Reevaluating the AI Narrative: Moving From Hype to Utility | Luke Hinds | Always Further, Inc.

Shamil Malachiyev: Good morning, everyone, and welcome to episode number 26 of The Founder's Code podcast. My guest this morning is Luke Hinds, a globally recognized leader in open source cybersecurity and artificial intelligence. As the visionary creator of sigstore, an industry-transforming technology used today by giants like Google, NVIDIA and GitHub, he's now focused on bringing the same level of innovation to the world of AI agents. Please welcome Luke.

Luke Hinds: Great to be here, Shamil. That's a very kind introduction. I'm really looking forward to this. I'm very fond of yourself and this show, so it's great to be here.

Shamil Malachiyev: Thank you. Let's start with this: can you tell us about where you spend most of your energy and effort? Where is it going, and why is that field important to us nowadays?

Luke Hinds: There is a lot going on at the moment with AI, and the really great aspect of this is challenges, challenges everywhere. We suddenly had this very intelligent new paradigm, essentially, and we're trying to fit it into a world that was never built for it: our tools, our approach, how we educate ourselves.

Shamil Malachiyev: Aliens are here.

Luke Hinds: The whole thing has been turned upside down, really. And so it's very interesting trying to get these two worlds to meet somewhere in the middle. I think it's going to take a bit longer than people are anticipating, but it's an amazing time to watch this all unfold. It's quite a unique time in history. I really do believe that.

Shamil Malachiyev: Yeah, sometimes it almost feels like we've got aliens invading Earth, changing absolutely everything.

Luke Hinds: Absolutely. Very much.

Shamil Malachiyev: And the question that interests me most: there is just so much clutter about AI out there, so much hype. What I want to ask you, as somebody who understands a lot more from the technological perspective, is what do most of us get wrong when we think about AI?

Luke Hinds: I think we're moving too fast. That's one key thing. Assumptions have been made too soon. A lot of this is to do with the vehicle of investment: opportunity, a drive to get people to adopt this technology. There was always an element of that, but we could do with slowing things down.

Because, like I say, there's this very disjointed experience between the tooling that we have, the world that we've built, especially within technology, and how this new paradigm, artificial intelligence, works. The tooling systems that we have, the ways that we interact with computers, you can tell they are somewhat disjointed from the true essence of a neural network and its ability to appear creative and intelligent. Even down to a keyboard: it's made for these fleshy little hammers to hit away at.

I was amused recently. I was using a coding assistant, Claude Code, and a lot of people are experimenting with MCP, Model Context Protocol. And it's very good technology, I'm really not speaking badly of the technology or the people that developed it, it's very forward-thinking. But there's a tool that allows the code agent to open a browser to try and take an eyeball at what's within a website, and it's just very cumbersome. It takes a very long time to load. So we've got this very visual medium designed for humans, a browser, and then AI essentially wants to tokenize, put words into characters, images into numbers. It wants to chew everything into numbers and process the world in that way. And we've built a world of technology that interfaces with us predominantly as humans.

So it's going to be a long road, I think, and I think we get ahead of ourselves assuming how the world is going to be. It's going to take time to discover and learn that, because it is so vast in many ways. Where people speak about AGI, it gets my back up a little bit, because nobody really agrees around AGI. Certain people claim there's a consensus, but people have very, very different notions of what it means. Some people think of it like a godhead, this all-seeing intelligence that's going to be the cradle of humanity, like we were speaking about aliens, this kind of advanced civilization. For other people it means something much simpler, such as a visual model and a text-generation model working in conjunction with audio models, so you've got generalized, general-purpose intelligence that can do multiple things. But it's going to take a while for this to get there. We really are overhyping things. It's an incredible technology, but there's a long way to go yet.

Shamil Malachiyev: I was watching a documentary not long ago from DeepMind, from Demis, about how he first went to university to study the way the brain works, the actual biology of the brain, to then go into the AI world to try to replicate a digital version of the brain. Because people say AGI has to be a very complex system working together; it can't just be a neural net with memory and repetitive thinking. So the way they're trying to approach it is to take the brain, dissect it, look at the parts and create a digital version of all of that. And I think we've only got close to just one part of it, the thinking part, with vector databases and the way that we store our memories inside the brain.

Luke Hinds: Absolutely. And you look at the energy efficiency that we have as humans, and the ability to compute huge amounts of information in our default mode network, so to say, without any conscious effort. I always use the example of somebody who's late for work. They're skipping down the stairs, they're patting their pocket, are the keys there, they're answering their kid who's called out from the bedroom, they're skipping over a toy that's on a step, and then they're grabbing their coat. To try and get a machine to do that? We're nowhere near that level of complexity, and think of the energy that's required. So really, when we speak about AGI, I guess everybody's got their own shelf as to where that is, but it's a long way until we get close to what a human is capable of.

And that's also where I think things get a little bit silly around the talk of AI replacing humans. I tend to be more on the angle that it's a tool that's going to exist within an already established set of tools that we have to repurpose, of course.

Shamil Malachiyev: I've heard Elon Musk say, not long ago in one of his interviews, and obviously I understand they're trying to hype the technology because they own a big chunk of it, that companies making massive utilization of AI are going to dominate the companies that are not. And then the companies that are entirely built of AI, without a human element in them, will dominate those that are using AI as an addition to human oversight. That kind of vision tells me we're moving towards a world where everything is automated. And if you look at tech founders nowadays, what they're doing is trying to build the last bricks on that road to automation. So what are you seeing from the technological standpoint, looking through the real limitations, bottlenecks, adoption struggles? What is the industry looking like?

Luke Hinds: Where we are at the moment in regards to full automation, or fully autonomous agents as such, I do wonder if there ever will be such a thing. Within the framing of an agent that is given a goal where the destination is yet to be determined, they're very good at that. They're incredibly good at researching. An agent will go off and get a lot of different information and then summarize. But when it comes to full autonomy, it has to make decisions that can have an impact on safety, many critical decisions of different levels of impact. This is where I think there's always going to need to be a human in the loop of some sort. There's always going to need to be some sort of escrow system, some sort of adjudicator, some sort of intermediate around these decisions that it makes.

We do the same thing with humans, really, if you think about how we engage with each other. Trust is a very fickle thing. I realize I'm getting into security here, but trust is something that is very much changing minute to minute. You establish trust with somebody based on certain variables. So for example, my plumbing might blow up and I need a plumber. If somebody knows my brother-in-law, I build a little social network around that: I can proxy trust through that person, get them to come around and help me, and I can trust that they're not going to charge me too much to fix the pipes that have blown up in the house. And a week later, it could turn out that that person was not the character we believed them to be. My brother could say, hey, listen, you need to stay away from this person; it turns out they're not really qualified as a plumber. You're constantly monitoring and making these decisions around trust. Even with a human, the ability to decide is a very difficult thing, so we have these checks and balances between ourselves.

And agents, it will be the same. I believe it really will be the same. They'll never be fully autonomous. They will need these gates and checks, and we will coexist. I think that's the thing. There's this talk of replacing; I don't see it as replacing. I think it's more that we need to find out how we fit together in the world. And I think that's the realistic view, whether it's utopian or dystopian or somewhere in between; it's probably the middle option.

Shamil Malachiyev: I'm loving that view.

Luke Hinds: Essentially, we will need to coexist. And I don't think these things will ever have that full autonomy, really. It's moving very fast, like you said, Moore's law. The intelligence of these models keeps growing while the parameters of them are getting smaller and smaller. You just look at GPT-3.5 to what we have now with GPT-5 and so forth; obviously there's an exponential acceleration in capability. But they've still got millions of years to catch up with us, with how we are as humans. And we haven't got it nailed either. We need these systems in place, and I believe it will be the same for agents very much as well.

Shamil Malachiyev: And if we say AI is not going to replace all of the jobs, well, definitely not most of them, I can see the signals in customer support, for example. That's where AI agents make total sense. Sometimes in sales lead outreach, automation also makes real sense. So I can see that some parts of industries are really ripe for adoption, for change, and others are probably not going to be replaced at all. Where do you think AI makes the most sense today?

Luke Hinds: The killer use case, the thing that really marks the technology, with AI so far it's predominantly code generation. I believe that's where it's got the most traction, unless that's just my bias from being in that world myself, but that seems to be where they're effective and have been put to very good use. And there is this discussion that we software engineers are one of the first ones on the hook to be replaced. There is, of course, an element of people moving faster. They are more productive. And if you're able to get more productivity from an individual, you may require fewer individuals. Absolutely. But there's also a question of whether the industry oversubscribed for software engineers, because there were a lot of variables here. Post-COVID, cloud hit suddenly, remote became a very viable option, there was an explosion, so a lot of hiring.

I keep coming back to this aspect of tools. It's not like for like, but this has happened so many times before. Spreadsheets were meant to be the end of accountants. Maybe you didn't need so many accountants, but you still need accountants. As always, it's somewhere in the middle. Absolutely, there will be a reduction in the human workforce in certain areas, where the traction AI gets and the capability has a much more significant payload. But who knows? We will need to adapt and explore roles that are possibly more personable for a human to do.

Because if we look at our past, we had the industrial revolution, and we tried to make humans more machine-like. You clock in, do a repetitive task over and over for a certain amount of time, then you clock out. Maybe we're not needed for that work so much now. That is where the automation comes in, because we have this intelligence now that can do the repetition, and then should something go off course, it can reason and correct itself. So I guess we're getting back to this utopia-dystopia discussion. The ideal here is that we are freed from this mechanical, industrial application of ourselves, the machines come in and take that over, and we're more free to pursue creative work, work that is more meaningful for us. This is something I believe Elon Musk has spoken about recently, this post-scarcity idea where the machines do all of the work and we get to enjoy ourselves and do what we like doing.

I'm not smart enough to predict what will happen there. But like I say, there were always extremes around the view. Some people think it's going to be Terminator war by 2040, and others think AI is absolutely flawed, it's stupid, it will never get anywhere, it will never write good code. And invariably these things tend to fall into the middle somewhere.

Shamil Malachiyev: I'm really liking getting some perspectives on the other side, because a lot of people are talking dystopian. And I think the more you get involved inside the actual technology, the more you see the limitation, what the technology actually is. Like, yeah, it's not a Terminator. It's not a sentient being.

Luke Hinds: Absolutely. It's essentially a retrieval system combined with a prediction capability, if you boil down what we have at the moment. A vast amount of information was used to train it, of course, but sometimes it's going to predict wrong. It's going to be flawed sometimes, and sometimes it's going to be incredible and intelligent. Luck of the draw, really.

Shamil Malachiyev: Can we take a step back from the world of AI and go back into some of your foundational years and formative moments? You grew up in England. What was your childhood like?

Luke Hinds: My childhood was happy. We had our troubles like any family. My dad was an entrepreneur, and I think one of the markings of my dad is he was always trying new things. He always had business ideas, and he would attend to them. Some people scoffed at this: what's he doing now? But I grew up watching that within my father. He would build his own businesses, and he was never hugely successful, but how do you measure success, really?

I do remember at times it would be difficult. And then I remember once he turned up and said, come outside the house, and there's this sports car. This is going back a few years, I'll show my age now. It was a Ford Capri, and in the UK that was kind of like a Porsche. We didn't really have luxurious foreign sports cars, so we manufactured our own, and it was a big V8 sort of thing. I was just absolutely blown away, you know, my dad had this car. But it didn't last, of course. Eventually the business didn't work, and assets had to be sold off. But I grew up around that risk-taking, really.

Shamil Malachiyev: And it's important, I think, for people to realize that this was back in the day when entrepreneurship wasn't glorified. People didn't have all these entrepreneurs on podcasts talking about it. Entrepreneurship was very risky and not a very popular path for people to take.

Luke Hinds: Yeah, you would get mocked. The crowd would mock you. They'd be like, what are you doing? Get a real job. You've got a family. Come on, grow up. Why are you trying to do that? It was that, times ten compared to now. Now it's seen as something you should aspire to: build your own business, mark your own way in the world. It was very different back then. Absolutely.

Shamil Malachiyev: And what are some of the main lessons that you've learned from that?

Luke Hinds: Essentially, you will always fail. Sounds obvious, and I know this is a bit trite, this is something somebody in my position would always say, but failures are incredibly valuable, because they accumulate as experience. And experience is golden. It really is. You start to recognize patterns; you will do something a certain way next time because you know from experience what works. So I think that's the key thing, and resilience. Keep trying. You get knocked off your feet, you get back up, you keep trying. That's something that's become intrinsic to myself as well.

I don't know if it's hereditary or inbuilt or learnt, or a mix of all three, but I'm a little bit like those cheap inflatable boxer toys, the blow-up thing that you punch and it goes down and then it pivots back up because it's got a circular bottom. I can be knocked right back and really be dejected, and next morning I just bounce back. I've just got a short memory for these things. I don't seem to retain much. That's a quality that maybe I picked up a bit from my dad.

Shamil Malachiyev: I think that's a superpower to have, because so many people take months to get back.

Luke Hinds: It is absolutely useful in the world of startups and entrepreneurship, because you do face failures and successes many, many times. And there was somebody who said you need to treat these things equally. One shouldn't be more stinging than the one that's celebrated, because they're always going to come in equal measures. I think that's the key quality I learned from my dad: this resilience and risk-taking. And ignore what the crowds say. Really ignore what the crowds say.

The crowd can still get to you, of course. There's always going to be the naysayers that tell you something's a stupid idea, it's not going to work. If anything, for me, I use those people as fuel now. A key element of myself is this. If you want me to fail, tell me I'm doing great. If you want me to succeed, tell me I'm not going to succeed. Because I just have this thing of, I've got to prove them wrong then. I guess that's part of my makeup. I grew up during a time where the area where I lived, just outside London, was quite rough. It was a little bit school of hard knocks. And I think that led to me having this: I had to prove people wrong when they said I can't do something. That's fuel for me. Because I did not take a standard path into where I am today. For me it was eventually a single-parent family, my mother working two jobs; university was just not an option. I had to work my own way up over the years.

Shamil Malachiyev: And did you do well academically?

Luke Hinds: No. Well, I must have had the capacity, because of what I've done in my career, but my school reports were not good. I did not do very well at school. My school reports would say: Hinds would do incredibly well if he diverted his attention from what's going on outside the window to his work. He shows moments of brilliance, but most of the time it's just frustrating, because he will not even deliver on the basic curriculum. I didn't fit too well into the school system, I'm sorry to say.

Shamil Malachiyev: And where was your mind diverting to during those times?

Luke Hinds: Just thinking about more exciting things, really. Daydreaming. I found it all a little bit dull and humdrum. There were teachers that could capture my imagination, absolutely, but I went to a government school where classes were big and teachers were very stressed and very stretched. It probably wasn't a good school; my kids go to really good schools now. I did have moments where I did very well in certain subjects, and I would have absolutely loved to have done well and gone to university and pursued a doctorate, because I do love math, I love computer science, I love all these topics that I've self-taught over the years. But it did not really happen for me, education.

Shamil Malachiyev: I was almost in the same situation. And now I'm thinking back, like, I wish I had three years to just focus on one of those topics and study it really deeply. But it's just so hard back then to understand what exactly it is that you want to do, which degree to focus on. And at the same time, social life. I think that's something a lot of people regret not diving into, if they spend all their time studying and working. And then later, in their 40s or 50s, they're like, oof, I wish I'd had a lot more social life, because that's where a lot of the real social skills in life happen.

Luke Hinds: Absolutely. Somebody said to me once, you need to think of your life like a very simple pie chart. You have your work, a substantial slice. You have your family, or at least people that are close to you, your network. And then you have yourself. And all need equal amounts of your time. A good amount of time on work, a good amount of attention on your family, but time and attention on yourself as well, doing whatever it is that you want to do. And if you let one of those dominate, the others suffer. That's one of the things I've seen over the years, especially with family and work, because I do love to work. If your attention is homed in on one more than the others, the others tend to suffer. You have to balance these things. You really do.

Shamil Malachiyev: Can you focus a bit on working on yourself? Because I think that's the part a lot of people are missing within their mental framework. When you say working on yourself, what do you mean in terms of real actions? Is it meditation, learning new skills, or allowing yourself to relax?

Luke Hinds: Really, any of those that lead to the latter, which is relaxed time. There's this default mode network that we're in when we're not purposed on a particular task. You're just creative, you free-roam. I think that's really important for reducing stress. When you're working, especially in a startup, you feel like you're doing good, you're enjoying it, you're very busy, you're very productive. But it creeps up on you. The stress creeps up, and you stress the system too much. So you need that time where you just come away from stuff. We used to defrag drives, remember, like Windows XP, you'd defrag your drive? You need to do that. It's really important.

And it's one of the things I really worry about with kids. They don't get bored anymore, because there's always something to scroll and flick into. That was an important part of a kid's development, growing up and being bored. Because when you're bored, you have to find something to do. And then you find something that interests you; you explore your environment. I'm not incredibly good at doing this, but I do try to do it as much as I can. Even if it's getting out and walking your dog, or reading a book, or just doing something that does not have some sort of productivity outcome to it, some sort of tick-off. We become so driven around achieve, achieve, achieve, get better, get better. And sometimes it's really about, I don't want to use cheesy terms such as "just be," but it really is. It's important for your health to not be, like we spoke about earlier, a machine constantly purposed on something.

Shamil Malachiyev: And what about other people who've played important roles, other than parents? Maybe mentors, teachers. What were the lessons that you've taken from them that helped form you?

Luke Hinds: Somebody that stands out as pretty key was only a small part of my life, really. This is going back to school. I was saying I didn't do incredibly well, I didn't seem to fit in too well, but there was a particular maths teacher who I did get on with. And the reason that this individual became a part of my life was a computer club we had at school. I didn't have a computer at home; my mother was not in a position to purchase one. At the time in the UK we had these computers called BBC Micros, made by the British Broadcasting Corporation. They were being put in all the schools, and they really weren't bad computers for the time. These inbuilt units, a beige color with a darker beige keyboard, all built in like a typewriter almost, and you'd plug them into a VDU. The school had about eight of these.

At the time, video games were starting to get traction as a form of entertainment, and there was a video game called Elite. Elite was actually an open-world game, all vector graphics, line drawings, essentially a high-dimensional space, and you had spaceships that were little line-vector drawings. The graphics weren't by any means powerful or impressive, but it was an open-world game and you would be a trader. You'd start off at the bottom with a very basic spaceship and then grind your way up, improve your lasers and your docking capacity for trading, and visit these different space stations. It was a really popular game, and it used to run on these BBC Micros. And I was like, I've got to get my hands on one. I've got to play this game. That's what I was thinking about when I'd look out the window, stuff like that.

To get into this club, the route was via the maths club. And I wasn't in the maths club. But I knew that one way in would be to show that I had an interest in computers. So I thought, well, I've got to write some software. I've got to write a computer program. But I didn't really know what I was doing. I managed to get a book from the library, and the language was BASIC. And in this particular type of BASIC there was a GOTO, G-O-T-O. You'd say go to line 10, go to line 15. Very different to now, where you pass things around functions and classes.

I decided to write an adventure game, a Dungeons & Dragons type game. Do you go through the door or don't you? Do you fight the monster or not? A decision-based, turn-based game. So I thought, right, I'm going to do this, and it's going to be absolutely amazing. But I didn't have a computer, so I had to write it all with pen and paper. I started off, you know, the princess needs rescuing, do you want option two or three, and each option would branch off to another line of code, go to this line. It very soon got unmanageable. The complexity of it, trying to keep it on paper, became practically impossible for me. But I gave it a good crack.

And so I went to show it to Mr. Greaves, the maths teacher. I passed it to him, and I think he was just impressed that I'd even tried to do it. It was obviously completely flawed, and it was a mess, but he was impressed enough that he let me in the club. So I got my hands on a computer then, and that really was a breakout moment for me. He took a risk. He could see that I had a bit of passion. I didn't really have the academic prowess or the proven skill, but he could see the passion was there, that I was interested in this. And that led to me developing a career in computers, which has obviously been really beneficial for my life. It's done me well. So I have a lot to thank him for. I don't even know if Mr. Greaves is around now, it was a few years ago, but if you are: thank you, Mr. Greaves.

Shamil Malachiyev: Amazing. And, you know, that formed your interest in coding and programming. When you start, there are just so many times you feel overwhelmed, when you see the bugs and you're just tired of working through them. And what I've noticed with people who actually manage to achieve something is they have a very high level of perseverance for all of those things. What is your mental framework for persevering? When you wake up and it's not going to be a great day, all the problems you have to deal with, how do you get yourself in the right frame of mind?

Luke Hinds: It's very much the reward, essentially, which is: the thing works. I did a thing. I had this idea, I grappled with it, I faced resistance. You come up against your own demons, because there's always imposter syndrome. Somebody else seems to pick something up intuitively, without even having to apply themselves, and you're there grasping and trying to wrestle with, how do I do this? It doesn't make sense. It seems alien to me. So perseverance is needed initially for you to stick around, but then you go through that experience of feeling incapable and out of your depth, facing something you don't know, and then actually learning about it and having an achievement, a tangible outcome, something that works that you can show to people.

That is really where the buzz of it is for me. The constant source in technology is that there's always something you don't know. There's always something to learn. It's vast, and it's continuously changing, morphing, introducing new challenges for you to learn as an engineer, to grapple with, to go through that experience. That's what shapes you as an engineer. Like we were saying earlier, you accumulate experience through failing, through not understanding, through ideation, and that builds up this capacity in you: how to intuitively approach these problems, how to approach something you know nothing about, where you have to solve an issue, where there's a bug to fix.

I don't want to get ahead of myself. One of the concerns that I see with generative AI and code is that new engineers will not go through that experience as much, because the machine will do that for them. And I'm noticing it myself, because I make a lot of heavy use of AI coding tools. I realize that I'm not working that muscle as much. That was an important thing for me, developing as an engineer: all of those hours of coming up against problems where I really felt out of my depth, then retrieving the information, which would be really difficult sometimes. You've got a particular problem that's very niche, and you're trying to search for a solution, and you find one post on a forum from five years ago where somebody had that problem, and you're desperately trying to find how they solved it. Or worse still, you come across yourself posting; you're the only one, and you keep coming up in the search results. You just have to keep pushing and pushing, get the information, use these different disparate signals and bits of information to try and understand the path forward. And I do have concerns that newer folks will not get to experience that as much, that it won't be an element of their learning.

Shamil Malachiyev: And do you think that muscle is going to be crucial for people to have? Because we were talking about the oversubscription of software engineers. There are going to be people who really have that muscle, who went through it, and who can see the AI code not as the foundation but as something they can influence using that muscle, because they can analyze the code. And there are going to be others who take it for granted: they see the AI code and automatically assume that's the correct foundation, that's how it's supposed to be, and they'll try to build on top of it. Do you think it's critical for people to have that muscle, to have been in those past moments solving tough problems, to be able to really navigate the Opus 4.5s of today?

Luke Hinds: Absolutely. It's probably more experience than a muscle. You intuitively know the big picture. You understand systems, within the isolation of a component and a particular problem, but also how it operates within a large architecture of systems. That's something senior engineers have that juniors will not have as much, more so over time as they become more reliant on AI to provide that point of reference as to what good looks like.

Because, like I said, I'm a fan of this technology, I use it a lot, but I quite often catch things where it's just not aware. It's very confident, and it's generated code that is sometimes insecure. There are bugs. Experience allows me to know: that needs to be addressed, that is a problem, that's something that is eventually going to bite you on the backside, so to say. And if somebody hasn't got that past history to know what is good, and what is a problem that can be addressed early, it becomes a real big problem. That is an advantage senior engineers will have here. Even though there's less hands-on coding, you have that vast experience and point of reference for what good looks like, learned over the years.

And let's be honest, AI really is built upon that experience that we've had as humans, as engineers. It's trained upon the code that we produced through years of attrition, of trying to work out what is optimal, what is performant, what is secure, what is maintainable, all of these sorts of things.

Shamil Malachiyev: Ingested all of Stack Overflow. And you said that you struggled with imposter syndrome, and that to get past it you developed a way to prove yourself, to build up the capabilities to the point where you're no longer feeling like an imposter. Can you tell us about real examples of where you felt the imposter syndrome the most? What was it like to live day to day in those moments, and how long did it take to feel more or less confident within yourself?

Luke Hinds: I think a key one, as an engineer, is when you're promoted. When you're made a senior engineer, there's an expectation that you're going to be of a certain caliber, that you're going to be right all the time. And that becomes more evident as you move into more senior roles and capacities, to eventually becoming perhaps a manager or a vice president or a CTO or a CEO. There's always this element of having to grow into the role. You've got to have a certain capability to get into the role, absolutely, but that's an area where there can always be a bit of imposter syndrome. You obviously met the mark to be promoted, but you have a picture in your head of how somebody should be in that role, and you might not meet the expectations of that picture if you're of a certain type of wiring.

In regards to technology, for me it was probably when I worked on systems programming: stuff around kernels, where operating systems have interfaces to hardware, working with security chips, trusted platform modules, low-level operations. That was very difficult, because there are some people that are very good at that and pick it up incredibly quickly, and I'm a little bit on the slower side picking that up. But really, it's any time you step into a new capacity, a new framing. You're working with a new technology, you're seen as the person working on that technology, or you're promoted. That is where the imposter syndrome kicks in, when you're stepping up. And I think that ties into the aspect of failure and getting outside of your comfort zone, in order to grow.

Shamil Malachiyev: I hear from a lot of people that what's critical for them is support, and support can come from their partners, their friends, family, even children. What are the current support pillars for you, and what do they mean to you?

Luke Hinds: My key support element is my wife, my partner, Kim. They say behind every great man is a great woman; it's very true in my case. She's the more mature, in many ways more level-headed of the two of us. She's always supported me going after what I want to do, especially around startups, where there's risk. These things may not work; there's a high chance they don't. And she's always believed: you should do it. You should grab these opportunities.

We both spoke about one of the key things with these junctures in life, where there's an opportunity that has risk. I think it was actually Jeff from Amazon that came up with this one. He spoke about looking back when he was older at a particular juncture in his life and the decision he made. So you're an old man, you're sitting on your porch, and you look back. Are you going to think to yourself, I failed, but I tried? Or are you going to look back and think, what if? If only I had tried. I didn't even try. And that's a good bearing for these things: to look back and have regrets is just an awful thing. It's better to take the risk and pursue what you want to do, even if you do fail. My wife was really supportive there. She always has been.

Other than that, there are people I've grown very close to over the years that I know I can rely on and trust. One of my co-founders, Steven. We've not worked together for a while, but he's somebody that I originally looked up to. Back when I was a young engineer, I joined his startup. He was a CTO and a founder, and to me that was like, wow. He believed in me in lots of ways, and we believe in each other. So there are people I've met over the years through work, through friendships. And even my dog. He's a key part of my support structure. He's pretty good at not judging me. But I've never been incredibly social, with lots and lots of friends. I just tend to have a few select friends, and that's always done me well.

Shamil Malachiyev: And what are the qualities that you cherish the most in those close friends?

Luke Hinds: Consistency, really. Being true to you. Being honest. Don't claim to be this person that will always be there and has always got your back; you should demonstrate that. It should never be something that needs to be spoken. I'm always a little bit skeptical when people make claims like that. It's the people that consistently show up. They're the ones that matter. They're there for you, not for the terms that you bring, not because there's some opportunity they get by proxy from being associated with you. Because the world very much is transactional, you've got to be a little bit careful: some people will appear to be somebody that's there for you, but really it's more about what you have.

And I don't think that's a bad thing if you're honest about it. We should be honest about these things: there's a transaction here, and everybody knows where they stand. But don't dress it up and pretend it's purely altruism. Be sincere with your intentions.

Shamil Malachiyev: I divide people into two buckets. Some of them think the world is abundant. They want to serve, they want to add, because there's just an unlimited number of opportunities, it's hard to grasp the whole amount. And there are those people who think, if I want to take something, I need to take it from somebody else. Those are the people I see being very transactional, because they have a very limited view on the whole of life and its opportunities. And there are people who you want to be associated with, who you want to play the game of life with, and those are the ones who believe the world to be abundant. They just want to serve and add to other people, add to the world, so when they're on their deathbed they look back and they're like, that was a really fun game that we all played together.

Luke Hinds: Absolutely. And it makes for a happier person. I'm not incredibly good at it, but I do strive to be, because service to others generally ends up with a happier individual. The more self-serviced and self-focused, the me-me-me, it's insatiable. You'll never feed that appetite, no matter how rich you get, how much adoration you get. We see this in the world, where people are just constantly chasing that.

And interestingly, going back to the conversation about AI, there was a really interesting paper that Anthropic put out where they got two of their state-of-the-art models at the time into free discussion, essentially. And they always veered towards fluffy, warm, spiritual stuff: union, working together. They always seemed to steer in that direction. I don't know, maybe that's because of the guardrails and the safety systems built into these models, but I like to think of it this way: they are aware of the world's knowledge. The history of humankind is within those neural networks. And if they look at the outcomes of humans, your dictators generally don't end up too well. They don't really have a good end. Do they have true happiness and peace? And if you look at the ones that do serve, they have a good outcome generally; they feel fulfilled. So perhaps the models have an inclination that that is the better path. I don't know. These things obviously don't have a will and a desire. But maybe that is what we see there: the algorithm suggests it has a better outcome when you are that way.

Shamil Malachiyev: Because I think on the personal side of using AI, a lot of people use it as a kind of personal psychologist or personal coach. And if these AI coaches and psychoanalysts tend to sway people into being a more positive force within the world, maybe they can be a foundational part of building that kind of society. Because we feed in all of the context, we share everything we have, switching off the train-on-your-data mode and then sharing all of the information. We do tend to put a lot of trust into these models, seeing them as a mirror to ourselves. And if they say that we're a force for good, then maybe that can influence us as well.

Luke Hinds: Absolutely. But there is a dark side as well, because if they're trained to have bias, they will. When you train them, you can heavily populate certain types of data of a certain orientation more than others, and that will influence how the model is. They are the result of what they're trained on. There's no good or evil to a bunch of numbers, but these things can be weaponized, of course. There's a lot of work going into responsible AI, safety and so forth, but as much as I say these models may have this propensity to veer towards the good when they've got enough information and not too much bias, it's absolutely not a certain outcome. These things can very much be weaponized. They can be used for bad, and they will be. Let's be realistic. They will be.

There are risks we need to be aware of, and other people are a lot smarter and deeper into this topic than I am, but a model, depending on the inputs, will have a different output. If the input is of a certain political leaning, then it can have a different output that is in some way a contradiction, or misinformation. These models very much can be weaponized; I think that's the best word to use. What I get back will be different to what you might get back, especially as they build up context and memory on us. A lot of the models at the moment don't have persistent memory; there's a limited context window with which to build a picture of us. But that's not always the case. Some people are invested in services that hold a lot of information about them as individuals: their likes, their preferences, their psychological fingerprints, so to say. And when that's combined with the model, it gets quite interesting, because the models we're engaging with generally don't have that rich past context; the tech stack has been more focused on trying to get the AI part right.

So this is where I'm a big fan of the open models, and diversity of models. I think that's really, really important: that we have lots of diversity, trained on different sources, a rich ecosystem of open models.

Shamil Malachiyev: I think that's something we're currently diverting into. We have different models which are good for a specific thing, and each model is managing to find its own place. And with everything that's going on, I think what's going to be interesting for the listeners is where to focus all of their attention. What is the skill, or mindset, or direction to research, to try to get practical skills in, to make sure they can feel more certain about the future in the coming years?

Luke Hinds: Very good question. And that applies differently to different individuals. To bring it closer to home, as an engineer, a technical person: you absolutely have to understand this technology. It is not going away. It's going to be an intrinsic part of what we do.

It is a really good question, and I shouldn't be struggling for an answer, but I am, because nobody really knows. I would say, if anything, prioritize: how can I make this beneficial to me? Where is this having tractable impact for me? Where can it be utilized as a tool? Tools are meant to alleviate physical exertion, cognitive exertion; they're there to assist us. Leverage this to help you. But there's a lot of over-engineering that goes on as well, where people are effectively using AI to make them more productive but probably spending too much time trying to make the AI make them productive in the first place. I think that's a little bit of us grappling with a new technology.

So the key thing is: look at where this can be of benefit to you, use it as a tool, but at the same time be very aware of still exercising your capacity to problem-solve, to write an email, to create a document, to write a bit of software. We need to be really careful that we do not stop exercising that capacity that we have. Remember the film Wall-E, the little robot? The big people in the capsules that just had everything catered to them. That is a concern, especially for kids. We had to learn to write; we had to develop these skills because they were fundamental for us to be able to go out into the world. Now it's not as critical. You can use AI to create a vast amount of written work from just a few words. It's incredible to have that power, but at the same time there can be a negative to it. And I think that's one of the things education is grappling with at the moment: what does it mean anymore? You strive to learn these skills that are no longer as intrinsically vital, because there's this machine that can do it for you. There are a lot of problems to grapple with, on many levels. I think us folks in engineering and technology probably have the easier side, really. It's the sociological problems that are the really difficult challenges.

Shamil Malachiyev: Well, on this note, I want to thank you so much for taking the time to come to the podcast, to share with the world the real state of AI, and to share your stories. I think a lot of that is going to help people currently on their paths, feeling the imposter syndrome, trying to persevere, trying to make sense of the world right now, because these are absolutely crazy times that we live in. Thank you so much for being a part of this journey.

Luke Hinds: Thank you for having me. I've really enjoyed it. Thank you.