Latest / Hardly Artifical Podcast / Are We Ready for This?

Transcript

Are We Ready for This? · Hardly Artifical Podcast

• Derek Major: Welcome to Hardly Artificial. We explore tech and the human experience. Sit back, relax, and let's dive in. Okay. All right, everybody, welcome to episode number three of Hardly Artificial. I'm Derek Major. Mike Bukowitz. And we're live in well maybe not live for you, but we're definitely live ourselves here in San Francisco at the AI conference twenty twenty six. So today we're just we're doing it in person instead of vir virtual like we typically do. lots to cover off, of course, and lots of inspiration we've had this week too. So we've been here for the last two days and today's the last day of the conference. So lots of cool things that we've seen, different ideas, all kinds of vendors, like companies you'd never heard of. It's funny when I talk to like all these people, I'm like, Hey, how many years have you been in business in like four years? I'm like that's that's the nature of this business though, right now. Totally. You know, a lot of the a lot of the products that we were talking about, it was interesting, you know. one guy was just telling me about how they're so specialized now that they've actually shut down a arm of their business in the medical world to basically focus on transistors and being able to develop the software chips to support the technology, which the fact they're doing that big of a shift tells you everything's shifting right now. Yeah. No, it's crazy. Like it's just interesting to me how you know, everyone's got a different angle with it. And the one of the sessions I was in today, how they talked about The the first phase was a chat bot and then it became coding and now they joke hacking. but I think that, you know, there's lots of things that are coming along there. But so I maybe let's kind of start there a little bit on on that topic, like before we get into our stories. But what's one thing that's surprised you in the last two days? I think what surprised me the most is you know, th everybody's so hyper focused on the technology that they're not really thinking about the humans that are gonna be interacting with it much. Now there has I'm not saying there hasn't been like you know, accounts of hey, you need to talk to the you talk to the user. But there hasn't been a whole lot of how do you do that. Right. Or like what are some techniques, you know, things of that nature. I think the biggest thing that I'm realizing though is I don't know if anybody really knows an effective way yet because it's still pretty fresh, still pretty new. But there's people out there thinking about it. Like the last session I just went to, he's thinking about it. Is he doing it successfully, continuously, repetitively? Yeah. Not yet, but I think we're getting there. Yeah. What about you? I was surprised, I mean, you know, if you're thinking about going to this conference, you know, we talked about this is it's very technical. I compared to when I did Saster AI back in May. The thing that stood out for me is that it's very technical. and there's like a whole section of companies that are doing like governance. And I'll talk a little bit about that in a second, but AI I was surprised how many companies are trying to do governance. Because I mean we were joking about it at launch It's like AI really is and I've no really I'm really you know, dumbing it down here, but sure you're throwing text at it at an L L and it's spitting back text. Now everyone's like, Well, how do you monitor that? How do you manage that and all that? So I was you know, that that really started to me as how technical and how you know people are just trying to solve something so simple, but could have such implications on a v wide variety of things, right? I think I feel like that's leaning into the commonality of what happens with everything in this world. And that is, you know, the overengineering because they're concerned about every little thing being perfect. And I I I don't know, and again, I you know, I've talked to a lot of vendors here, I've listened to a lot of talks. I don't know if anybody knows what level is right to get this productive. Because there's so many risks that are unknown at this point. Like, you know, we brought up a while back the hugging face incident. That brought got brought up how many times in your sessions? Like Yeah. Every time every time somebody's talking about it. So surprise nobody's sued anybody. Yeah, yeah. But like that's the thing though, is like everybody's concerned. They're bringing it up as like a data point. And then everybody just kind of goes, Yeah, I know about it. But nobody's going, Well, hold on a second. Are we really ready for this? Like I don't know. Yeah, exactly. And I don't know if we are. And it's it's you know a funny I told you that my funny story, I again come from Canada and the States. You know, even even the the the border guard to me, you know, like Right. You know, I went up to the border to the border guard like you do and they asked, Well, where are going? And I told him I'm going to San Francisco. And he says, Okay, what are you doing there? And I'm like, Well, I'm going to an AI conference. And then he looked at his screen, said, okay, I'm good. I'm done here. And then he he sits back in his chair and he's like, So tell me, Derek, think that AI's gonna kill us? You know, like Terminator days or what? And I'm like, first I'm thinking is this a trick? Yeah. But then we ended up having an interesting short little conversation. I'm like, I really don't know. It's like he says but obviously my point being is that it's on the border guard's mind. It's not everybody's mind. Like they're all a little worried about it. And I think I I read a poll yesterday that in the United States, eighty percent of the population believes that AI should development should be stopped. Which is significant. You know? and I'm sure there's what you can dig into the details of what that means, but I I'm maybe it's not completely full stop on everything. But yeah, I think people are the the point is is people are concerned. Yeah. So I mean, you know, seeing all seeing all the data and seeing the adoption, seeing the failure rates and we'll get into that more of that here in a little bit. But yeah, I you know, I get it. If you don't if the unknown is scary. It really is right now. And I don't I don't know how to comfort you guys here listening about what's gonna happen. I I can tell you right now though, there is good things happening out there and I think we're gonna get into some stories that'll share some of that perspective here today. Yeah, maybe I I know one of mine might be kinda fifty fifty, but Okay, three of the four. Exactly. So Okay, hey, well, hey, why don't we segue into that? why don't you kick it off with your first story? Go ahead. Yeah, you what I'm gonna pull that up here because I wanna make sure to quote this correctly. But I was on I was going into some of the the meta rounds that are happening right now, and I'm going back to the goggles specifically, and I know that that's been a topic that you even brought up recently. And what's fascinating is they're actually doing a connect to hearing aids now, which I think is fascinating. So around September twenty third, they and I'm just quoting this right now, and they announced that A glasses will double as a hearing aid. Yeah. In July, they had the FDA clear it too. So that means that not you know, in clinical testing, 142 people have performed as well as hearing aids fitted by professional. Nobody needs a clinic visit, nobody needs to worry about like medical. It's basically, you know, US dollars of 149.99 and being able to be HSA and FSA, which is the health health spend account, eligible, which is huge, right? So my question to you is, you know. The device basically doesn't necessarily replace hearing aids, it supplements it. So what do you think that's gonna do? I mean, I I personally I could foresee the folks that and I'm gonna speak on my behalf of my parents, you know, they're already at that age where they're starting to look at, should do I need hearing aids? Right. Right. What if that's is that changing the stigma or is that just putting another grouping towards something that may or may not stick? I mean, it's still new. We don't know. You know what that pops in my head is like AirPods are, you know, especially now with can't I can't recall I can't recall off top of my head, but I know I don't know if it's the new ones that just came out or the AirPods four which were out came out two years ago and it's more of a software thing, but translation was one thing. But I've heard that some people I think it might be approved. Well, you what? Maybe it was released two years ago and I think it's recently finally been FDA proof at least in the States. And For exactly that. So I can see a lot of people applying that. I've always thought though, like do you want your like you want to be wearing AirPods? 'Cause there's that thing if I if you have your AirPods in and I'm trying talk to you, there's always that part is he actually listening to me? Is that a generational thing or do you think that's everybody? Yeah. I I don't I don't personally like but when I see it, I think the same thing, right? Like, you're not paying attention. So you think you think can my kids one day will be like, whatever, he's got airpods in, he's listening to me. I mean You know, you brought the I'm I'm an Android user. Yeah and I'm I'd use a Mac. I'm I'm I do a lot of this. But the Google Pixel also has that similar with the Pixel Eleven release recently. Yeah, there's the option of doing the live listening. And I think I can't remember it's not called Rambler, that's another thing specifically. But basically you can hear the translation, right? live now. It's not as a delay or a thing that happens we have to submit. So it's a similar function and maybe it's gonna start being accepted because Yeah. Maybe it's a commonality that folks are gonna have and really, I mean, they're let's be honest, the Bluetooth are starting become very yeah n non noticeable. Well and I wonder well, I and I I know people that use hearing aids. Like I got a friend of mine that uses them and he and he uses not because of I guess it's more just enhances hearing, not necessarily I don't know if he's clinically determined to be having hearing loss, but I think he does just have better clarity and it's just the ones that wrap around their Bluetooth and he can configure them off his phone. so I mean, I think those use cases are there, but I mean AI does that make it does it increase the clarity? Like maybe someone I'm just throwing this stuff at the wall here, but maybe someone has an accent that's different than yours, maybe? Does it actually translate the accent? I'm just throwing that to find out. Yeah. If anybody you guys know about any of this, I mean w I know it's fairly new and they just released it recently, but it'd be fascinating to hear in the comments around what are your thoughts on this? 'Cause it's an interesting topic and honestly it's impacting human life as we know it. Totally. Yeah, exactly. One of my stories is right from here is is robots. So when we got to the conference, and I'm sure Catherine's gonna show the clip of this, but there was Gerald. Gerald the robot, the humanoid G one retailing for I think around twenty eighteen twenty thousand dollars US. yeah he's Just hanging out, walking around, doing his thing. at one point the video you're seeing is him doing some karate kicks and just when I was filming that, I'm like, Is he gonna come after me? So it was very interesting to see like that. I haven't seen a bird, maybe some of you have seen him in person, but I don't think you and I. It's kinda like, huh, interesting. I mean they they picked the most non threatening name ever and then the guy started doing ki karate demonstration. You're going, What what's going on? I know. Well, and I I sent the clip of that with that name tag Gerald on it. Gerald that's a good friend of mine. So he got a kick out of that. I said, here's your new AI overload. Good thing you guys have shared a name. So maybe you guys I mean maybe maybe we can help download his brain and that Gerald. I don't know. Sorry, Gerald. both Geralds. but you know, it was interesting going through that and even the talks today talking about the robot side of things. So yeah, there's that that to me it it's almost like a parallel trick. but the real stuff I think is when they got into some of the sessions and when we talked about how do you train a robot? One guy was talking about the difficulties of doing that. And says, Yeah, you can go and train a robot, but the reality is that when you're training this thing, how do you train it? It's a lot more difficult. Like it showed a picture of a really messy kitchen. And like how do you get into the kitchen and all that stuff? And it was, you know, it's certainly a bit of an issue. Yeah. So you train it to do that. So what they gave an example of was like put put like a headset on or something, a recording headset, whether it's say your rebands or whatever, then you show them doing the actions. So you record the actions of what you want the robot to do. And then what you do is you share that data with the robot, that video, and so this is how you sort things out or whatever else, whatever process. And then that's a way to train it, which I thought was quite compelling. So think about it a robot with an AI in it. Right. Share that video with it. And then you tell it, okay, can you go ahead and start producing the doing this work for that? And it makes a lot of sense. So I thought that was quite fascinating on how that works. but yeah, I what do you think of robots? Do you see robots coming into play soon? Like I think they already are, honestly. Like there's there's you know manufacturing, the e-commerce industry, you know, they're already taking advantage. I I mean even one of our clients recently, I did a site visit and I watched, you know. And again, it's it's minimal application right now. It's it's it is spending, right? You to invest. are you is helping them now? is that taking away a job? No, there's actually a guy monitoring it now that doesn't they're not, you know, threatening the fact that it might get hurt, there might be a safety protocol issue, it's very controlled. Well, what I thought was fascinating, going back to that presentation you talked about, was the earlier statement of we're almost learning our robots are learning as if. They're four year olds to a certain extent. And that that metaphor really hit me because when I think about AI, think about technology, think about even just the human interface with this technology, everybody's skipping the step of like, hey, it doesn't know what it doesn't know. It just assumes that the system is going to know what to do. So I think you're bringing up a good point is when does it come time to train them as if they are humans? And is that training going to be the new realm of not a job taken away, but a reframing of what we're actually going to produce? So I think there's something there. And I and again, I'd you have to dig into it a little deeper, but what are your thoughts on that? It'd be interesting to see like the example of a video. Like this is the does the AI take the video and say, you know what? I get the idea. Let me try to do a better right? That's fascinating. I think that's so yeah. So that robot thing kinda hit me there with kind of an interesting one, but yeah. obviously the karate kicks and all that. It was entertaining. I I took a video too and I remember it was dancing and it was a lot more less threatening. Yeah, yours look way more less threatening. So it was like waving a few. All need to know now. Actually know funny last thing thought on that one. Okay. Yeah. So whoever there's a contest they're running here and whoever gets the most points gets to actually win this robot. And he says, Good luck taking me pole on an airplane because he fights and kicks. And they walk saw the video. So good luck getting that stuck around an airplane. That's anyways, what's what's your next one? All right. This one's kind of a combined one and this is related to kind of going back to the last episode we were doing and talking about the medical side of it. What was fascinating was is that they had hardware and this is an AI teamed up to understand exactly the neuroscience behind the actual You know, it's like the spoke and gesture. So it's kind going back to the training of the of the robots, but it's also training people in some regards. Right. The biggest thing that happened was that there's a software that's actually talking about the neurons, the protein design models created by this specific enzyme. And they paired it up where they're finding that a lot of the work that's happening, the researchers are able to check in major quantities of data to then get to the result a lot faster. Right. But here's the kicker. This is the part I really want to focus on with you is there's a balance of September 27th. There was an I healthcare IT today roundup that says 70% of healthcare workers feel AI makes care feel less human for patients. Now, going back to what I just said about the partner partnership, we're seeing the results faster. So in my eyes, getting to a result or a diagnosis faster is going to give us a better chance of being preventative instead of reactive. But they don't they you know, they're not wrong. It's gonna be a lot more augmented. So I guess let's start there. Like what what where what side would you be on it like? I like I've always thought preventative is okay. depends on the I mean obviously us being Canadian, you Canadian, you American, like it's interesting when you compare healthcare systems too on how they work together. or work differently. I imagine in the States there's probably this this feeling of of like going to the market as cost mining thought. It it is. Yeah, it is. So yeah. And in Canada it's like well, it's public health care. The problem typically is can't find a dog. So availability. So availability is an issue. So it's not so it's not even so much of a cost thing. So a lot of people end up going to emergency and all that whenever they cough. And it it ends up plugging up the system. So it's it's a kind of a different thing. I think in both cases though, the s the the underlying issue is the same thing. Yeah. So how can we so I guess I'd be on the side of like how can we actually enable make this is kind of a pathway to help still start enabling people so they have preventative measures so they don't need to go see. Yeah. Like I mean you have to do it with caution, like I'd probably want something endorsed by a health provider. FDA FDA in the US, right? And just some regulation, something. Yeah. Like I need sure he's board that but there's like simple things like, Yeah, what do you think about this? And I don't know if you have this down here but in Canada, at least in British Columbia, we have a a hotline is called centers. Got it. Yeah, you can call them up if you're concerned about something. Yep. And they'll tell you what they want to let's start with that. Yeah, you know, I think the gathering information and the the the logic and the thinking that is housed within that tribal knowledge is gonna start being more related to actual references, right? I think that's where my head goes. It's not necessarily we're gonna replace the humans. I I wanna be very clear on that. I think what's gonna happen is we can and we're doing this in our own company, I'm doing this in my own life and my own ventures where We're trying to get back time to focus on those relationships. So I think what you're bringing up is fair. It's like it's kind of a balance, right? It's always going to be a balance. Now, are there going to be people that prefer one or the other? Absolutely. Right. I know the generation that's my generation and higher are gonna probably want to have that texting feature within a bot just to get the answer. They're not gonna want to pick up the phone. But others are different. So I think we have to adapt. We can't just say we're in this extreme, we're at this extreme. It's What is the what is the environment needed? Yeah. And how do we do that in a way that's gonna impact people positively but still get to the result we're looking for? Totally. Yeah. Yeah. What's your what's your other story that you got today? This one's again from the conference. AI stuff, but yeah, obviously based here. I mentioned earlier at the intro there, like, you know, governance, so As I was I spent some time going through a lot of the vendors yesterday and as I stopped each of them, they all have this these different governance tools that basically you know I dumbed it down. I said it'll send text over there and text comes back, right? they're trying to be the intermediary. Okay, so it's like almost think of it like a filter, right? Yeah. So they're filtering information. Then the idea there is like understanding what's being sent to the AI and what's being sent back. Trying to put some understanding around reason like you know, certain professed kinda kind of how they go. Yeah. But takes it even further. Some of them are really focused on like the software development lifecycle. So they'll have data on you know cost, which models cost more. Right. who spends the most out of your team. Yeah. So there's lots of data and analytical tools that are flushing out this again, this for relatively simple concept, but they're building all these layers and stuff. You're gonna see a lot more of this when they come in. But the governance piece was really interesting and what stood out to me is the heavily regulated industries that don't want to actually a guy put it perfectly says is the the regulated industries like your healthcare, insurance and all that. But the other one that really he said is that what ends up happening, a lot of their customers come to them because the bridge and their board said, Hey, you need to get some kind of governance in place that continue tools. Yeah. so They're also it's not just like regulation, but it's also even boards are starting to say, That's great you're using all these tools. This goes back to like shadow IT stuff we talked about in past. And yeah, and now it's like you need to actually put in place some tools and mechanisms to monitor what's going on. I don't know if I can say that. Is that another beep? Okay. Gotta believe it. Gotta believe it. Maybe you should put like a smiley face on my mouth too. Yeah, you know, stamp of disapproval. So any anyway, so yeah, the to me that was yeah, it was very interesting, those conversations, just like all the tools that are coming, all the only governance side. So that's it. So just to give you a perspective, for listeners, this is a great way of thinking about it. You're used to using Chat GPT, Claude, or Gemini, whatever you're chatting with. But once you hit that enter key, it's like the message is off. Yeah. And it comes back and you get a response. Yeah. There's so much that goes into creating that response. wouldn't you want to know why it got to that? Or you know, why the follow-ups got to a certain point? or why did it suggest a certain decision? Or if you're a team manager auditing your teams to understand right what prompts they did and why they got to those results, like I think that this stuff will be is ever evolving. It reminds me of call centers like 25 years ago when I was in that world. Mm-hmm. And we started bringing quality assurance and we're monitoring phone calls. It's screen reporting screens as well. So then we would analyze that. That was a human doing that. But now this seems like the same idea. It's like there's that quality assurance aspect. So 100%. You know what's funny about that? There was a fact that was brought up, but I believe Catherine, you're gonna have to check on this. The McKin McKinsey was the one they cited. And it was interesting going back to the heavily regulated organizations that are adopt adopting AI, is that they are actually the fastest to evolve. And I think that's a little bit contradictive and you know, it's it's a little bit nonintuitive. But I think what's fascinating about it is I think they have to because if they don't, they're gonna get behind on the regulations and the training that they have to do to keep it regulated to a certain degree. So I think that on top of the fact that we have an opportunity to evaluate not only the permissions, the security, then we can't approach can't approach AI the same way we approach the platform security anymore. That's not the same thing. But we're seeing the need to understand the reason, right? It's good, it's human nature to go why, right? We should be doing the same thing. And I think the stuff that I saw and I talked, I talked to you know a bunch of groups in the vendor just to get specifics around what are they doing. It's all the same thing, just different wrappers, just like we always deal with. But the the commonality was we need to know one, how responsible is it? Two, what's the efficiency rate for to make sure responsible money-wise? And three, what are the sources that it's pulling from, right? Because if you're pulling from, you know, I don't want to say it's probably for Wikipedia for any of the public models, but like if let's say it was. Well, that's not a regulated source, right? So I think I think there's something there. And I think you're right. This is gonna be the next level of how do you manage it effectively. Yeah. And what what can you see? Because you're still gonna need humans, guys. They're still gonna need humans to find out is this data really. Right. Exactly. Yeah, no, it was that was very fascinating. Yeah. Right, you still getting there's still things and we still have to manage things, but it's interesting you made a comment right there. It's like we like to know the reasons. Isn't it crazy when we say that Chat GPT is an answer? We feel comfortable because we got an answer. But we didn't ask why. Like so many people do that. They think, I got here's the answer. Or I mean I've heard so many times where people are like, yeah, my CEO came by and said that he talked to Chat GPT and that's how he made his decision. Like a really important decision. Kind of like Okay, all right. This is the other policy I think people need to be careful with is getting stuck into can't remember where I've been reading this. It was an article or a podcast I was listening to. and it's Cal Newport. Cal Newport's podcast. the and on he does deep work and all that if you haven't read his stuff before, but he was talking about how People are having conversations and it makes them feel like they're having a real conversation. So he's arguing that we need to stop having these chatbots actually feel like they're real humans because they're not. Because we start thinking that it's a real human. So so to your point, like it's how it's crazy how we agree with what they said, the chat bots, yeah, and just move on. whereas if you if you told me a fact right now, I'm like, well, okay, tell me a little bit more. Maybe I might go research a bit, but it chatbot tells us, we think, yeah, AI told. And you know what's funny is I never actually believed the answers they gave me, even when it's the models that I trained. Yeah. I always ask, check your work, or some variation thereof, because half the time and you know what? I'm gonna be honest. Eighty percent of the time it comes back goes, right, I missed that point. Right. Yeah. I think this is a good segue. I'm feeling feeling like we're about ready to get into the segment here. There we go. So this week what do we do? Prompt relaulet, yeah. Yeah. So this week we're gonna do what are we gonna call this segment? Quiz? Pop quiz. We'll call pop AI Jeopardy. No use it? I don't know if we can. Okay. So what we've done is we've taken a bunch of the recordings. So I use I use I have an AI recorder and I've recorded most of my sessions over the last couple of days and I've loaded them in here. So I can share my screen in a second for those that on YouTube. for those on audio podcast, though we'll just we'll be talking it through. But we're gonna have the AI actually ask us some pop quiz questions, five questions. So let me get Claude up and running here. What I've done is I've created a project. inside of Claude. And what I've done is I've put instructions. I say you're a quizmaster. There we go. in the spirit of Ken Jennings of Jeopardy, speaking of Jeopardy. We had to have a persona of some degree. We had to do we had to something something. And you can kind of see here I've given us some instructions. So read through all the context files. And basically I exported just ex a bunch of text files and threw them in here. So the only thing we have to do is actually Give the level of difficulty and how many questions. So what we're gonna do is let's start quiz. let's can't type in the sun here. Let's do five questions at low difficulty. 'Cause I feel like Okay. Okay. It's gonna be challenging enough. So let's go. Yeah. When we tried this the first time, just so everybody knows, the question that came up was so easy that we were a little concerned that it just looked like we're just making, you know, play ups. Like we gotta we gotta make it a little challenging here. Okay, so we're up and running. So question number one. In the open source models talk from Together AI, the speaker named one reason above all others for why companies are moving from closed models to open source ones. In practice, customers who got it right saw savings of up to 70%. What was the primary reason? A better data privacy, lower latency, cost savings, easier fine tuning, or access to larger context windows. So I wasn't in this session. So I'm gonna make a guess. What do you think? To me, from what I heard over and over in other sessions, the cost savings kept going up every time. Cost saving? I think I'm with you on that one. I I and that's and just so everybody knows, the op the open models are the ones that are starting to take over the market. Yes. They the big but there are other elements that play into these open models too, like performance. See? See what let's see, let's see what our AI does here. C honing. It says honing right now for those on audio. That's just correct. So the speaker cost savings was the main driver of open source adoption with customers seeing up to 70% savings once they got model selection and deployment rate. So you're worried that I was lisping. I'm always there to speak. And I mean I was confused for a second. So and I was there. Okay, well, question number two. Ready. You want to run this one through? All right. In in the Amazon talk, your agent's best thousand tools, the speaker gave a simple formula. A harness is the prompt, skills, tools, and memory connected by a loop. When you add the model to the harness, you get this. Ooh, that's a good one. I think I remember what it is. What do you think? I think it was agent. Is it agent? 'Cause the harness creates an agent. Yeah. I I I think I agree with you. I don't it it's funny. I mean you could call it co pilot if you're Microsoft, I guess. Yeah, but that that's a product, not a yeah. That's not it. We're talking structure here. So I agree with I think it's an agent. You know what's funny? We could say that we could do process elimination. Is it this or this? That's how I used to do the exams with the kids. It's the A B testing model for you guys that want to know about more about AI. All right, B. B is survey says It got stuck itself. While it's thinking it's thinking really art. It's really hard. We might never know the answer. here we go. Correct. The formula was model plus harness equals agent. The speaker set of harness pulls the prompt skills, tools, memory together in a loop, and that for most use cases you can define is in a simple JSON structure. Yeah. Amazon bedrock and core. There it Yeah, yeah. There it is. Okay, question number three. In the Google talk Beyond Toil, the team built an AI powered platform to review security change requests. Humans went back and checked the agent's decisions afterward. When the agent got it wrong, the team updated the playbooks it used. But how often did the humans agree the agent made the right call? Okay, I wasn't here for this one, so it's just a pure guess. But I know I think I know what the answer is though. Based upon what I'm seeing in trends. And based upon what we just talked about, which is agreeing with what it says because of the nature of the human language, I think it's ninety percent D. I feel like that's exactly what they said. So let's see. D Mulling thinking. Correct. Human reviewers confirmed the agent was right about ninety percent of the time. For the other ten percent, the team updated the playbook so the agent would get it right the next time, which means accuracy should keep climbing. That's really fascinating. I I love that fact. We need to make sure to put that in the notes. That's a really good Yes. That's a good one. All right. On the fundraising panel, one panelist laid out AI's killer apps in order. The first was chatting. The second was coding. He said the third, which explains why cybersecurity stocks are doing so well, is something else. What did he name as AI's third killer app? Killer app. And I see C is I mentioned it earlier. I mentioned it earlier. was it robots? These are all my notes. Was it robots? You mentioned robots earlier. No, I also mentioned something else. I said there's three stages. well I I I hope it's the right answer. I'm gonna look like a fool. But the first one was chatbots. Then it became what I say, chatbots and then I forgot now. Chatbots coding. Coding, there you go. And then hacking. I didn't know that. Let's see if it's C. If it's C, I wanna know the reason. I'm actually really hopefully. I hope it's C. Otherwise looking like that fool. As you as you explain that all correct. There go. He said AI is very good at finding vulnerabilities and stealing passwords, which makes hacking its third killer. He added that attackers have the edge because offense needs no permission, while defense requires a whole organization to agree on how to respond. Okay. All right. Are we going for the perfect score here? Let's see if we can do this. In Peter Norvig's Beyond the Frontier talk on coding, he pointed to a famous software figure who was cautious about AI six months earlier. That person now allows AI generated code in the Linux kernel and tells anyone Who doesn't like it to go away? Who is it? This is the founder of Linux. see, you would know that. I think it's it's Linus. It is Linus. Yeah. It is Linus. I think we hit five out of five, dude. We did go easy. So imagine if we did the the whole mailword. All right, so correct it says Norvig said Linus Torvalds went from skeptic to advocate in about six months and now allows AI to the Linux kernel. Norvig used him to show how fast coding became become became AI's strongest domain. You can run code to check it. There's plenty of training data and programmers are already using it to fix their own mistakes. so we got a five or five perfect game. It says you went five or five and low difficulty. So you're clearly paying attention to conference. Wanna rematch? I think we're good. We don't need a rematch today, but not today. Yeah. But one thing on Lin L Linus there, I've now they brought it up, I I I was so he's been commenting quite a bit about coding and all that too. His argument is that AI is not much different than an IDE. So if you're writing code not so much an ID, but a compiler, sorry. So compiling code. He says it's it's no different than that. So he gets frustrated when people send a pull request like with a code change to the to the repo repositories for open source. Yeah. So you don't need to say this was made with AI. His argument is that you are responsible for your AI. So which I think is the right way of doing it. It's another tool, it just happens really fast. Yeah. So there you go. Cool. All right, let's stop sharing this. And we're back. And that kind of takes us to the closing piece here. So as we wrap up, any final thoughts on the conference? We got a couple hours left here, but yeah, I think I think one of the big words that kept popping up that I you know I even learned, honestly, and this was something that I'm taking a little bit seriously in terms of understanding better, is the inference and the foundational of data. And I it's something that I've been looking into more, and honestly, for everybody that's at home and listening or watching. You know, inference isn't something that everybody understands and I don't expect you to, but if you think about it in the nature of a human language, right, inference is basically your inferring information. So it's the agents inferring what it learns along the way. I think this is the next stage of our agents that we're going to be developing moving forward is if we can get this to learn in a way that is practiced and structured, not just creating latency and not creating a bunch of context that is unusable that's causing the cost to go up as far. I think that's going to be the biggest piece. Now, how do we do it? I think that's up for decision because the str the strategy's there, the concepts there. Yep. I don't know if anybody's figured it out. And that's what most of these vendors are talking about is you know, you know, inference. Totally. Yeah. Yeah. I think for me it was just it's just amazing to see all the new things coming up. People trying to it's like people trying to trying to solve a lot of problems that are out there. Yeah. And there's so many different ways to tackle it. like I mentioned the governance because so many different companies doing it in a different way. And it's really cool to see the innovation that's happening. I have no doubt a lot of these companies won't exist five years from now, let alone maybe a c a year or two from now. But they're all trying in every conference I'm going to, they're all new businesses, there's new ideas, new startups. so to me it's like it's really cool seeing all the innovation coming forward and all these new ideas. And yeah, that's pretty good. Yeah. So it's been a good conference. So Yeah. Absolutely. Cool. All right, sir. Well On behalf of you and me. Yeah. You listener, thank you for tuning in. hope you learned a thing or two here on with our our third episode this week. And I guess we'll talk to you again next time. Thanks again. I'll never back on. Are we? I'm starting to see. You did see that or or wondered if a camera got cut off on our video. So good luck to Catherine is piecing this together for us. Yeah.