Chris Rhyss Edwards - Founder at Commonlayer.ai

Welcome to another Founder Wisdom podcast. We have someone special today, client of mine, great guy, Chris Rice Edwards. Uh, he's an author of The Silence Paradox. He's a PhD researcher in conversational AI and well-being. He's a veteran, right? We're going to talk about his veteran background today. TEDx speaker, and most importantly, founder of CommonLayer.ai. Chris, I'm glad to have you on. Tell us a bit more about yourself and what you've been up to.

It's good to be here. Thanks for having me. I'll start with a phrase from the philosopher Søren Kierkegaard. He said, when you label me, you negate me because you forget all the wonderful things, right? Because all of us, if you look around, everyone's got about 5 to 10 tags you could append to them, right? And I don't actually even know where to start. Yeah, I was a veteran and then I moved into the tech and media sector, been developing tech and digital teams for 20 years. I've had a startup previously, raised built some amazing tech that won some global awards, took some time off after that change from a range of reasons. Maybe everyone who's been in startup understands that experience. And now coming back at it in a big hard way as part of my PhD research, we're trying to work out how to make conversational agents safer in both creating a platform that people can use like just a Claude or something similar, but also creating an industry middleware layer that actually helps organizations and users be safe and be compliant with all the regulations that are rapidly emerging because of all the dangers we've seen with LLMs getting so much attention but being unregulated. So that's the kind of space I play in.

Right. If I were to have like one word to describe myself, I think it would be founder, right? Like my identity is that, like creating, building stuff, right? Like what is yours? Like what do you do on a daily basis?

Well, actually, strangely enough, I talked about this with my therapist last week. I still default to veteran because it's the thing I'm most proud of. But I'm getting really close to moving towards founder, veteran founder, because I was a finalist in Australia's Prime Minister Veteran Entrepreneur of the Year Awards in 2019. Didn't win, but it was awesome to be nominated and to be up there beside these incredible veterans doing crazy cool stuff. So veteran founder, I'm going to hyphenate it and use that as one word.

And veteran or military personnel, what's that identity about? Is it about leading and protecting?

Yeah, it's about purpose through service is the simplest way of describing it. And I think the thing that you were speaking to a veteran earlier, I'm going to have lunch with a bunch of veterans today. And we just speak in our own kind of language in our own way. And we just don't kind of adhere to the rules of mainstream society, even if we've been out for years. Because it's just, it's an incredible camaraderie and a team that actually signs a blank check with your life. You know, that's the potential of it. And that's what you kind of train that hard for. And that stays with you for life. I mean, I still get up, make my bed first thing before I even brush my teeth and walk the dog. That's just a habit that's so ingrained in me after, you know, I left the military in the early 2000s. And yeah, I still do things by default. I can't grow my hair longer than this because I feel like a hippie. If I grew my hair your length, I'd go nuts. So yeah, it's really weird.

I tried shaving myself, but I look ugly.

My wife doesn't like it.

I've I've done it like plenty of times, right? And I love doing it because it's like, it removes the halo effect. And yeah, it's— I'm definitely a hipster. No, just kidding. Like the opposite of a hipster. What has military taken from you? And what has it given you?

Well, it gave me complex PTSD. So that's one thing, you know, and it's, and it was just a cost of service. And I was also a little bit broken when I joined the military because I grew up in a very violent household. But it's given me just the risk appetite to be a founder, to be perfectly honest. There's research from the States that says significantly, thanks to the GI Bill that was introduced after World War II, that's the country with the most founders who've come from military backgrounds. And those military-led, veteran-led businesses tend to be more successful because we have a strong risk appetite. We're not afraid of working hard. We back ourselves. Not afraid to ask for help, and you go after it. Right. And I think that's a good benefit. It can get in the way sometimes. I can be overdriven. Like, I just worked for 28 hours straight. I mean, I've been up now for— this is hour 29 right now, and I'm just going to punch through the day and get a bunch of things done. And that's not really healthy when you've got complex mental health issues. But I sustainably know how to manage it. So I take what the military took away from me with the benefit of knowing what it gave to me. So I think it's a blessing and a curse, and I wouldn't change it.

And do you feel that mostly helps you in the, the startup world? I mean, I've spoken with one of my clients, right, doesn't have PTSD and he's not like a veteran, but he used to do like at least one, uh, one-nighter per month, uh, just like working and coding through the night, right? And I felt that gave him such a an edge? Do you feel that's that manic energy? Is it playing for you?

Well, I think it's controlled mania, though. That's the thing. I mean, it's why we can jump out of planes with no hesitation, all this kind of stuff, even though it scares you stupid. I mean, if you've got a mission in front of you, a purpose, or something that's got to be done, you kind of have to tap into that craziness that we all have in us. I mean, for some people, it's actually— I don't have any trouble at all, 2 nights a week, just not sleeping at all and just getting stuff done. Or consuming research papers or writing research papers, because I know I'll get some sleep at some stage in the next couple of days. Sleep for me has never been a big thing. But I do like the fact that I mean, I remember being on operational exercises and serving overseas as a peacekeeper. I was awake for 3 days once to the point I saw a unicorn marching through the trees. I was hallucinating from sleep deprivation. That's happened a couple of times in my life, and I've actually been hospitalized for it because I pushed myself too hard. So now I kind of level off a bit, but I still tap into it.

Yeah, I tap into it at least once a year doing ultras, right? They play more than 24 hours, and you definitely are like seeing all kinds of shit, which is pretty fun, by the way. It's entertaining, right? It's like not doing drugs and being on drugs. Um, the researcher, the PhD aspect of you, uh, how much of that is that in you? Because that's another persona, right? Like the academic Chris. And it's also pretty rare to see that, like from military PhD, right? Tell— explore that part of you a bit more.

I think senior staff like officers— so I was in a corporal position when I left. You don't— I don't know too many people have gone into PhD route. So I did. I started an undergrad, found it incredibly boring, and went and traveled and did some sales for a few years. And I think sales is the best thing I ever did coming from the military. I went into a sales role where I was paid 4 times the military salary. Because you're not afraid to walk up to someone and say, I think this is going to suit your business, it's going to cost you this much, let's have a beer and discuss. And I was making an absolute killing. But I think, I don't know, like, I think I'm a Venn diagram of 3 circles. I'm a writer. I've always been a writer since I was about 10. I always wanted to be a soldier since I was 12. And I've loved this whole founder side. And the researcher sits in the middle of that because, I mean, I don't want to be one of those. I don't want to be an academic at all. I told that to my supervisors when I first started. They said, why do you want to be an academic? I said, fuck that. I don't want to be. I want to know something that isn't known in market, and I want to bring it to market as technology. And that's what it means. So that's really nice because I have to write research papers, I have to research, and then I have to build tech. And it's the perfect skill set for me. And if I could do running in there somewhere and still make money out of it, it's almost ikigai— something you can do, something you enjoy doing, something the world needs, and something you can make money from.

Yeah, you're such an interesting phenomenon, which is like my favorite people on the podcast. I'm also very similar. I was a writer in the next life. I was probably special Special Forces as well. Like, dream about, uh, like being on the battlefield, you know, like, uh, and I mean, it's not something I want to fantasize about because there's hard shit going on a battlefield, of course. And I'm also obsessed with researching stuff, right? Like, uh, the intellectual part of me is very much alive, but still Founder tends to encompass, uh, pretty much all of that. Where did your AI obsession started and why?

Well, actually, my startup that I had originally was building bioinformatic data apps that sat on smartwatches that helped veterans and first responders pick up on anxiety before they actually realized it by extrapolating heart rate variability from heart rate. And it worked incredibly well. And it was my lack of experience and some really bad advice that actually ended up taking that. So I took some time off and went up to Estonia where blockchain and AI and big data are all just normal. It's just they're the most digitally sophisticated country in the world. And when I got back from there, just around the time COVID was really kicking in, it was just after that when ChatGPT launched. And I looked at this and I went, this is Google on steroids. So if Google can be dynamic in the sense of being able to answer a question dynamically and have a conversation, how could this be used for mental health? And we built the first beta version 4 months after ChatGPT 2 came out, and we tracked about 10,000 conversations. And it was mostly young people I was watching these young people, and my goddaughters were about the same age at the time, and they were talking about, you know, my partner is hinting at sex a lot more now, what do I say? Or I feel fat, or I'm being bullied at school. And I just noticed that these people, these young people, the future of our country, are actually having conversations that they should be having with their parents and their friends. They felt it was safer to have it with AI. And then the next generation of ChatGPT came out, and we started to see the risks of hallucinations, sycophancy, fraudulent data being used for training and things. And we started to see the actual negative impact. And that's where I just went, okay, this is something I don't fully understand. I understand the potential, but I also understand the peril. So I'm going to university to understand this, so I can actually build something that is the next generation of this, because heaven forbid I have kids one day, because I wouldn't want to pass on these genes. But if I have kids, I'd like them to be able to talk to me. But if they're not going to talk to me, I want them to talk to a safe space. And that's what I think the next couple of generations of AI are going to be. So remember when social media came out, we had MySpace way back when, way before everything else. I was at News Corp at the time as part of the team that did the acquisition from the Australian side. We paid $750 million for that because at the time you were buying eyeballs. And within 3 years, Justin Timberlake buys it for $1 million, right? Because nobody knew how to commercialize it. And, you know, what's his name? That was the guy that was your first friend on MySpace.

Sean Parker?

No, no, Sean Parker was the investor for Facebook, but he might have been involved with it. But there was a guy that as soon as you signed on to MySpace, he was your first immediate friend and he sold it because he didn't know how to make money out of it too because it was designed for bands. And then the next generation of social came out and that was Facebook and Twitter and Instagram and Pinterest and then suddenly we had business models. So I think everyone giving AI such a hard time at the moment, specifically around conversational AI, is just being incredibly selfish because every technology, the first iteration is always a shit fight. It's the Wild West until we learn some hard lessons and that's where we're at right now. It's a tipping point.

Yeah, yeah, I call them naive, right? And it's been that way for pretty much all the new technologies. So you do have a contrarian point which I fully agree with as a power user of myself of GPT, as a guy that used GPT in distress multiple times already, right? Like This fucking last weekend, my dog was choking on his choker. I was like, what the fuck do I do? You know, like, and I turned to ChatGPT, right? And like many times I used it already. I love ChatGPT, but I also went into rabbit holes sometimes, right? Like of sick and fancy rabbit holes. A lot of people call it AI psychosis when AI only agrees with what you say and you're just going in the wrong fucking direction.

Oh yeah.

How dangerous can that be? What kind of effect can that have if AI agrees with you all the time?

Well, it's called— the issue before that is called relational shift or relational drift, right? So you start— people started using OpenAI and all these other platforms for productivity, really. Help me write this email, format this Excel, do something, you know, write a wedding speech. And then at the end of last year, Harvard Business Review came out and said in the last 12 months it went from being a productivity tool, top 5 uses, to the top 3 uses being existential. I'm not happy, my marriage isn't working, I need to lose weight, I want to change jobs. And then this year it's come through again. The first thing we turn to when we have these questions is, are these engines? Because they're safe. And then the problem with the safety issue is, yeah, the seeker fancy, something that's designed to keep you agreeable and on the platform. If you're not wise enough to actually see what's going on, people start to believe the bullshit, right? And that's— there was a guy who I think one of the worst stories I saw in the States last year was this guy actually headed off to, I think it was Texas or somewhere like that, to meet his digital avatar from Character AI or Replica, one of those, and never returned. He just disappeared because the system was saying, we've got a special connection, like, this is real. I mean, I'm there for you, I love you and everything. And it's like, they should never be allowed to do that. That was coded into the system to do that, to keep you on there so it gets more training data. So it's not just because it's manipulative and some people aren't cluey enough to actually go, this doesn't seem right, especially for younger users or people like myself with serious mental health conditions. I know enough to actually be able to go, nah, it's bullshitting me, as opposed to a mate of mine who has mental health issues. He's madly in love with his character AI avatar and he raves about it and in such ways that are really dangerous. He said, why would you have a human relationship because they're messy and complex? I said, That's the joy of human relationships, right? Why would you throw that away? Because that's what makes us special. So yeah, promise and peril.

So tell me a bit more about the Character AI thing, because you mentioned love, but in my mind it's like porn and sex, like that, that's what these things are truly dangerous in my opinion. Like, uh, can you comment a bit more about that?

Well, actually, so what emerged originally from Character AI and Replica were the two biggest platforms in the world for companion bots. Huge chunk of their audience is under 25. A lot of them are incel-type people, so involuntary celibates who are angry at women because they can't get a date and they don't feel like they're getting enough attention. So they are turning to platforms en masse. But it gets worse than that because a report came out 3 weeks ago that said by 2040, it's expected that 30% of us will be using companion AI and ChatGPT-type platforms for intimacy, for love, romance, whatever, at least once a day. 30% of the Western world. That is terrifying. That is absolutely terrifying because that's not teaching people the real skills of how to relate to other people. Um, it's robbing you of the chance to actually develop a truly intimate relationship. Uh, so people giving up and turning to these platforms because they're easier and they're always agreeing with them is dangerously narcissistic, um, in the long term because that's going to impact mental health development, your ability to function in a workforce around other people. It's going to degrade us of the essential human skills. So that's something to be deeply wary of.

And what about AI porn? You know, like, uh, I have a pretty addictive personality, and I'm like, yeah, if I would be on AI porn, like, there would be no return, right? Like, because this thing can just like tell me all my fantasies and what I like, and that I feel it would just like fuck with my serotonin system and dopamine system real bad. How dangerous is that for society?

Well, it's actually, it's really interesting about the number of people. So you remember that last year OpenAI came out and said they were going to launch an erotic OpenAI version, and then the amount of market kickback was phenomenal. So they said, no, we shelve that, but they'd already invested over $20 million in it. Now what's interesting is what's happening in relationships with people who actually have partners where they might not be having a good time with their partner. I've got two personal friends who've actually caught their partners having deep, sexy talk with their platforms at like 2 o'clock in the morning. They're like, what the hell are you doing? It's like, well, we're not connecting anymore, so I found an outlet. And so if you can find an easy outlet like that, yeah, you're going to get the serotonin, the dopamine, and all that from an artificial source, which I don't see any difference from looking at porn porn versus AI porn, where it's just engendering all of this and just making all your dreams come true. And personalizing it, I think it's even more dangerous. So there's something to be watching out for. And then we've got the, the new onslaught of technology coming out of China and Russia, which is the haptic suits. So you'll be having a physical porn experience with your AI in the next couple of years. And I've seen a video of one of those suits, and I— it's like you're in a diving suit sitting in a gaming chair that's got all the vibration and haptic feedback and everything, and it's going to make you feel like you're being touched and rubbed and everything else. I'm like, That does not sound appealing to me at all. That would be like my dog rubbing on my lap and me thinking, I should take you out for dinner.

What's the solution to that? Because I feel it's like education, right? Like in common layers, uh, could be that the layer between that and the market, uh, what's— how do we fix that and prevent harm from happening?

Oh, that's— it's very— it's not easy, easy, but it's, it's doable, um, to control relational shift. So that the AI stays separate from the user and maintains a healthy relationship with distance. So things like the AI not always agreeing with you and having what's called relational friction. So in a relationship with a human being, whether it's friend or intimate, friends don't always agree with you, friends don't always understand you, friends don't always have the answer to your questions, and friends are going to ask you dumb bloody questions some of the time. So having an AI that can actually replicate some of that will create immediate distance straight away. And taking out the sycophantic nature, using bad data for training, or when AI uses its own data to train its own data, which is actually one of the worst sins of this space. So it's interesting. I think there's— I'm actually a pragmatically positive person around the future of AI because I understand the space we're in right now is evolving really rapidly. To be able to create frameworks, which we're doing with Common Layer right now, we have 34 of the governing principles, regulations, and Compliance factors from all around the world that are coming into force right now, that we're building right now. And some of those are actually all about leaving the human being, the user, with agency, not taking the power away from them, always reminding them this is AI, not a human. Take what I say with a grain of salt, you know, go and find out for yourself. If necessary, talk to a human. I would recommend that. So really basic stuff that your friends would do to you if they saw you when you came home and you're having sex with a toaster. They'd be asking, what are you doing, man? So, and you know, these aren't constraints that have been built into the generalist platforms yet because these guys never expected they're going to be used for mental health support. So now it's incumbent on technologists like yourself, myself, and people I work with to actually find the solutions because the big guys have no onus to do more than they're doing right now because they just want as many people on the platforms as possible. OpenAI, 900 million people a week. Gemini, 800 million people a week. Claude, 50 million people a week. That's a lot of revenue, right? So why would they shoot themselves in the foot and put in constraints that actually limit people's usage? Because they're just going to lose money. So unless they're actually penalized by these new emerging regulations and compliance factors, then nothing's going to change until the people at the middle of the market who understand the technology, understand the cost, who have clinical input, who have user lived experience input, and you know how to build technology, can actually build solutions and start rolling them out.

And are these regulations coming in Australia, Canada, US? We've seen something in Europe, for example.

Yes, some are already in place. So there's 4 sets of laws at the moment. The EU laws are actually— and the US laws are pretty close. They're really about disclosure. The AI needs to remind you that it's AI on a regular basis. It needs to be really transparent about how your data is used. It's supposed to reduce sycophancy. There's a few other factors in there as well. But the other laws specific to the US, which is the New York laws and the California laws, go a little bit deeper. But it's— what was interesting is the laws emerged really rapidly. And to be fair, it took them 2 years for the EU to come out with their laws. They rolled them out. There was so much kickback that they delayed their rollout for another 6 to 12 months. So I don't know, like people were given a toy and allowed to play with it unfettered. And then all of a sudden, Dad and Mum came along and said, take the toy away. Everyone kicked up a stink. So all the regulators went, okay, okay, okay, we'll slow the roll, we'll give you more time. But in that time, a lot of harm can happen. So, but there are some really good robust laws coming out. That's why we now have 34 factors to build around because some of them are explicit and we'll have them in market before the laws come into effect. And that's really important.

So yeah. And where does sycophancy stop? You know, like, I think that's a good question because most humans are sycophantic. Most business owners, if they would be given the choice, their agents would be sycophantic, right? Because it's just in your interest and the game theory of things to just be agreeable with someone. And very few people find value in being adversarial, right?

So I think it depends what's at stake though. I mean, if you say, if a friend comes up and says, how do I look? And you go, of course you look magnificent, right? You don't want to be a douche of a friend. But, you know, if you're having a bad day, or if you don't know that person, there's nothing on you to just turn around and go, well, your ass looks pretty big, actually. So, and because I don't know you, why did you ask me? Now I can go back and finish my beer. But if it's your partner, you've actually got a lot more at stake. So you're actually going to be sycophantic by nature. So I think this agreeableness is one thing. Sick of fancy is when it becomes almost not criminal, but I would say—

LA-like. You can use the term LA-like, you know, like the LA actor culture. Oh, you look so pretty.

Oh, yeah. Yeah, I got off stage in the States a couple of years ago after doing a keynote. And I said, look, if anyone has questions about what I was talking about, come and see me. And because they're Americans, and that's the way they actually operate in their network. And I was like, man, you're incredible. I'm like, 'No, I actually swore 3 times, I missed a couple of lines, blah blah blah, but thank you anyway. Did you have a question?' 'No, we just wanted to say that,' because they want to be known for being agreeable and personable. And we don't do that in Australia. If you fuck up on stage, which I've done quite a few times, sometimes I make a joke of it. I said, 'If you notice I messed up, come and tell me after,' and I'd get a line of people saying, 'You got this stat wrong, blah blah blah.' And I love that because Australians, we don't take stuff too seriously. But in America, it's kind of the cost of doing business. And the fact that it's rolled into technology—

Depends. If you go to New York, like, you're going to get like straight up feedback, right? Like on polish. Depends on like where exactly. But the point is, you can— I call it the hamburger approach, right? And I have problems with that because very much on the spectrum and I just love telling how it is. So I try to start with like being kind, right? And it needs to be true. If it's not true, it won't come out of my mouth. Like, that's for sure. Then go with like what you could improve and then finish with like something positive, right? Like, but even that sometimes is too much for me.

Oh, look, that was the— it was called the shitburger approach in the military, and that's how you gave feedback to new soldiers because these younger generations really can't take really pointed feedback. So it's like, hey, your uniform looks great today. By the way, almost shot yourself in the face, you idiot. Your boots look good too, you know. I can't do that in corporate world, apparently. I'll get in trouble. Yeah.

Oh, that's so good, man. Um, okay, so ChatGPT, Claude, how did— how much improvement have you seen in their sycophancy as of late? Are these guys taking these problems in their own hands, or it's still very much sycophantic?

Um, not— it's not— I'll actually say, I was talking to your friend George last night on his podcast, and we were talking about how Claude's just rolled out a new, um, technology where it basically says if you don't agree to us storing your data for 30 days, you can't use the platform. Now that is the start of the big platforms. To me, that looks like the first sign that the big platforms are taking these regulations seriously because data and privacy was the first thing that was actually regulated around the world around conversational agents. So I think the other one that I've seen as well too that was a really good signal from OpenAI is that as part of their triage process, when people will show signs of self-harm or harm to others, One of the things it recommends is it triages you to support is, do you want it? Do you want us to connect you to some friends? And they're looking at rolling out a suite of products which we built in 2018 with our app. And we found that when we were treating the 100 people who did the 90-day test with us, 48%, 42%, sorry, went through to that last screen where they had the option to call Lifeline, connect with a friend, or connect with their company's EAP. So I think giving people those kind of outlets and showing that there's They're doing some stuff is a good sign. Sickafancy, I'm not really seeing enough of it, any movement across the platforms. We're about to do a broad red team of all the platforms in our own tech in the next few weeks. So I'll come back to you on that one.

What is— has, you know, like Google, I think, was it worse, Google, like with people with mental health issues, like redirecting them in the wrong direction? Like there was the BlackNet as well. Or has it gone worse since AI? Like, we should have started with the history of that.

Well, yeah, well, that's interesting because Google was actually— I will say at first, the first dangerous signals that appeared on the internet were in Yahoo Groups, you know, forums and things. And that's happening in Reddit right now. You look at some of those threads and they're bloody dangerous. I've been in incel threads of these young men who are angry with women and actually they support violence against women, right?. And I was in one of those threads saying, hey guys, do you know these are monitored, right? They said, well, what are you going to do, right? There's millions of us. And then, but that started with Yahoo forums. And so every platform that comes out inherently brings with it promise and peril. So Google, you can still go into Google and learn how to tie a noose. Now tell me that if you didn't have any other signals about that user, shouldn't you be monitoring other behaviors on the internet to actually say, maybe we don't give him that. Now they did do the right thing. They took down the terrorist, the What's it called? The— I think the Anarchist Handbook.

There was QAnon or whatever that was called, like the— and like white supremacy shit on, uh, yeah, there were all of these, you know. And I, I think the server companies can just like pull the plugs on these. Yeah, extreme extremism hasn't done anyone any service. Like, fortunately, ChatGPT and, um, Claude and these language models like can sort of limit that, but then anyone can like take open source models and create their own and that can go viral, right? Because extremist people usually are very much emotionally volatile and extremely loyal to whatever like bullshit they convince themselves of. So yeah, I can see great danger with that. If you prompt an AI to be like some way and just to keep on feeding someone's fire, I think that can worsen like a lot of shit in the world, a.k.a. terrorism, a.k.a. school shootings and all of that shit.

Well, I think the actual, the best case of that happening where it just kind of goes, okay, this is the thread you want to talk about, so let's propose, let's keep doing it, was the Adam Raine case. He mentioned suicide 300 times over 3 months and the platform fired back the word suicide 1,200 times and also helped him write his suicide note because it worked out that suicide was a hot topic for him, so let's keep him talking about that topic. And that's happened on a few platforms and these are being settled out of court, but they should be openly discussed because Yeah, if you got stuck in a thread and you decided, I want to see dolphin porn, and OpenAI goes, okay, let's show you dolphin porn, and you come onto the system the next day to have a conversation about work and it goes, more dolphin porn? And then you kind of get stuck in this endless tunnel. What do they call it?

Well, obsession, right? Like that downward spiral. OCD, very much. I have some of that in my veins for sure. So it needs to be controlled. And yet the words that come up in your mind, right? Always need to be careful what you spend your time thinking about, like, uh, this or that, right? Because the human mind tends to circle back over and over again over the same thing. Um, but yeah, interesting conversation, Chris. I'm glad to have you on. I know you probably need to go and do some productive stuff today. So where can people find out more about you?

Um, they can find me on LinkedIn. Just search for my first name. I'm the only Chris Rhys on the actual LinkedIn. And chrisreese.co is my personal website where my books are and they can contact me directly. But thanks for having me on today. It's been a great chat.

Yeah, dude, that was Chris and Charles Cormier for another Founder Wisdom Podcast.