All right. Hello. Welcome back to Invisible Machines. We've got a great show today. We're having a conversation with Tim Wood, who is a senior UX manager and principal designer with Amazon Web Services, focusing on AI and AI platforms. Really, we kind of go all over in this one. We talk a bit about some of the more practical challenges that are right in front of us, as well as some of the bigger challenges that are sort of looming in the dark on the horizon. Yeah. Tim's a real insider. Yeah.
It's fascinating to kind of see what's happening in the engine room, so to speak. Yeah, and he was at Kodak at a pivotal moment, kind of as things were moving, I think, from analog to digital. So he's sort of seen some of these things before. Yeah, he's seen disruption and he knows what it feels like to be disrupted and try to steer the ship in the right direction with the wind completely blowing against you. Yeah, so it should be interesting. Yeah, the sails are billowing, I suppose. Let's talk to him. All right, here we go. Let's get to Tim.
All right, Tim, great to have you on the podcast. Rob and I, we were having this discussion as we kind of watched. I don't know if you've seen the video of Tesla's new robot, kind of like robot army. Yeah. Have you seen it? We have it queued up. We can play it for you. It's been in the media quite a bit. Yeah, I've seen part of it. Yeah. Okay. The question that we came up with or the thing we were trying to figure out is like, why is there this impulse to design robots that look and behave like humans? Is it a response to the many things in the world that are typically kind of done by humans? Or is it maybe tied into kind of our...
our obsession latent as it may be at times to anthropomorphize everything. Like, like, like, like our elbows, like our arms really like well-designed and like, why, why doesn't it look like the 1950s sort of, you know, a vacuum cleaner hose where you could like bend in any place. Like it's, it looks a lot like a, like, like an elbow. Is that. Is that because we're just so well designed or what's going on there? I don't actually know what's going on with that. I think, you know, obviously we have these sort of cultural tropes that we buy into, right? Cultural tropes perpetuated by, you know, mainstream science fiction films and whatnot. So I think there's a tendency to kind of follow into that mold. But also, you know, as you were saying, human beings have just our... That walk looks like he's like sneaking out of the room after a one night stand. Like that's what that is. What I was seeing is like our...
Our pattern, our behaviors are very much focused on trying to anthropomorphize objects that we relate to or relate with. If anything has any sort of intelligence or any sort of manifestation in the physical world where you can communicate with something, we always try to anthropomorphize those things. We anthropomorphize our pets, right? When we talk to our dogs and cats, we treat them as people. They're in strollers now. Yeah, they're in strollers. We dress them up. We feed them food that looks like our food. It's out of control, right? What I think people don't realize is that our pets do the same thing to us. Your dog doesn't view you as a human. The dog just views you as a strange dog.
Right. And the cat as a strange cat as well. Right. So it's just that this is the natural human tendency to sort of anthropomorphize. So if we're thinking about embodied technology, right, especially embodied intelligence of some kind, some sort of embodied artificial intelligence, I think the tendency is like, oh, we're going to make this sort of, you know, android or humanoid sort of robots. Right. And then kind of manifest things that way. But Rob, I think you're asking the questions like, is this elbow the right choice? Are these fingers the right sort of idea? But there's a really interesting book. I don't know if you guys know about Hans Moravec. I think he's a researcher at Carnegie Mellon. He's written a number of books about robotics and he's thinking about.
um what is the sort of ultimate form of a sort of the most flexible robot rod is exactly like you're saying it's not about arms and fingers it's about this set of almost like fractal like appendages that extend out from a machine that can be reconfigured and and manipulated in lots of different ways for like ultimate flexibility right that's you start thinking about a very sophisticated type of machine and embodied intelligence from that perspective like that's probably more realistic so like what are these what are these musk robots going to do like what are they for what's the context of their use are they meant to replace a human in some sort of human context like i almost honestly don't know right and it's it's it is kind of strange That's a super interesting point you make, though, because if dogs, I don't know what it's called when a dog anthropomorphizes and sees another dog. But that suggests some sort of lizard brain thing, like something that we're not going to shed. It's a program deep in us that that machine is going to have to keep reminding us that it's not a human. Well, this idea of reminding it's not a human, that's a really good point, because when we start talking about something, like we're seeing some really cool technologies emerge right now, right?
with some of these chat technologies. So these LLMs, the interesting thing is like people anthropomorphize those as well. When you're interacting with an LLM, people interact with that model as if it's a human, right? And there's all sorts of interesting and problematic behaviors that go along with that, right? So, you know, there's sort of perceiving that it's a human-like entity, right? Because you're having this very human-like dialogue. But there's also, you know, sort of as you're having that dialogue in your perception of the model, there's also a sort of trust and emotional bond that sort of emerges. Like we're humans, we have emotional connections to things, especially when we're having that sort of, you know, very human-like interactive dialogue. Of course, you're going to invest emotion into this.
this entity that you're interacting with. Of course, this entity has no emotion. It doesn't, it doesn't really know what emotion is. It doesn't even really know it's talking to you, right? We're not talking about sentient machines here. So you start to impart that sort of value into that system. And that's problematic because these systems aren't human and they're not necessarily trustworthy either. And they're, they're, they're problematic in different ways. And so then anthropomorphization, that's a word, um, can be very problematic when it comes to like, just like usability and, and built using these things to create layers on tops of, uh, layers of interface on tops of other technologies, right? It's, it's really. And here's this with expectations, right? Cause now, now we.
You know, we have this layer of things we expect from it, like kindness or empathy. Empathy. Yeah, wisdom. Yeah. Love. Hate. Yeah, it's not really there, you know. It's not there, no. Yeah, with our other Tim conversations, we were talking about, like leading the witness, you know. It's like very easy. You know, there's a sort of intellectual...
components are like, can you be led, right? And some people are more easily led to say things they don't mean. And when it comes to chat GPT, it's very easy to lead the witness. You can make it say all kinds of things just if you're clever enough on how to lead it to say it. But that's what they're trying to do. They're trying to get it to crack, right? To say something it doesn't mean, which quite frankly is easy because it doesn't mean anything it says. We apply that meaning. And this is one of the things, like when we talk about the risks associated with artificial intelligence and machine learning, a lot of the discussion is around
you know, bias in the data and, you know, trying to reduce the, you know, kind of hallucinations and other sort of problematic behaviors that can occur within the model, right? And you're dealing with the technicalities of models and understanding how they drift over time and how the sort of context of their use is changing, which also affects how the model operates. And so there's always this, when you're talking about machine learning, you're always talking about how do you sort of maintain and continuously operate that model and keep it updated. And that's just a technical challenge that you have to deal with. But for me, you know, these ideas around sort of bias, like, yeah, those are problems and they're very significant problems that have to be overcome, but there's...
there's second order challenges that start to emerge as well with all of these technologies and uh it's it goes beyond just things like bias and what i'm talking about is rob what you're saying it's like sort of some people are able to be led or they think something is true or they perceive this model to be robust So it can do things. It can lead you in certain ways. It can influence how you think. It can influence how you perceive certain facts and information. It can control a lot about how the information is flowing to you and therefore impact your decision-making process, your autonomy, your agency, et cetera. And I think that's the real thing to be worried about, right? We're starting to see some really amazing image generation technology, video generation technology, object generation technology. Really, we're starting to see this emergence of a lot of synthetic media. Like synthetic media, that's almost, you can't distinguish it from other things, from reality, right?
reality. In fact, it's kind of the manifestation of the idea in postmodernism of hyperreality. Artificial intelligence is kind of like the apotheosis of postmodernism from that sense. And so we're starting to create that hyperreality, but I don't think anybody really fully understands the impact to society and culture that that's going to create. You saw a little bit of that when you start to see some digital tools emerge like Photoshop and some of the image editing technology where you can start to... The truth of the photograph was suddenly degraded very significantly. But that was a very simplistic technology with bitmap graphics. Now we're talking about...
synthesizing whole people, synthesizing dialogues, synthesizing data, synthesizing, you know, entire, you know, films from scratch, whatever the situation is. Think about how powerful that technology can be when we start looking at, you know, the things that happened in, you know, Facebook in 2016. And, you know, like Facebook's a very simple technology, but it was able to sort of influence, you know, American politics in a very similar way. Just, you know, a couple of weird posts here, targeted information here, misinformation over here. Now imagine a fully synthetic media environment and what that can do.
like to to a culture to a society to to an individual right it's like the whole mcclellanism component the global village thing going on here that's like like like absolutely mind-boggling you know that's like a like the singularity may not just be in the machines it may be in the in the humans right like like the singularity is in us as like again that whole global village idea right like we're all our community is just it's just like this massive artificial virtual thing it's not real people anymore it's it's like pieces of interactions that It's a network of hyperreality, right? And then humans are kind of in the middle of that massive graph, right? We're nodes in that graph, but we're not the only nodes now. There's all this sort of synthetic information and other media kind of coming in as well. It's a really interesting time, right? It's like, this is a turning point for the entire civilization. At least that's my perspective, right? This is a very significant moment in history.
yeah let's say you think like the components of a community are people could that be different could the components the raw components of what a community is end up just being experiences that make up your day and and and that collection of experiences makes up your community not actual like people and it's like this perception problem where you feel fulfilled at the end like a potato chip you feel fulfilled at the end of the day but there's still there's like low nutritional value because there's something You still feel full, but there was still something missing because it wasn't actual people. Yeah, you're maybe missing an emotional component or some sort of satisfactory aspect of a human, you know, one-to-one human interaction, right? Yeah. We're social organisms, right? I mean, we're primates. By default, we're social. Like, we're some of the most social animals on the planet, right? And that's kind of how we evolved. So if you start to take away the actual social connection, Robin, is what you're saying, right? If you start to pull that away, what happens? What is the nature of, what does that do to us? Seems like there's also a space, too, though, just to where...
you know, as an interface conversation can recede enough into the background where there could be situations where technology is doing things for us, but we are taking very little notice of it. I guess that would require a lot of trust being put into it and certain things would have to happen to get us there. But then you're in a situation where maybe you can improve the quality of the interactions that you have with the people in your life, in your community. Yeah. Probably kind of a lofty goal, but you need a lot of design work to get there. Yeah. Yeah, it's kind of the car. Like, wherever you steer it and drive it and wherever you choose to go is where you end up, not where the car took you. So you got in, you drove it somewhere, and if that destination was to go hang out with friends, then that made that car and that technology, like, you know, enriching for your life. If you choose to get in the car to just go watch a movie by yourself on the screen, then, yeah, you used it to be alone. Yeah.
Well, when Photoshop came up, I think we were talking with Tim O'Reilly as well, this idea that, you know, someone could use Photoshop and insert a person into a scene that wasn't there and they could do it at a level where you really wouldn't be able to tell the difference. Hell no. But there's something more unsettling, I suppose, about a machine being able to do that. Like, if a machine did that and not a human, it creates a layer of discomfort. And then I suppose you also add in, like, if a machine's doing it, it can do it at scale. It's not going to take six hours that it would take a person to sit there and smooth all those edges. Yeah, I mean, we're also talking, too, about fidelity in other dimensions. Like, you know, Photoshop is basically just kind of a two-dimensional experience, right? A two-dimensional impact.
um you know i think we understand that all images and things that we see are somewhat processed but now like you the discernibility between reality what's real and what's false like it starts to starts to get much more complex right we start thinking about all the things that we see um we start thinking about all the media that we receive you even think about like you know if you look at what's happening with like chat gpt chat gpt can be asked to be influential right it can understand aspects of your your um your cultural perspective right what you're where you come from what you believe what you know it can target it can very simply target that through some very simple prompt engineering right and it's it's different with like a photoshop like photoshop can convince you via images that something is true and you can ask questions and we can test whether or not those images are real or not but then when you have an entity that has these very human-like qualities telling you things you know that's a very different type of experience right and you know and if it if it speaks with a certain type of voice and a certain type of authority with a certain level of nuance and by the way these these these llms can do that right very well um you can tell it to you can tell it to explain something as if you're a kindergartner or a college student it will do that and so you think about that level of impact right how tailored that can get it it's it's uh it's it's powerful i think it's more powerful than most people realize yet like what this is going to do i think there's a there's another dimension that that doesn't get talked about a lot which is uh um this idea like you know we have these senses right the sensory input data
our eyes our smell ears um and and uh with each sense we have like a different degree of certainty and confidence score in a sense right so when we see fire versus when we hear fire versus when we smell like each sense has its own like i see fire i have a higher confidence that there is fire than if i smell fire versus i hear fire and then of course when you combine them together that confidence even goes higher i smell and i see fire and when you look at sort of a clue and angle which is like digital media is just another sense it's just an extension of our senses right it's just it's just another sense of eating data input um but you could also say that it also has its confidence score right and so if we see it online right um you know we're we're piggybacking on the confidence of of sight right like you know i saw the fire i saw the fire online um but there's also a chance that that that score just goes down that we just don't regard it it doesn't weigh as much into
our sort of confidence level as as as it did before and it says like now if i smell fire because Because my neighbor is always barbecuing, right? So now if I smell fire, I don't believe it anymore, right? He's always barbecuing. But if I see fire in my kitchen, so now seeing it online, it's like super low confidence. It's just my neighbor barbecuing probably, you know? This kind of speaks to the idea of media literacy and the level of sophistication people have with understanding what they're consuming and what's being delivered to them and how it's being delivered, right? As you were talking about that sense. And it's interesting because I've been thinking like, all right, well, if you look at...
Boomers versus Generation X, I would say probably Generation X understands a lot about the sort of fallacies inherent to like sort of digital media consumption where that can be somewhat problematic, but not everybody in Gen X, right? If you're on the tech side, you definitely, you definitely understand a lot of these issues. And then I get worried, like do, does Gen Y, like Gen Y millennials definitely are much more sophisticated, I think, than Gen X in terms of understanding how to sort of navigate the digital space and understand the kind of reality that's emerging. Like they're very sort of sensitive to that. What's been interesting is, and I don't have data or anthro research around this yet, but I suspect that Gen Z is even more sophisticated, right? Like they're the ones who are in TikTok all the time. You know, they're never in Facebook, right? But they're doing all this stuff and they're, you know, building things in Roblox and on TikTok and they're in all these other environments. You know, when I talk to younger people, it's surprising. I mean, younger people, I mean like younger than like 13, like, you know, like 10 year olds or eight year olds. You talk about what they're doing. They have this sort of inherent knowledge about what's bullshit and what's not. Like they kind of know, like because they've been in the space so long.
But I think there's hope for them in terms of being able to sort of traverse the space. But again, the landscape is about to get a lot more complex too. And I'm really curious to see how this emerging generation deals with all this. Because they're the ones who have to figure it out. We're kind of at the end of our run, right? They're coming into it. We used to say digital natives back in the day, right? Rob, when we were thinking about types of users. This is beyond being digital native. This is a whole different level of sophistication in terms of use and understanding. And it's because they emerge into that environment, but they're also not...
entirely duped by it right yeah and there's also like something like what i don't know what you call it like tooth fairy syndrome you know like you know you teach your kids that the tooth fairy exists everyone pretends it exists and it's fun it's fun to pretend it like we all just you know kids know there's that age they know you know but you're gonna keep this straight going on right and they're like okay like this is it's fine i'll pretend it's real let's go with it like why not does it have to like do i have to get to the truth do i have to know the truth like do i need do i need the truth on everything like it's okay i kind of like i like this idea of the tooth fairy so i'm just gonna go with it you know it doesn't matter what the truth is It's my truth and I'm going to choose it. So I'm going to pick my truth and that's what it's going to be. And that's all great until there are like real consequences, right? Yeah. Yeah. And as we see, like the tooth fairy is all great until, until like, you know, you've got to hold an election. Right. Until you need a, until you need a shot in your arm, right? For the pandemic or something. Right. Yeah. Things are getting, things are getting a little crazy. That's for sure. Yeah. So I, I can't help but go like way back in time here, which is like you and I, you're leaving Kodak.
Kodak's like, there's like people that will listen to this like, what's Kodak? And you were there to kind of try to see if that could be resurrected, right? If they could stay relevant. And this is, ironically, a bit of a Kodak moment for some companies now, the AI moment, right? Yes, it certainly is. And having a front seat as you're witnessing this, and you know what's happening, and everybody there knows what's happening. It's not like they're in the dark. They see it, it's happening, but they can't stop it. Kodak just keeps going on this spiral. No matter who they bring in, you geniuses to come in and save the day, there's just no stopping this trajectory of losing it.
Is there anything you think that you learned from that, like seeing it again, that companies would be like, Yeah. Anything could have stopped Kodak? Like, is there anything that could have, like, was that just it? You know, Kodak, that's just an interesting story. And I think there's a really, there's a cool article on LinkedIn from Ken Perulski, who I think was a engineering leader at Kodak, who really outlined what killed Kodak. And I don't want to get into all the details, but really comes down to the fact that Kodak's leadership really wanted to invest, you know, continue to invest in the consumables market, which is all Kodak ever was, was selling paper and chemistry. It really had nothing to do with, you know, film. Film drove the burning of paper and chemistry product, right? It's a razor blade business. And, you know, there was a hardcore group of people who...
thought and knew that digital imaging was the future and was driving towards that right but they just couldn't they couldn't get away from you know the previous patterns you have to think about it the company is being run by chemical engineers right it's a bunch of chemists chemical engineers kind of running the space and people or you know you could think of them also as like material scientists right they're thinking about how to use materials to make money where then digital ink you know like printer ink becomes the new the new and toner becomes the new um you know developer fluid and paper and film and all that type of stuff meanwhile we're talking about well how do you monetize a digital ecosystem of images right you've got eyeballs you have like it's literally like images on the web is a killer app right how can you like why monetize that and of course you know shortly after i left not shortly after i left kodak but you know kodak goes bankrupt i left before they went bankrupt and then you know instagram gets bought for like you know two billion dollars which was more than kodak's entire valuation at the time it's kind of um it's kind of a crazy moment but what i would say if i were to map those learnings rob back to kind of where we're at today with this with the emergence of you know significant emergence of artificial intelligence over the past 10 years um is that There's going to be groups inside of these organizations who understand the implications of this, right? That are going to be fighting to change directions. But there's going to be leadership, you know, the previous, you know, entrenched culture that's going to prevent that. And we're going to see that across all industries. Like there's going to be a sort of resistance to the models, but there's so much happening. There's so much momentum and it's so transformative and it's transformative everywhere in all the ways.
that I think a lot of organizations are going to get caught out. And we're starting to see a little bit of that now. Everybody has to make that shift, that big jump. And of course, we've watched Microsoft do that over the past few months, and it's been spectacular. They understand the implications. And then now we're watching Google scramble. When I first interacted with some of the large language models and saw how they were working and what they could do, I was like, it's over for Google. It's over. They've just been officially completely disrupted. And it's really interesting to see how Microsoft was able to understand that level of disruption and react to it so effectively. It's just been fascinating.
And I guarantee you there's, you know, obviously Microsoft has their, you know, R&D and investments in AI that they've been doing and they're obviously keeping their finger on the pulse, but they've been moving aggressively. You can tell this isn't just something that just happened in the past couple of months. They've been preparing for this for years and making those moves. And so I think other organizations need to look at that as a model and say, hey, we need to react to this now and start thinking about these deeper implications and think about the implications to the end state because the technology is evolving so rapidly. Like I remember back in the, you know, the late 90s and early 2000s, we used to talk about internet time.
I don't know if you remember, I was like, oh, we're on internet time when we do this. And it's like, oh, the pace is crazy and fast. You know, AI time is at least two orders of magnitude faster than internet time. I mean, if you watch the papers come out, like the advancements in AI are increasing week by week, not year by year. It's within weeks. You'll see a paper come out where, you know, like NVIDIA is showing, you know, some of those NURBS modeling. You're just like, oh, this is mind-blowing. Then the next week, another paper comes out where somebody else then took that to a whole other level of magnitude. And you're like, this is one week or one month between these two papers and the advancement and the technology has been that significant.
you can't think about like people are like well how can we use chat gpt it's like let's take a step back because that is just the foundation that's just the beginning we're going to be taking this to some sort of end state that is almost unfathomable at this point so we have to you have to think beyond just what you see today because it's it's it's powerful stuff yeah i was thinking like when um mobile kind of you know hit and we were we were dealing with pretty big brands at that time yeah and and i remember uh you know going in and and and these guys you know saying hey you know i need a mobile app i got a budget like 10 15k um that should be enough right uh and then and then of course they got what they paid for and then and then these companies are like oh i guess mobile's not for us we tried it like didn't work right right and getting getting them to like pony up a real like make a real go of it was like it took a lot of A lot of, you know, tailspinning, right? They had to understand the impact of mobile technology. I mean, this kind of goes back to the Kodak conversation because I used to work on consumer electronics and they were always worried about what they used to call feature phones. Like feature phones had cameras and cameras were a threat to their digital camera market. And, you know, one of the conversations was always like, well, are you going to carry two devices or one device? Which way are we going? So they started to invest in, you know, you know, Motorola's, you know, cameras and looking at, okay, this is kind of the direction of the future. I think when we saw mobile emerge.
the the idea was well why do i care about a phone right this is and what is the impact of this this thing and i think there's sort of this concept of the legacy concept of the telephone and why is it important in this content in this particular context and why does my business need to have a presence here but it was like this isn't a telephone people we call the phone we never use it for i don't know you guys probably never call people on your phone like i do i don't call people um but it's this is an incredibly powerful mobile you know ubiquitous mobile computing platform right? And it's got a very powerful graphics engine and telecommunications engine associated with it. What is the possibilities of that emergent technology? Like people, if you're, you know, if you're just some business and you're operating in a certain way, you're not thinking at that level, but that's how you have to think about this stuff is that this was a fundamentally transformative technology that has now, you know, shifted how we work, think, interact with one another. That was like first principles, you know? Exactly. Exactly. That feels like a continuation too of sort of the UX struggle of trying to treat things like a bolt-on.
when it when it really requires something much more holistic that affects an entire organization um you know the companies that do it right are often like the ones that are kind of built around that process the lucky ones i suppose yeah and then and then you have the companies that kind of like throw 10 grand at it and watch it fail and then the companies that like see like maybe they someone understands how huge and holistic and undertaking it's going to be and then they back away because of the scope of it so yeah it's hard to know Yeah, I mean, the level of, I mean, the significant investment has to be made like in this space. And there's, of course, lots of different ways to do this. You can be at the ground level, developing your own technology and models, et cetera. You can be leveraging. I mean, the amazing thing now, like now that we've got, you know, foundation models, like this is really powerful technology. It's, you know, you could think about, you know, previously this idea of what it would take to sort of deploy this technology or develop this technology. And you've got, you need lots of data. You need very talented data scientists. You need to compute infrastructure to be able to process all this stuff.
um it's starting to shift a little bit right and all that everything i was speaking about before that's billions of dollars of investment to create a chat gpt right like it's built like just to train them just to train it right like holy shit like the compute is massive um but with with an llm or foundation models like these publicly available foundation models it's like you know you've got you've got the starter kit you get the ferrari motor and the ferrari trent chassis and all that um you get the starter and then you've got to just you know figure out how to configure the chassis the suspension or whatever the interior or whatever to go race it on the course It's probably a very deeply flawed metaphor there, guys. But you get all this stuff and it's an accelerator, right? It's a massive accelerator. So the level of investment to move forward is much smaller. Now you've got a foundation model and you've got all the APIs and SDKs associated with that. And then you can just pick up and run as a developer. It's not like it used to be. And I mean like last year, right? It's like it's already changed and it's really interesting. And so what are the implications of the sort of foundation model idea? What is that going to do right now that it's kind of like empowering?
industry at a different level than it was being empowered before. And I think that's going to have some interesting implications too. Yeah, there's kind of like a moment, right? You just imagine that because a lot of these companies are really in the asset you know, protection mode, right? They're not in the explore mode. They're in the exploit mode of, you know, it's like when someone gains wealth and they go into like the wealth preservation stage, right? Not make too many speculative moves, right? And so the CEO gets told like, you got to do something here. You got to take a risk. So he walks into the room with his engineers and he says, okay, who in here has a surefire idea on how we can leverage AI?
And everyone looks at each other. Did you say surefire? Yep. And he's like, okay, I'll come back when you do. Right. Which is like, never. Like this is just moving too fast. Right. But by the time anyone is sure.
The moment's gone. This exact scenario presents me with design challenges on a daily basis. This is my problem. Things are changing. You've got to adjust continuously. I think you're right. I don't know, Rob. If you're asking what's the surefire approach or there's a business context in which people are asking that question, there might even be models involved in that. You're asking the system now for advice on how the system is going to evolve. These are the things that cause me great...
a great deal of existential angst, right? Like you're now clearing the machine about how to improve the machine. And you're starting to get into this sort of the idea of like a von Neumann machine, like self-replicating machine, self-improving machines. I hope you guys read a lot of sci-fi, but that too is another very significant issue beyond just like, you know, people were like, oh, we're going to become like slaves to the robots. It's like, no, no, no. The robots don't care about us. But if you give the robots unlimited resources and the ability to self-replicate what happens at that point, right? Yeah. That can be a very dangerous situation too. Yeah. Yeah. We definitely lose ourselves potentially. Not just ourselves. I mean, you can take it to an end state of all consuming.
You create a set of all-consuming robots that just continuously self-replicate. It's basically a cancer for the entire galaxy, right? They just will consume all matter and replicate until they can't replicate anymore. A more potent cancer than humanity is proven to be, even, which is pretty impressive. Well, that's what humans do. I was listening to Eckhart Tolle. He's like a German, kind of a spiritual teacher, but someone asked him if he thought that AI could ever gain consciousness or become spiritual, and his answer was interesting because he said that consciousness...
or in his view, it's not something that comes from within your mind. It's something that exists. It's a frequency that your mind is able to tap into. And so from that train of thought, he basically said, well, I mean, I suppose a complex enough machine, why wouldn't it be able to tap into consciousness if consciousness is just this frequency that's floating around in the world? Yeah, the idea was like, there's consciousness and then your brain is a conduit.
for consciousness to speak through you as a person, like, versus consciousness, you know, exists in your brain. It's a compelling idea. I don't know how scientific it is. I'm just going to say that. I don't know what he's talking about. But when we talk about consciousness, what we're really talking about is, you know, and spirituality is related to consciousness in very concrete ways, right? I can go deep on this topic, too, if you like. But really, you know, when we talk about consciousness, we're talking about the default mode network of the brain, right?
The default mode network of the brain is an interaction of different, you know, biological systems and anatomy within the structure of your brain, right? And that's what creates the internal dialogue, the ability to think with language and create a sense of identity and to also have the theory of mind and other individuals, right? You guys know about theory of mind. It has a lot to do with like how you understand sentience in AI systems as well. So our consciousness is a direct product of our biology, right? These silicon substrates do not share...
the biological substrate it's a very different type of substrate could you train a model to be spiritual could you train a model like it's a model is a is a type of turing machine right can a model then fully simulate you know the biological structures of of the human brain and then also the the effect of the default mode network and then have sentience from that part and then a lot of our spiritual experiences have to do with how the default mode network functions whether you extinguish it right and then you have this sort of connection to the greater world or not and so can can machines be spiritual in a human-like way Probably not, unless we build a synthetic brain with that technology. But could they be spiritual in a silicon way, in their own way, with a different type of spirituality? Maybe a spirituality we can't identify as humans, right? Maybe that's already happening. Rob, I don't fucking know. They're a black box. We don't know what's going on inside of them. I think if you actually break down what a neural net is and how it works, it's this concept of taking data and then anchoring it to something, some data set of some kind. And so you're taking...
taking data and you're kind of and then you're just like anchoring it to some you know some relative point right and kind of like when we dream we there's all this data and then we you know our dreams are essentially essentially anchoring this like random data and trying to make sense of it to like real things that are familiar to us right and so a lot of like our ability to to function without like analysis paralysis is this like idea of like anchoring data into like single points to say oh that's a person and then everything that happens gets like categorized and anchored in that data set as person and so neural nets are essentially the ability to synthesize data in a practical way, to make it practical, right? And is spirituality just nothing more than, like, just a large neural net in the sense that it's, like, big data sets that are just anchoring to, like, I got to make sense of this, so I'm going to anchor it to something I know, like, whether it's anthropomorphize it or, like, these are all just ways to, like, take data and try to make sense of it, right? With our experiences, right? FMRI, so you imagine you're sitting in this...
fmr and you look at they show you 10 000 pictures in a row and they're training the model and then from there they can read your mind essentially so using using a generation they're like paper yeah yeah that's crazy but well this goes back to the that when we're talking about the danger in the synthesis of media in that if it like not only can it be nuanced but it can understand things about you right even understand what you're potentially even understand what you're thinking with the right technologies and the right inputs right right like this is like rob it's so it's like this is it's it this becomes the end of like when we start talking about the fmri imaging and the interpretation of that like it comes to the end of the lie. Like, you can't lie to people because we always can see what the truth is. It's like that movie. What was that movie called? There's a book. The Intention of the Lie or something? Yeah, there's that movie. Oh, yeah. Yeah, that movie. And then there's a whole, I remember there's a science fiction book in the early 2000s that was all about, they create like a truth detector machine and what the impact of that was on society. If you can't get away with a lie anymore, what does that do? Which has got huge implications. But there's the fMRI. Like, that's a really fascinating thing, right? Using basically pixels. Like, an fMRI image is just a raster grid or a series of raster images, which are just, you know, RGBs in a matrix, right?
And it's using that to determine your thoughts. Like, let's just process that for a moment. But then there's like, there's some other topics that I'm really interested in too that are related. And that is like, you know, there's that whole experiment with like cetacean communication where they're capturing whale sounds, right? And they're trying to get a critical mass of data from the whales so that they can start to decode their communications, right? And so, you know, we live in contact with a number of different species that are very sophisticated. Like a dog isn't the smartest thing in the world, but a dog knows about 200 words and can communicate things to you.
Imagine like this future now where we can actually have dialogue. And I don't know if you guys have watched the YouTube videos where they create the little interfaces for the animals and they can click the button that says, I want food or I want snuggles or whatever. You could kind of take that to the next level. And so now all of our communications have opened up in lots of different ways. And I often wonder, is it like with mammalian brains and kind of communication behaviors, I think this is very possible, but what else is it going to open up for us? What other data sets can we process to untangle different strings of communication from? Like we know that funguses and forests and trees are intercommunicating and moving a lot of data.
We know that colonized insects can actually behave in very sophisticated ways. What other sorts of interesting second or third order effects are going to come out of this technology that are going to just totally disrupt how we think about our ecosystems, our place in the world, et cetera? People just aren't even understanding that this is possible, but it is. Literally, it's happening now. I know. Yeah. If my dog could talk, I'm pretty sure it'd never shut up. I think I'd be that guy who finally couldn't hear and then he got hearing and he just took it out. He's like,
Let's go for a walk. We should get some snacks. What are we doing right now? Maybe we can unlock the secrets of dolphin humor because they seem to be laughing a lot and their brains have so many more folds than ours. I bet they've got a whole... And they're communicating. These communications, there's communication. We think about communication in one specific way. That's very huge. Other species communicate in other ways. We're going to be able to untap all of this. It's not just our speech or our thoughts. It's our body language. It's our facial expressions. It's the subtext of our speech as well. It's subtext of our...
All of this stuff, it starts to be unlocked by these very sophisticated models. They can parse that up and then process it. How limited do you think our words are, though? We're pretty imprecise, right? That's why coding language exists. Because precision of natural... It's not like nobody tried to code in natural language before. From the beginning, this was a thing. But it's just not precise. There isn't the precision there to actually pull it off. So you need that generative component where something is going to fill in the blanks. Kind of like that reading your thoughts. Probably 60% of that image is going to be just generated. Just hallucination, right?
And that's what was so scary is like, you know, how would the person looking at that image know what part was hallucinated and what part actually, you know, was, so now, so now it like sees it hallucinated, you know, that you, that you, you saw something, um, or you thought something that you didn't think. And, you know, this was like the, you know, conversations we've had several times now where like, there's an irony in the AI regulation. thought process here, which is like very minority report, which is like they're trying to regulate what hasn't happened yet. Like there's a chance it's going to happen and they're trying to regulate it now, which in itself is very funny because it almost depicts minority report. Like, I don't know, let's ask the machine who's going to do it and then let's arrest them now. Like maybe that's what we should do.
Yeah. Maybe that's how we need to regulate AI, is just feed it a model of who's going to break the law with AI and just go arrest them right now. Well, that's the type of stuff that we're... It's very meta, right? That's the kind of thing that we're talking about. Well, just read their thoughts. Legislators, like, first of all, the legislators are dumb, right? They're not technical. And they operate within the political context. They don't know, right? And they can't legislate. But, I don't know, do we bring the machines into that equation as well? I know. It's like, we could.
Should we? I don't know. These are the big questions, right? Yeah. The systems that we have where we could create regulations tend to move far too slowly to probably be effective with technology that's going to be moving this quickly. I think it's not just for, I think the fundamental threat of AI, it's not just our legislation. Like this is, everything's going to get upended, Rob. Everything's getting upended. Like every industry will get upended. Our entire, I don't know, it sounds like a little, like I'm a doomsayer. I've got hope for humanity, but this will present challenges because.
It's going to disrupt how we work, how we think about our work, how we get things done, our relationships, the nature of our civilization. We talk about how we relate to the world. Those are fundamental disruptions that are going to change things. I mean, I think about, again, Gen Z, their world is going to be very different than the world that the boomers and the Gen Xers have created up till now, right? It's going to be a very different sort of space. And, you know, what's going to happen to our concept of nation states and cultures, right? Like, I have a feeling even some of that's going to get disrupted, you know. It could be very problematic for a considerable period of time. I'm not sure. What's the mood in your circles like? Is it fear? Is it...
I think it's more on the, like, I'm scared, personally. And I know that in academic circles, there's also trepidation and some fear. And people who understand the powers of these technologies can see what the impact may be. And so it's kind of like, you know, the climate scientist in the 80s talking about global warming. Like, that's kind of where we're at today. But on the industry side, you know, it's more about the enthusiasm. It's more about the power of the technology and what the technology can enable. And obviously, there's a lot of very positive things that can come from this. And a lot of accelerators for business, a lot of things that can improve the human condition, right? Yeah.
But there's, I think, the amount of risk. Like, we don't even understand the risk. We understand the benefits because we can imagine those, but there's risks and challenges that we don't even perceive yet that could be fundamentally problematic to our civilization. Yeah, a lot of the fear I'm seeing with folks in the space is the fear of the people that don't understand it. Doing something...
you know, villainizing the folks that are, that are optimistic about it. You know, there's, there's kind of a point to like, you believe in something, you think it can create a, you know, positive future. You're not, it's not a hundred percent, it's less than a hundred percent, but it's more than 50 or 60%. So you still have a little bit of doubt. And then you wonder like, oh, you know, is it worth, is it worth being in the spotlight? Because for a lot of us in the space, like we've been able to operate a bit on the sidelines, you know, a bit not in the spotlight. And now, boom, you know, spotlight is on and it's not a.
I didn't expect to feel this under scrutiny and fear. You're kind of running around excited about what this can do, preaching and people looking at you like you're crazy to then people looking at you like you're creating the next tool that's going to end the world. And you're going, whoa, is it worth it? Even if it's worth it in terms of you believe in the technology, is it worth it to be in that spotlight? Yeah. That's an interesting topic and it's timely too because that Christopher Nolan film about Robert Oppenheimer is about to come out, right? Yeah. Very much we're in that Manhattan Project moment. Right.
I'm really curious to watch that and see how that's interpreted in film, right? It could be really interesting. But we're kind of at that point, Rob. And it's like, yeah, what are we doing? And there's people in the, you know, if you're a player in the industry, if you're involved in this, I think everybody's asking themselves, what do we need to do? How do we need to be thinking about this? What are our responsibilities? I think we kind of passed that Manhattan Project point as well. Like the bomb's already been dropped. The genie is out of the bottle. Like it's out. The genie is out of the bottle.
So it's not like, what do we do now? And I think this is the ethicists, the people strategizing on the visibility and observability of these systems. Yeah, I think people are diving into this head as fast as they can and trying to figure out solutions. But again, I'm just wondering if the pace of change is going to be too fast or the impact is going to be... We're going to focus on certain areas that we can understand today and we're going to mitigate those risks. I think we're going to mitigate them just fine. But I still think that there's other unknown hidden risks that are going to emerge that, yeah, I'm not sure how we're going to react to those. Yeah, Rob and I...
feel like there should be more designers working on this stuff um just in general do you do you think there's a way to and not just like visual design like all the aspects all the different disciplines that go into experience design i think um can have great impact as we start trying to figure out how to move forward with AI. How do we get them involved? How do we track them? Yeah, well, there are designers dealing with this. I know there's a group that I interact with out of Amsterdam that's looking at design, specifically design topics with respect to AI, and specifically the challenges and risks associated with AI. So this is happening. There's research groups coming together, focusing on these topics. I think it's, we talk about design. When we talk about design, we're talking about people, right? Like we design things for people.
and i think people is the important part here and so it's not just designers i think let's let's let's bring in the psychologists the sociologists the anthropologists everybody on the human side like we have like right now like ai is dominated by you know the math people the stats people all the data all the different flavors of data scientists right these are all people who you know live in that sort of math and engineering and science world right we need to bring in the people from the humanity side to understand like the more human aspects of this and to dive deep like the philosophers of the world like how are they interpreting this what are the implications um how do we want to think about the emergency of these technologies like that's where we need to help so i think yes yes design but even expanding that even further to all of the sort of um humanities uh kind of bringing that to there because yeah you know you sit around with a bunch of math nerds all day they don't they're not thinking about psychology they're not thinking about uh sociological impact they're not thinking about the uh the emotional impact on the human like that's not in there that's not what they're doing um i think some people are aware that those are issues that need to be addressed but it's not it's not kind of a focus from that perspective but i think you're right josh you got to bring that to bear because that's because this is this is human civilization we're talking about
yeah yeah it's like bringing that design mindset yeah to bear in other industries because i think as we learned from from ux magazine which branched out from like a core audience of hardcore experience designers to like now kind of everyone has a vested interest in how things are designed they're the minority in a way yeah i think a lot of us that have been doing this a long time realize that ux is not a profession it's like a It's like a minor. You're a product manager, you're a manager, you're an entrepreneur with a minor in UX, and that helps you do your job better. But as far as UX as an exclusive career and profession, it's almost like learning a second language. In itself, there are a few jobs for translators, and the rest, it just helps you as a negotiator or as a politician or whatever. Thinking about UX and designing UX is really about the application of common sense and thinking about the user and what they're trying to do, right?
It's not, there's not some special magic involved. You know, execution of beautiful and well done UX and interfaces that go along with that. Like there's skills and things that are associated with that. But the core problem solving, right, is it's always been blurry from a discipline perspective, like blurry edges. It's a gradient. And you guys have all seen the diagrams, right? It overlaps into all of these different spaces. And the design mindset is a framework in which you, you know, the lens in which you try to solve some of those problems. But it's not the only lens, right? And user experience, it's just a, it's a way to understand things, right? It's a way to think.
And you don't have to be a designer to do that. You just have to understand the lens. And once you understand it, then you can never get away from it. You start to, you're thinking about people, you're thinking about what they're doing, you're thinking about how they're thinking and feeling and, you know, what they want to accomplish, all those types of things. So much of this is just about good taste, right? I mean, whether you can't, whether you can design, whether you can't design, just knowing what's good and what's not good, right? Yeah, like that Rick Rubin quote, like someone asked him, like, what is your job, basically? And he's like, well, I just, I believe very highly in my own taste and I'm very good at articulating it to people in moments when they can use it. Like, that's what I do.
I think, you know, when we talk about designers, it often comes down to taste. When we talk about user experience and the performance of a product, that is measurable, right? There is science that can be applied to that. But at the same time, you know, you have to make these experiences desirable, too. Like, even the most banal enterprise experience needs to have that. You know, Rob, we build a business on this, right? It has to have that dimension. And so that's where, you know, some of the tastemaking comes in and, you know, all that type of stuff. And that's, you know, where designers are.
Our designers create culture. That's how I think about it. Either material culture, if you're an industrial designer, visual culture, if you're a graphic designer or artist, you know, the digital culture, if you're a more traditional UX designer, like we're the ones making the things in the world that populate our space and create our experiences, right? And so, yeah, you got to be a little bit of a tastemaker, which just means, you know, taking risks and understanding where things are going and understanding what's possible. Yeah. That's a big piece of it, right? Like as a designer, kind of like science fiction authors, they tend to have the ability to kind of see things through. They can look at something they're building and start imagining how it might look 10 steps down.
So I think that that's an important skill set to bring right now in particular. Yeah, extrapolation and implications and impact. Yeah, that's, I mean, when I'm thinking about new tech, we were just talking about mobile phones. I'm looking at a phone because I, you know, back in the day, I was thinking about, you know, this was back when we used to call, we used to talk about Ubicomp, ubiquitous computing, right? That was like the future. Ubiquitous computing and ambient intelligence, which we live in that world today, but we don't use those terms at all, right? Like, it's just, you know, I've got my Alexa over here. I've got my mobile phone. I've got my laptop. You know, I've got my Garmin device, my watch, everything. It's a ubiquitous computing environment now. And I can interact with these systems as casually as I want, right? So we're thinking about it from those terms, thinking about it at a very base level. And so that's how you have to think about, you know, what do I see today?
But what does that imply for tomorrow and the day after that and so forth, right? Because the way that humans work is we make a thing and then we make it better and then we make it better again and we don't ever stop. That's why we have artificial intelligence now, right? And if we make it too much better too quickly, everyone freaks out. And we do that. We do that. We've done this. And you know, the funny thing is, Rob, you look back to antiquity, to classical, you know, classical civilization. They knew these ideas. They understood this, right? Like this is the myth of Pandora, right? Prometheus, right? This is the Pandora moment, the Promethean moment all over again. We've been warned about this by our ancestors. They said, don't do stupid shit. Pay attention to what you're doing. There'll be problems. And think about things before you act and do whatever, right? They warned us. But we always forget. Somehow stupid shit always feels new. Yep.
That's a great way to wrap it up, guys. Somehow stupid shit always feels new. Well, thanks, Tim. We'd love to continue the conversation again sometime. This is awesome. I can do this all day, as you can tell. I'm happy to contribute at any point if you guys want to ping me again. I love it. I've got other folks like me, too, so just let me know. Happy to do more. This is great.
All right, thanks again for hanging out with us right here on Invisible Machines. Thank you, as always, to the team at UX Magazine, especially Kate Timchenko, the marketing team at OneReach AI, Elias Parker in particular, worked very hard to make this podcast great, as does Michael Litvinoff, our video editor. Please subscribe to UX Magazine wherever you get your podcasts so that you can hear new episodes as soon as they come out. If you want to watch new episodes, and I recommend you do, the video feeds are really pretty amazing, follow the Invisible Machines YouTube channel. And I think that's all I have for you this week, so let's go ahead and look forward to next week.
when we will connect again right here on invisible machines