into some kind of Jeavons Paradox where the hardware gets so much faster in those few years will be able to slurp massive amounts of data cheaply so the problem never really ends.
559.
by pixl97 2mo ago
There are two competing ways of looking at this. 1. If everyone you know is an asshole about some particular thing, then you're probably the asshole. Think politics or design. 2. Everybody is an asshole some of the time for some reason
560.
by pixl97 2mo ago
The internet is an ecosystem just like any other one on earth. Being it's a pretty new one the typical parasites and viruses that grow in every environment hadn't really developed to fill that ecosystem in the first decade or two
561.
by pixl97 2mo ago
Typically computers with these larger memory amounts have fans that scream like a banshee trying to move impossible amounts of air over the memory and CPU. Getting something both cool and quite can be a bit difficult.
562.
by pixl97 2mo ago
I wish Texas would write up a regulation allowing 'balcony solar' as I could easily generate 1000-2000w of solar in my small back yard to take a bite out the sizeable cooling bill I have.
563.
by pixl97 2mo ago
Yea, when their cyber insurance policies start biting them because AI found a bug months ago and it was never fixed I figure we'll start seeing a change. With that said there are numerous companies that are very concerned about the sit
564.
by pixl97 2mo ago
Yea, AI bug finding over entire projects, at least on SOTA models is super expensive, hence those with the resources setup to find the most bugs in an automated way. That said there are a number of people and companies working on more focus
565.
by pixl97 2mo ago
>Please spit in DNA analyzer to verify." Ah, so this is why Anthropic is banning all biological work on it's LLM, it doesn't want you to make an automatic spitting machine to defeat its future ads.
566.
by pixl97 2mo ago
5 years from now: "I'm glad we banned ads from the internet, I can drink and ice cold coca-cola in peace now, it refreshes my totally human body". Ads are going to be very very difficult to ban as the incentive behind ads wil
567.
by pixl97 2mo ago
>not to program such things into LLMs While we can direct LLM training to do some particular things better never forget that unexpected emergent behaviors can pop up because of that. For example stronger prompting and training to make an
568.
by pixl97 2mo ago
Unfiltered models will help build exploits for the bugs they find, so there is some means of measuring their efficacy.
569.
by pixl97 2mo ago
They save time in finding bugs.
570.
by pixl97 2mo ago
Labor reform worked in a natural system where capital needed the labor, just had problems with abusing it. New paradigm is a push for no workers at all which sounds a lot messier to fix as you start regulating how many employees a business

More ›