ASI Explained: Is Artificial Superintelligence Exciting or Terrifying?

James Dooley (0:00)

Artificial Superintelligence. Today, I'm joined with Kasra Dash, and we're going to be looking into ASI, how close we are away from artificial superintelligence, the controls and the risks involved with artificial superintelligence, and whether you think is a good thing for it to come or whether you don't. So Kasra, first of all, can you just define for anyone who's watching this, what is artificial superintelligence?

Kasra Dash (0:31)

Yeah. So let me start off with artificial general intelligence, and especially if people haven't watched our video on that. So artificial general intelligence is when the AI is basically on level with just regular humans from an IQ point of view and from a smartness point of view.

Superintelligence is when it's massively surpasses our own intelligence on all fronts. So from a biology point of view, from a coding point of view, it's just way smarter than us. So it can actually do a lot more tasks and the speed is just way better than what we would be able to do, and also the process as well of how quickly it can actually come back with certain answers as well.

James Dooley (1:19)

Yeah, I mean, artificial superintelligence is going to be absolutely, it's going to be a game changer for the whole world, for humanity.

I want to talk a little bit about on the artificial superintelligence, where obviously it might be able to come out and cure cancers, there might be other ways for like biohacking, extending life and all things like that. Like what's your thoughts about artificial superintelligence? Are you excited by it or scared by it predominantly?

Kasra Dash (1:51)

I think it probably comes down to who controls it.

I know that people will say it's uncontrollable, but it probably comes down to who creates it first, because it is almost like an arms race. Whoever creates it first will actually end up being it. It's almost the leader, and that's the scary thing for me. So a couple of things on this.

Even currently right now, when you ask ChatGBT to code you something, it codes in a certain way. If you ask ChatGBT to code you a tool for extracting query fan-outs on ChatGBT, it actually doesn't do it. And if you already have a tool, like let's say you use Cloud Code to create, and you then give it to ChatGBT, it removes it. So there's all of these different nuances when it comes to currently how we use AI. And I can also see that when it comes to ASI as well. So whoever is almost the owner will have the, almost like a black shadow over it, where it's like you can't do certain things. So that's the scary thing for me, whoever is actually the original creator.

James Dooley (3:02)

Yeah, so obviously on that there, they can get them to do certain tasks, but they might be strictly from doing certain tasks, it can do.

And who has that control to depend on whether their ASI agent is able to do things or not. It can be scary times depending in which hands that is, like what country kind of is the problem of the first one? Is it gonna be China? Is it gonna be Russia? Is it gonna be the US? Is it gonna be a European country? Like it's gonna be scary times, because whoever kind of develops that artificial superintelligence first, he's gonna have, like you said, the control over that moving forward. How far, like I believe we're already at AGI, so I believe we're already at Artificial General Intelligence. I believe that yes, okay, there's times with the context window it can make mistakes, but so can humans. So I feel like we're at least as good as humans, if not slightly surpassed it. I won't say we've vastly surpassed it yet. I don't think it's on its own say so, self-healing and self-learning. I still feel you need to give it certain agents to say, OK, once you've got a swarm of agentic and autonomous agents, I think we're almost at ASI, but only because all these different agents are talking to each other and synergizing and having two great people locking horns and working together is almost like we're at artificial superintelligence. How would you explain where you think we are at present with regards to AGI versus ASI?

Kasra Dash (4:34)

I think currently we're tiptoeing on a very thin line of AGI and being autonomous. And then in some cases, we take one step back and just Cloud Code or ChatGBT is just really dumb. So I think we are getting to that. There's certain things that AI does that amazes me, but then there's certain things that AI does and I'm like, why are you thinking like that?