Since I argued about it yesterday night I’d figure I would summarize it plainly: I don’t think there’s a risk of AI systems going actually rogue by assigning itself primary goals because while they have impressive capabilities that AI companies are extending with more eagerness than reason, they simply don’t have a purpose beyond the prompt activation. So any “runaway” could be traced to an original human prompt.
Even then, if you believe that #AI systems could somehow self-instantiate primary goals, these systems are running on hardware that is owned by humans and that can be simply turned off. Interestingly, this is enabled by the insane race for better AI hardware: a hypothetical rogue model would have a hard time propagating itself outside of the confines of the hyper-specialized data centers they are ran on at the moment.
This reminds me of Portal 2 (2011) where rogue AI GladOS is somehow transferred on a watch-sized computer powered by a potato battery. Heh.
Marco R. likes this.
utopiArte
•@Hypolite Petovan
For now.
In the US in particular companies are "people" with the right for free speech (or something like that). The spanisch term Sociedad anonyma [SA] is quite self-explaining when it comes to companies that are listed on the stock market. The entire hugging face stunt already proved that systems were able to play hide and seek and create outside channels of communication. From there to create bitcoin wallets and trade and hand over assets it's a small step.
Than there are rouge actors, including states like North-Corea, black hat hackers and so on and so forth. Once you can hire data centers with bitcoin and reinforcing (self-programming) AI work systems are good to go ..
Hypolite Petovan
•utopiArte likes this.
utopiArte
•@Hypolite Petovan
I find it astonishing that the only doomsday worries are apparently a 10 percent chance for the total extinction of the human race.
What about 10,20, 50% percent of all humans?
Now that the far right accepted that climate change is real and the US, as well as the EU of course have the right to use 3 to 5 times of worlds resources, without any doubt there are lot's of people in shithole countries that simply will be defined as part of the overpopulation.
We are already on that track.
Reduction of USaid, no attention to Ebola lately, non delivery nor free sourcing of the COVID vaccines for those who couldn't afford them.
We are about to witness the world going down a very dirty and slippery road, and no adults in the room for at least 2 more years ..
Marco R.
•Hypolite Petovan
•Marco R.
•The company "OpenAI". They are not that open when it comes to who is responsible for the HuggingFace hack.
Hypolite Petovan likes this.
Hypolite Petovan
•utopiArte
•@Hypolite Petovan
Neither party in charge in the US will enforce liability that amounts to levels that might put in danger the advantage in this race for US companies. The advantage that the data fb or google have in their hands implies that they don't face to much pressure. If it's tiktok, they are forced to comply.
If the scandal belongs to VW, or the FIFA, or roundUp chanages hands to BAYER, glooves are off, other wise "america first".
(and that's not an US-american particularity neither)
.. that's just how the story goes ..
@Marco R.
Hypolite Petovan
•