2 ай бұрын
2 ай бұрын
2 ай бұрын
2 ай бұрын
2 ай бұрын
Пікірлер
@danielroberts3484 11 сағат бұрын
Claude 3.5 would like to have a word
@lordalmightyjr 22 сағат бұрын
Man so we’re all just gonna be sitting around hoping these absolute beasts just stay friendly and happy and don’t rinse us at any second. Cool boys
@backgroundcharacterj5071 23 сағат бұрын
Ok but hear me out how cool would it be to build a giant robotic snake
@jakemeyer8188 Күн бұрын
When I'm like, 90 years old, I'm going to be in one of those generations that people will say, "wow, the changes they've seen in their lifetime is amazing". For context, I'm 44. I grew up before cell phones and internet, but will die in the age of Artificial Super Intelligence...
@person737 Күн бұрын
Ilya launched a new company to deliver safe super intelligence, and that only and anthropic just launched claude sonnet 3.5.
@bladestarX Күн бұрын
I think the hype around AGI has a lot to do with keeping everyone engaged, and maintaining billions in investments. The truth is that most people and companies have yet to start implementing what we already have. A pause or slowdown in AI development is probably a good thing, as it allows for a shift in focus to using and implementing existing capabilities rather than researching more. I am not saying to intentionally slow it down, but it would not be that bad. We probably have enough to spend years developing tools and ways to scale and implement what we have.
@ollep0lle Күн бұрын
That was very interesting, thank you! Also very beautiful scenery.
@sadscientist9995 Күн бұрын
when you hear about "safety" just understand what it really means is control.
@benjaminallison4644 Күн бұрын
Why would we need more data? Humans and other animals learn by being in the world. Clearly the algorithms have a ways to go.
@neorock6135 Күн бұрын
The minute we have an AI that is able to write code *better than us,* irrespective of its abilities with other tasks, the entire paradigm shifts, no? I realize algorithmic gains are only one branch of the tree, but the efficiences gained with near perfect coding as well as the resulting recursive self improvement, has to stand a great chance of exceeding the gap between what we have now & AGI. Of course this is all entirely dependant on us getting to an AI that can code better than us. However, considering AI can already code, current models reaching that summit in the foreseeable future, seems more likely than not.