I know that everybody thinks the robots are the best things, but I don't think robots are better than humans because robots can't make robots. And if you just make a machine that can make robots, then that would just be absurd. And I don't know why you would do that even. That's my point.
Bye. And this is Shogi, and I am just telling you my observation of this. This is Brandon. Welcome to Concerning AI.
Welcome. Here we are. It's 100 degrees here. It is.
It's hot, important, Laurie. I noticed in our stats we've been getting some overseas listens a whole bunch. But yeah, so we're going to, you know, tonight when we say things like Portland, we might want to say things like West Coast of the United States here. So you know what we're talking about.
Oh man, international. We're international now. Yeah. And let's go.
Let's just say it could have been. Yeah. It still feels good to think that there are people around the world that are part of this. That's right.
So yeah, so that was my son Shogi. And he thinks absurd to think about robots making other robots. Now is it that he thinks it's absurd that people are just as cool if not better than robots because we make more people or because we make robots? I got the feeling that it was because we make robots.
We're their creators. We're their creators. And so how could they be better than us? That's a good question.
It's a philosophical conundrum I think. So yeah, he says that robots can't be better than people because we make robots. That's fascinating. That's an interesting thing.
I mean, it seems obvious, doesn't it? That's correct. It seems obvious. Yeah.
He's seven. I don't know how many adults believe that many I would assume. I would think so. Most.
Well, yeah. I mean, are we better because we make robots? I think at this moment that's probably true. I have.
I mean, better is also a value judgment. But I think in terms of hierarchy, I think humanity has always thought that it was at the top from the time we were able to think that thought we had thought that hierarchy. I want to make a bookmark on hierarchy. We'll come back to that.
Yeah. So I think if you were the better, it's better means we're on top. We're sort of the king of the food chain or whatever, the top of the food chain. Right?
We're the super predator. Yeah. Let's just say we're the most dangerous to life on earth at this point. So clearly we have to be the best if we're the most dangerous.
Yeah. And we've had shows about what is intelligence. And we had a hard time figuring that out. We tried to put the definition on that.
And we tend to think that humans are the most intelligent species that we know of. Yeah. I still feel that way. Yeah.
Based on the way that we know how to measure intelligence. We know that we're missing a whole bunch. It's entirely possible for sure. But in the observable universe, the creatures that we know how to interact with, we seem more intelligent than them.
Definitely of things that think at a similar rate to us. Yeah. Because who knows? Maybe trees think really slowly, but think awesome thoughts.
But we would not be able to keep up with it or even we just can't. That's not a conversation or a thought pattern we could follow. Right. And so if we then put all non-biological processes into this mix as well, we wouldn't call them people yet.
And we would say we're more intelligent than they are. They could be tiny things that we don't know about that think way faster than us for whom we are mountains. But again, they're not on our detectable range. Right.
But I was thinking about the ones that we've built, any intelligence that we've built. Yeah. Yeah. Yeah.
Yeah. Yeah. Yeah. Yeah.
Yeah. That's a weird thing to say better human beings. Let's say. Well, yeah, sure.
I mean, intelligence is not like calculator can do math a lot. That's not what we, I don't think that's what we, I mean, I think it's the, well, you're trying to trap me. No, I mean, I'm saying that right now, today there are certain narrows. We're not the best at go anymore.
But they're more intelligent in that. They're more intelligent. So does that make them better? I mean, no, it's not generally intelligent.
So yeah, I think that's the thing, general intelligence, that's what this show has been all about. The narrow intelligence, I guess maybe I would ask the question was that making this statement, is it possible that we could be in danger from a narrow intelligence rather than a general one? Sure. A narrow intelligence that also has built in the ability to learn and therefore maximizes a function.
I mean, a paper book maximizer could be a very narrow intelligence. Well, I mean, we could clearly be in danger. I don't know if it would be existential danger, but we could clearly be in danger just even from my, you know, one of these financial intelligences that goes off. Yeah.
Actually, we had recently there's some thought that that was caused by algorithms interacting badly with the sort of failsafe built into the system, you know, as an unexpected dynamic that it happens. Could we crash everything? Could we, you know, accidentally with a narrow financial intelligence? Sure.
But that's probably not an existential problem. No, if we crash the little financial system, we would recover from that relatively easily. Yeah. I can imagine more than decades.
Very worse. We went through the bank, you know, around the banks in 1929, had the great depression, and that was pretty bad for a lot of people in the US and around the world, but we recovered from that. We could have a much, much worse scenario now, but it still wouldn't wipe all humanity out. It could be horrible.
I mean, it could really crash. I think it's possible we could really crash away, but it just wouldn't wipe all the people off. Yeah. And there'd still be a lot of infrastructure and knowledge and things to build from and a lot more space.
Right. So we really want to build the robots and therefore some of us, by some measure, we're better than the robots. For now. Yeah.
For now. Yeah. But it is absurd to think that robots could build robots or machines could build machines, but that absurdity may soon be with us. I think so.
Yeah. As I was thinking about this question, about machines building machines, I was thinking about what we talked about last week around feature design, right? That we have a lot of that paper. Yeah.
So if you want to listen to last week, you'll go ahead and listen to what you want to, but we were talking about features, feature design, features are everything. And you were sharing that if you were trying to just recognize speech, you have to decide. You have a data stream and we could look at five second clip and say how much energy is there, particular frequency, or you could pick five seconds and just look at the waveform and then move over 10 milliseconds and look at the waveform again. And I was thinking about that thinking, this is an absurdly difficult problem.
I'm glad we were able to bring absurd into the show often. Yeah. Word for the week. So it's me, like it's absurd.
Like the problem space is so large and our human brains are so feeble that this question of feature design feels impossible. Well, clearly it's not impossible because we are doing machine learning today. This is one of the areas where a massive breakthrough is possible, I think, in terms of actually trying to automatically learn features. We don't know how to do that really well now.
But that is a, and when that happens, that's going to make a big difference. So let me ask a question. If I'm a new researcher, clearly I don't know anything about this stuff. Like I'm the novice here.
I don't know anything about this stuff. But let's say I was a graduate student of learning, machine learning, and I would learn some techniques. I would learn from what's been happening before. I would learn from choices that the pioneers in the field have chosen before me.
What features sets do they choose? Or how do they design features? And then were they good? Were they bad?
Although it's probably easier to learn from the ones that were good. The ones that were bad, I understand, biased. They don't publish negative results. Yeah, for sure.
And so that's a difficulty that we may have to repeat and learn over and over and over again in the negative. It seems like a good idea. I wonder why nobody's ever said that it's not a good idea because it's hurting and work for us. Well, nobody wants to publish that.
But it would be really useful. I actually made a blood post about this very issue at one point thinking about in a company, I would consult for companies in a company. It would be great to have everybody publishing their negative results. So I don't have to keep repeating them.
Well, that's kind of like the lean startup approach anyway. It's kind of a little bit about that. Do a quick experiment, document it, say what you learned. Yeah, so I was thinking, what if there was an AI that could do that for us?
That could be designing the features that we were using from some other problem. Does that break through your speech? That's the one I'm speaking on. I think that that will be, you know, I think that we've had some kind of slow progress on that for on that for on.
But it's still an area of black art. And when it's not, that's going to be massive. When it's actually part of what is learned. But I have no idea how to do that.
We don't have it. I don't think we have any idea how to do that at this point. Right, but that's exciting because you could imagine that we could fly away. Yeah, absolutely.
It doesn't seem, it doesn't seem, yeah, it seems like a positive feedback process too, where, you know, this is kind of the intelligence explosion sort of thing, where you start to learn features that allow you then to learn more sophisticated feature learners themselves. Right, right, because if you can take the feature design process that you use in one problem and use it another problem and it actually generalizes, I would guess it for decades. None of them generalize. We just had to start from scratch every time.
But as soon as we start finding something that's at least a little bit generalizable, then everything can kind of ramp up from there. Yes, that's maybe so. Maybe so. Should we switch this topic too?
We have two topics right today? Yeah, here are the king's segues and there's one right there. There's one right now. Well, I've been holding this, my fingers in a diamond shape here for a while and so I've been a little distracted by that.
So this diamond shape means to two things. It is hierarchy and it also is like the delta symbol for change. So that brings us all around to the event that I was at this weekend, which was awesome. Okay, it's awesome.
So the event I was at this weekend was called Shaping Change and it was sort of organized by I would say the Octavia E Butler legacy network. Okay, so for those of you specular fiction fans out there, you may have listened to our show about speculative fiction. I'm trying to use brain and highfalutin languages that are saying sci-fi. I appreciate that.
He gave me such a hard time when I used that. Yeah, I appreciate it. It is nice to sort of appropriate names that have been used as spiritual stuff. If you have, I'll start using sci-fi, you start using spanking fiction and all will be right with the world.
Yeah, if you have missed any, you know, if you have missed Octavia Butler in your exploration of speculative fiction, you really miss an outline. Oh my god. Yes. Rectify that post-haste.
Yeah, she is probably the most influential author, period for me. Yeah. Period. Like any genre, any thing.
And one thing you might notice if you look at speculative fiction writers is that they tend to be a certain you. Yes, and they have a certain anatomy. Yes. And she doesn't look at the mold.
No, she's a woman of color. Yeah. Yeah, an introvert. Yeah.
And just an amazing mind. And I'm just so thankful that I've gotten to read her. Anyway, so the reason that we thought we might include this as another topic is because the name itself, right? Shaping change has a lot to do with intelligence.
Right? That is sort of what we decided generic intelligence was, was the ability to do future theory, which is shaping change. Yes. And shaping change comes from a series of books, one of her series of books that was about like a religion that was created where the central precept is God is change.
Yeah. And I read that book. I was influenced by a lot or the two books, the parables of the sower and the parables of the sower and the parables of the touses. Those are the two books that really influenced me.
I didn't like that part. That part didn't speak to me at all. Yeah. I've had a hard time with it too.
And I've gone back recently and read them again. And I still have a hard time getting myself in that mindset. I still don't know how I feel about the language. I don't really like the language.
But boy, change has been at the front of my mind since we've been doing this podcast, especially because we've really asked two questions on the show. This show was about two questions. One is, should we be concerned about the future of humanity when we think about AI? Right.
Is AI an existential threat? And we've not been able to rule that out at all. It's not yet. No, but we want to then ask the second question.
The second question is the one that's about it. Right. How do we deal with this massive change and the uncertainty around it? And I think what I came to the...
I'm kind of narcissistic. I like to put myself at the center of the universe, the center of things. Like somehow I'm going to be able to like shape things. Like this AI thing.
I'd love to tantalize myself with the idea that somehow I'm going to make a difference on this. Right? Yeah. And what I really came away from this weekend was I was just blown away by all the amazing awesome people who are doing all this incredible work.
And in an area where I have kind of put myself at the center so far of my experience of this area and kind of recognize that maybe the role for me going forward with this community, at least, Octavia, those who gathered around Octavia Butler, is not one of being in the center, but is more one of finding a way to be supportive from the side. And that got me thinking a little bit more even about like, you know, we are trying to figure out how to deal with uncertainty. Us two privileged white males who have really not had to deal with a lot of uncertainty in our lives. Right.
Really. So maybe we should be looking to communities that have had to deal with a lot more uncertainty over the years and have come up with solutions and ways of dealing with those things. Yeah, maybe that's where the wisdom is. I wonder what it would be like to actually ask for some advice from these kind of luminaries in this community.
They're super nice and super, super helpful. I bet you could get some help. Future is uncertain. Very uncertain.
Yeah. And what do you do with that? How do you live in spite of the fact that you are not in control of... In the cards, they may be, they may appear to be, you're both of stack against you against your whole community, against everybody you know.
Right. And so certainly we have groups of humans that have experienced that to a much greater degree than other groups of humans. Right. Yeah.
And in this world of, you know, this future of AI might be all humans. Although it could be at different stages affecting different groups of different ways. For sure. Yeah.
I have an idea from our conversation where I was sharing with that about this super, super cool event that I was at this weekend. Hopefully look forward to over the next weeks of the months, maybe getting a guest or two from that community to talk to us about, you know, living generatively in the midst of a life that is uncertain and where, you know, maybe power is not handed to you. Yeah. Awesome.
I'm thinking about future topics and thinking about, you know, you listeners out there, what would you like us to talk about? You know, what are your ideas? We have swum various directions at different times. We've tried to get really geeky with academic papers and really digging into some technical stuff.
We've done pretty far out in a limb in terms of philosophical stuff, what is consciousness and what is intelligence and those kinds of things. We've tried to speculate in some ways about the future. We've looked at fiction. We've looked at a whole bunch of different ways.
And now we're considering this, like, how do you, you know, how to live with change? Yeah. The massive change on this area. You know, what would be interesting to you, I actually read across the paper this week from the DeepMind folks.
It was from DeepMind and it was from Miri, I guess. And it was about the machine intelligence research institute for those of you who are. It was about the idea of a big red button. Imagine you're a researcher and you're training the AI and at some point things go off the rails and you want to be able to push the big red button to stop the thing.
How do you not have that process, because how do you not, how do you create safety in that process? Well, so I got this paper and I was like, wow, I don't understand. I don't understand what about this. This is so, you know, I sort of felt like I did last week when you were trying to talk about walking through the thing about learning.
But anyway, so there's academics that we could try to dig in on and, you know, that would be interesting to folks. And if you're an expert in an area like that and you want to be on the show, that would be fun too. Yeah. Go to Facebook, like concerning eyes, the name of our group and you can get in touch with them.
Make some suggestions. Suggestions about what would be interesting is we think about. Yeah, yeah. Give us answers.
We love answers. We mostly have questions. So we'll share them. We might have to share them in the form of questions.
Whoa. Like Jeopardy, right? Jeopardy. I get it.
And deep mind is, I think it's really good. Jeopardy, or is it a difference? Lots of them. I got to get my AIs right.
Yeah, what is Watson? I think so. Ah, you got it. Yeah.
Alright, so we'll see you next time. Super. Bye.