240 - Distressed by the AI stuff around

I’m not stressed by AI itself. I’m stressed by the insatiable greed of those who profit from it, even if it means sacrificing large parts of the population. I'm also stressed about how ruthlessly it can be abused to cause deliberate harm.

In this episode I'm not taking you into world of fire science, but rather into my own thoughts on how the AI revolution influences our lives. And I was influenced it just last week - through a phishing attack on the IAFSS, and through reading a very disturbing piece of fiction I found on the Internet...

In the episode I comment on the targeted phishing attack against our association that used well-researched details and a cloned voice pulled from public audio. From there, we step into a stark forecast of near-term AI disruption in white-collar work. Agent teams can already write, review, and ship production code in loops, compressing time and cost while jolting stock prices across entire sectors the moment capabilities drop. 

Then we get specific about our field. Some tasks in fire safety are ripe for automation—code interpretation, routine calculations, device placement, and documentation—where speed and consistency help. But holistic fire strategy is contextual and slow to validate, with scarce, standardized case data and long feedback loops. Buildings are messy, multidisciplinary systems; that friction is a temporary moat against full automation. The larger risk may be macroeconomic: if AI compresses demand and margins across white-collar industries, construction cools, and safety work gets squeezed. Paradoxically, low digitalization in construction buys time, making it harder to train and deploy one-size-fits-all models.

I'm still to large extent positive Fire Safety Engineering won't be directly disrupted at the same scale as Software Engineers got, but as a part of a larger ecosystem we won't be untouched either... I hope the version of the future that plays out is more optimistic than the one I got worried about.

Read the Citrini piece here, if you have not yet: https://www.citriniresearch.com/p/2028gic

Update: A week later there has been a lot of works that have refuted the dark scenario in Citrini piece, like here: https://www.businessinsider.com/ai-job-losses-wall-street-strategist-citrini-research-citadel-securities-2026-2?IR=T

Some are showing Software Engineering job openings rising etc. Perhaps world is not in the dark scenario - I highly hope so! 

----
The Fire Science Show is produced by the Fire Science Media in collaboration with OFR Consultants. Thank you to the podcast sponsor for their continuous support towards our mission.

Transcript

Setting Expectations And Context

Wojciech Wegrzynski 0:00

Hello everybody, welcome to the Fire Science Show. Today we will be talking about AI, but this is not your normal fire science show episode. There's no fire science in it. It's not a scientific record of research or anything. It's just some of my thoughts that about stuff that happened with the within the AI in the last week and they have genuinely frightened me. So I thought, well, I I feel an urge to record this and share those thoughts with you. I thought about such an episode already when MoldBots came out, but then uh I was just a gimmick. I thought uh no point in doing this, but last week, wow, that that was crazy. Uh if this is your first Fire Science Show episode, that's probably a poor choice of starting your journey with a podcast. There are 239 instances of real science in the podcast, so you'll find them below. And if you're a long-term listener, I think uh you may actually uh enjoy this. I felt the need to talk about uh AI because I feel somehow connected to the topic. Perhaps I'm not a creator, I'm not uh I'm not the leader, I'm not someone driving the AI efication of the world, but uh I'm at least an active observer or maybe an early adopter and uh in in that, or at least people put me in in such a bracket, if I may. In the Fire Science show, if you look at the podcast, in the first 10 episodes we had two AI episodes. Uh episode 20 something was one with Mz Nasser about how to enter the world of AI. That was way, way before chatbirds and when it was so much more difficult. And actually, uh Nasser, what he said back then, it was that he claimed that he's uh just a few years ahead of everyone else and he's just trying. And today, look, five years later, he's uh the undisputed industry leader in AI, not just in fire safety engineering, but in civil engineering. Actually, I have good news because he's coming back to the podcast in very, very, very soon. So we'll have a scientific view over AI, not just my rambling very, very soon. But let's let's continue with rambling. We have uh shown you chatbots very early. Well, just after they came, we're discussing how chatbots could be used for the benefit of fire safety engineering. We've discussed implementation of other AI tools, we've discussed how to set up your own AI LLM models. I've presented my own view on where it's heading and where it's gonna be useful, where it's not gonna be useful. So a lot of stuff has been shown about AI in the Fire Science Show and and by myself around in different places where I show up. And uh I'm usually extremely enthusiastic about it. Or if not enthusiastic, I'm realistic that it's gonna be used for a good and it it gives us opportunities. That that's what I see, opportunities. And and this week, two things have happened that that frightened me really. Like I my view has shifted. Perhaps I should start farming. Maybe Bronze Age was not that bad, actually. Like really 180 perception shift. It's I've recovered, I'm a bit better now, but this was quite strong for me. So one is uh a fishing attack on IFSS, quite a big attack, to be honest, and uh it involved me to a level uh way, way, way beyond my comfort zone. And uh second thing was uh a report, an online report essay novel, I don't know, with fiction work by Citrini. I mean it it's a work of fiction, but so is uh 1984 by Orwell and look at the world today. So yeah, sometimes those things materialize, and if that Citrini reports materialize, that would be a hell of a shock for economy. It's a report on on the uh disruption by AI into the economy, and basically quite catastrophic in its view, and this prompted me to think about uh what AI means for fire safety engineering again, and and what it means for the broader economy, and at broader economy point of view, what does it mean for fire safety engineering? So, those two things I would love to share with you in this podcast episode. If you stayed with me this long, there's a good chance that what's gonna show up after the intro will be interesting uh to you. So if you would like to hear my views on those topics, uh stay with me, let's spin the intro and jump into the episode. Welcome to the Firescience Show. My name is Voycze Vingzinsky, and I will be your host. The Firescience Show podcast is brought to you in partnership with OFR Consultants. OFR is the UK's leading independent multi-award-winning fire engineering consultancy with a reputation for delivering innovative safety-driven solutions. We've been on this journey together for three years so far, and here it begins the fourth year of collaboration between the Fire Science Show and the OFR. So far, we've brought through more than 150 episodes which translate into nearly 150 hours of educational content available, free, accessible all over the planet without any paywalls, advertisements, or hidden agendas. This makes me very proud and I am super thankful to OFR for this long-lasting partnership. I'm extremely happy that we've just started the year 4, and I hope there will be many years after that to come. So, big thanks OFR for your support to the Fire Science Show and the support to the fire safety community at large that we can deliver together. And for you, the listener, if you would like to learn more or perhaps even become a part of OFR, they always have opportunities awaiting. Check their website at OFRconsultants.com. And now let's head back to the episode. So in the intro, I've mentioned MoldBot Revolution. I'm not sure if you followed that one. Uh those revolutions come and go very quickly these days, but basically, one person has created an AI agent tool that you run over your own computer, MacBook Mini, actually. And uh basically what it does, it's uh like a self-uh guiding agent that just lives in that computer and does tasks that you do, like kind of a person. And uh what was interesting is that uh one of people were giving them uh access to sensitive information, you know, like credit cards, etc., like ridiculous stuff has happened with that. And also another thing was that uh someone has created an online forum for those mold bots to chat uh with each other, and that was also quite a thing. But then I thought, yeah, okay, that's an AI gimmick, that's uh you know, just a funny thing, just an interesting thing, uh, but not really world-shifting thing. Uh interestingly, since then, that happened like literally two, three weeks ago. It's really recent. And the person that was an open uh source project, but that person has uh apparently has been acquired by OpenAI for like one billion dollars or something, and they're just developing it now, and uh it caused an MacBook mini shortage all over the world. So yeah, that's a ripple effect in the economy, which is kind of relevant to what has scared me this week. So I I said uh I was frightened, uh I was generally frightened, uh, and two things have happened that cost that. I'll I'll give you a story. The first one, uh it was Friday morning. I woke I woke up and uh I had an early morning meeting, and uh, as I move into the meeting, I see my mail inbox flooded with messages. Well, flooded, that's probably a big word, but I see at least five different uh messages from my colleagues uh from the IFSS um members advisory council board and the trustees of the IFSS, so let's say the leadership of organization, and uh those emails range from hey Wojciech, there has been a scam uh attempt with your name in it, to uh hey, I'm really sorry for your loss. If there's anything I can help you with, uh like let me know. And I'm like, like, what the hell is happening? Apparently, IFSS has been a target of a large-scale phishing attack. Um, not the first time, it has happened before, not particularly difficult because all we do is public, so the names of the advisors and trustees are well known. The relationships between them are well known as who is who. There's a lot of public information that allows to track us and that allows to you know link us together. And and basically the phishing attempt the what was disturbing is that the phishing attempt was surprisingly well crafted. There was someone impersonating Professor Nayan Lu, who is the president of the IFSS. Uh, and uh in that email, the impersonator was letting know the people who received the email that I am in distress, that I am flying to a country, and there was the name of the country from which this particular person was coming from. So people from Japan got an email that I'm heading to Tokyo, and people from Australia got an email that I'm heading to Brisbane. And in this email, I'm in the distress, I'm stuck in the Philippines or or something, and uh stopped by the border, and I'm heading for a funeral, and I really need help, and he cannot help me because he's flying to Canada. And if that person who was a target of attack could call, uh, and there was like phone numbers to Philippines uh and and and help me. I was decently crafted uh email, like not your usual Nigerian prince scam thing that included a lot of genuine information, like someone injected a lot of genuine information into that. And this was targeted to a person. It was not like a generic message sent to everyone, it had information specific to that particular person, like the country, um, the direction, etc. Like it it it was frighteningly well-crafted message, I would say. And uh when I first saw chatbots, my initial I I I I kid you not, my first thought about Chatbot was my god, scammers are gonna get good. And and this felt like like that. Like this phishing attack felt like a a much bigger improvement over the the scamming attempts that I've seen in the past. But this is not enough to frighten me. I mean, it's it's a just a phishing attack. I've replied uh to IFSS that we are under attack, I've issued a warning, uh, we've sent an email to the membership, uh I thought case closed. But then I was chatting with some colleagues because more people were reaching out to me, and actually two people said that. But one person told me, like, Voczy, but are you sure you are safe? Because because they are now calling me. And I'm like, no, I'm sure I'm safe, don't answer. And then another person said, You know what? I just got a call from Philippines, and it was with your voice, and you were very convincing in that call, and that that's when I broke. Like, seriously, that is when I broke. Like, I I joked with some people that wow, this phishing attempt is really good, but imagine if they could clone the voice or use AI, you know, to enhance the scam and uh etc. And and I was thinking, yeah, that's a hypothetical scenario in three years, but no, that was the scenario that was in the play right now. Like they did that, they cloned my voice, they started playing those messages to people. Uh and wow, this this was like really bad. I mean, I I was nowhere no way connected uh to that attack, but I somehow feel bad if if if someone lost money or or was uh a victim of a larger attack uh because of their you know relationship with me, because they trust me, because they like me, because they felt uh they would like to help me. I mean that's horrible. Like that is truly, truly horrible, and this made me sick to my stomach. And uh I was not the only one. There was uh later in the day Europeans started receiving messages with my friend Sinan Huang in the same kind of sequence of of stuff. Not sure if they they cloned Cinean's voice as well, but uh it it appears it has been a part of a larger you know attempt just targeting our small association. But man, this like this is ridiculous. And it it it is it is so so easy to to make an artificial voice. It is very easy to create artificial voice, one that sounds quite real. This is in fact an AI-generated sample made straight from a text. Like this one was not even good, but it it took me like 10 seconds to create that. That's how easy it is. And you know, my job uh speaking at a podcast makes me especially vulnerable uh to those uh identity theft uh attempts. So uh one thing that I really really really need you to know is that if I ever call you and ask for money, don't give to them. Man, that's gonna backfire one day. But if I ever call you and I ask you for money, uh please, please, for love of God, do not give me more money. Please be vigilant. This the way how this is accelerating, the way how this is spinning up, and this connects you know to the cloud bots, this connects to the agentic AI because today it's not a single person who has to sit down and send those messages. No, you can set a friggin' MacBook Mini and tell it a scam 1,000 people and create an AI-generated voice to convince them to do that and just reach out to them. That's that's how easy it is, and this will unfortunately be better and better with time. It's already frighteningly good, and it will be better. I'm really speechless about that. It's uh please please stay safe online and and uh you know we we all have to be at a higher level of vigilance compared to just a few years ago. So yeah, that was my Friday that set a mood for the entire weekend, as you can imagine, and uh I was dealing with this kind of backlash of this fishing attack for the rest of the day. And then I think on Saturday I had a pretty disturbing read. So I found an online piece, an online paper uh called the 2028 Global Intelligence Crisis. It was written by someone calling themselves Citrini Research, apparently some some sort of financial uh equity investing uh firm that provides some financial research and commentary. And uh it's obviously a work of fiction because uh basically what it is uh they provide you a report written in 2028, which is a year in which the humanity is in a crisis and they trace back how the crisis has happened. That's that's the the the way how the stories build. It's linked in the show notes if you would like to read it. I think it's quite a decent read, and uh while I appreciate the fact it's a work of fiction, as I said in the intro, uh 1984 of Orwell is also a work of fiction, and yet uh in many places of the world it exactly unravels. A lot of uh Netflix Black Mirror episodes are works of fiction, and disturbingly a lot of those are coming to life in the real world in real technologies. So um while I appreciate this is a word work of fiction, it is disturbing because I'm not saying this is something that will happen, but I cannot exclude the possibility of their predictions coming to reality, maybe not at a full scale, because I I still think there are bottlenecks, which I will I will discuss later, but uh the it's it's the direction is is probably quite realistic. Um so why the why this particular report has uh has frightened me and and why I I started reviewing my position on AI after that. The thing is, if if you ask me that one and a half year ago, one and a half year ago I was giving you some predictions on AI for the future. Usually those things don't really work out well. And uh the ability to predict future is not that easy. If you asked me one and a half year ago about my views on using AI for programming, I would say that it's pretty damn good. It's pretty damn useful. I I'm using it all the time. Since the chatbots came up, I am using AI in my Python programming for my scientific research, scientific work, and uh always felt this tool is really really strong and really really useful. But uh I'm quite primitive in the way how I use it because uh if you look at the environment right now, the whole field of software engineering is currently replaced by AI. It's not that person goes to a chatbot and asks them, Dear chatbot, please write me a software for financial analysis of SP 500 or whatever. It doesn't work like that. But people s today set up agents. Agent is like an AI tool that has kind of a life of its own. You give it, you know, a set of rules, you give it uh a goal, a task, a bigger task, like you tell it, write me a software to analyze SP 500 live, and this agent is self-guided, so it knows what's the task, it assigns the goals or intermediate goals to itself, steps of of the process. It it it figures out the process and then does it step by step, and for those steps it crafts its own prompts and uses LLMs to you know build up chunk by chunk, and you don't really supervise it, it's just doing it on its own. It can take you and it can take an hour, it can take five hours, but eventually it comes back to you with with a working thing. And this is not yet enough to replace a whole field of software engineering, of course. It's it's a more complicated gimmick but gimmick. But now people have found that you could technically set one agent to create the code, then you create a second agent to review the code, a third agent to criticize the code, a fourth agent to optimize the code, a fifth agent to integrate the code and test it, and then you know you you're running an entire team of software engineers on one task, and they're talking to each other and they're working on it without your supervision. All you did is send uh a prompt to them on what they're supposed to deliver. It costs quite a lot because they are burning through tokens, and tokens is the currency of uh using AI through interfaces, so it's it's quite costly to run that, but still that's like a small fraction of costs you would have if you wanted to hire real people to do the same goal. And largely the outcomes are are really good, like they are on the human programmer level outcomes. Now, even the AI companies they said they're moving towards this type of engineering rather than having old school humans writing code. And this kind of disruption, this is changing the entire field of software engineering. Similar thing, maybe not on that scale, has happened to uh artists and and graphics designers, etc. So easy to create good graphics today with AI. And and today the coding it this is is uh strongly, strongly influenced by by this uh flow of agentic AI, which does the jobs of those uh of those people. And this goes into this citruni report. This is basically the start of that citrunny report because you see two things. One one is that you can automate jobs and replace people doing those jobs, that that's one thing. But the second thing is the uh economic impact that replacement has. So basically, when uh Anthropic, which is the company behind the cloud or uh an AI agent, if they drop a new functionality of their software, suddenly there's like drops on the stock market of companies which will be disrupted. Literally, like yesterday, I IBM has dropped because uh Anthropic has mentioned they can program in cobble. So it has really two avenues in which it's dis disrupting the economy. One is directly, one is through replacing opportunities. And it's quite easy to replace human workforce with AI when you kind of specialize it for the task. And once this happens, it disrupts the space within it, which it happened. What I mean by that is that if there's you know an industry that uh charges you one thousand dollars for a service and it hires two hundred thousand people, each of them earning a hundred thousand bucks a year, and suddenly you create an agating AI tool that can do the same job almost at the same level, but for a small fraction of costs. One is that suddenly those two hundred thousand people are perhaps going to be unemployed or will have to significantly change their lives, and two, I don't think you're you're able to charge a thousand bucks anymore because now you're competing with others who also do it for a fraction of cost. So the cost of the service falls down drastically. And uh the whole Citrini report is is that this is actually what's happening right now to some of the services, and when it happens to one service, that's pretty bad for that service. But when it happens to the entire economy, like the white collar economy, all the jobs imagine all the jobs that could be automated become automated. We are talking about the majority of workforce out there, which is victim to disruption. We're talking about insane amount of companies who absolutely shift their revenue streams and suddenly this revenue is not going to them. I mean the the Services become cheaper, the money is going elsewhere. This is a large-scale economy disruption that can pull banking and housing and other stuff collapsing to one uh one after another. A whole like systematic change in the world uh way how the world lives. And you know, uh Elon was uh speaking about such things some time ago. Uh Elon Musk has said that it's pointless to say for your retirement because we will be living in the world of uh abundance. That that's what he said. A world of abundance. Uh and and his logic was that everything is gonna be so cheap and so well developed through the AI tools and AI support, then there will be a very limited need for human jobs and human work. And actually, the wealth created by this AI revolution will be distributed among the human population in some sort of universal income, and basically you're gonna have everything you could ever need and want, like utopian world of abundance. And uh yeah, that's that's kind of cute, but uh I like observing the world around me, I don't think that's how the world works. Uh Charlie Mungo said that the world is not even driven by greed, it's it's driven by envy. And and what I see is you know, those optimizations they don't lead to create a little utopia for software engineers, they don't lead to create little utopias for anyone. It's about you know directing revenue streams to smaller and smaller groups of people and you know profiting. And those people who lose their jobs, that's just you know, as a cost. That that's they're they're the victims of the process. Who cares about them? Now, if this happens to the entire economy, we're we're kind of screwed. This this kind of frightened me in that Citrunny report. Is it fake? Is it like unrealistic? Is it hyper-optimistic about this assumption? Probably it is, it's it's it's quite strong, and I think there are bottlenecks. One really strong bottleneck is energy. That's you need a lot of power to support those uh AI tools to train them in the new fields, and uh if everyone is using them in an increasingly large scale, we're basically gonna run out of electricity to power that. So there's a hard cap on how much we can automate with AI based on an electricity power. And I would assume some legislation, perhaps some taxation will come up to soften the blow and perhaps move the world a little bit into that uh general abundance. But indeed, this scenario is something that that could realistically happen, that could realistically unravel. So now why I'm talking about this, why why this is relevant? Because immediately after I've read this paper and I had a chain of thoughts about you know the future and the AI and how it is disrupting the world we know, I started thinking, okay, if those software engineers have been replaced by AI largely, can fire safety engineers be replaced by AI at large? That that was my thought. That was my immediate concern. How about my job? Am I at risk of being replaced like that? And this chain of thought has led me to, let's say, an optimistic view that I don't think it's very realistic in the close time span, whatever foreseeable time span. And there are reasons are a few. One, uh we're kind of unknown to the larger society. Show me a work list, uh, list of jobs that counts as fire safety engineer as a job. Uh we're hidden in the shadow, the society doesn't know about us. So we're unlikely to be uh an obvious target for automation, you know, because we don't really exist in the heads of those people. Uh so uh that that's that's what gives us uh a little bit of safety. Uh this not being exposed. But uh joke jokes aside, in terms of software engineering, one you had an abundance of resources online, like you had Stack Overflow where people would be creatively solving each other's problems in programming. That's a gold mine to train AI models on. We had um abundance of code, a lot of softwares that work. You know, we we had that all available in the cyberspace where you could steal it. Let's be clear, those AI models are built on theft, so you could easily steal that and train your models on that. And once you did that, you received a model, and then you could run rounds of improvement on that model because in software engineering you can create software, you can test it. What works you can keep, what doesn't work, you can drop, and you know, in in a kind of evolutionary way, you're able to optimize the product to receive a really well-working one in the end, in a very short span of time, because those iterations can come very quickly. Uh, if you think about fire safety engineering, there's not that much of that particular problem solving. There is a lot of knowledge, there is a lot of information about how stuff works, there is a very sound and solid theoretical foundation of the discipline which you can uh go for, but there is not that much case study problem-solving things available online easily. So there's not such an abundance as there was for software engineers. Therefore, it makes training your initial models so much harder. And the second thing is the iterative, you know, recursive loops, they're not easy. We're talking about buildings, we're talking about projects. You cannot implement it and expect the next day you're gonna have outcomes and improve and improve and improve. Now it takes years to get buildings approved, accepted, built, tested. So those feedback loops also are kind of much slower. So I think there are mundane tasks that could be automated. I think I know the ability to read code, like the whole code consultancy, this is perhaps endangered. There are design tasks that could be automated, like distribution of pipings or cablings, or or even jobs like uh locations of sensors and sprinklers, etc., that this this could potentially be automated. But at large, you know, the the FAR strategies, the FAR scenarios. Uh I'm not sure if this is easily is it's possible to easily automate that because of how much they are connected to the project. So um despite you know this this uh kind of interesting outcome for software engineers, I still think that fire safety engineers are still to some extent safe. They're definitely safer than many, many other jobs, which would be high on the job ranking, which would be very visible, which would have those resources to train the models and solid economic incentives to kind of work and replace them. But now uh that's just one side of the story. The other side of the story of that of the Citrini report is that but what happens to the fire safety engineering if there is truly a massive global economic crisis? How how does fire safety engineering work? And you know, I I know some people who worked in 2008 uh in the last major economic crisis. It was tough. In Poland, I personally we had some smaller crises 2012-13, was not many jobs back then. It was tough. I think uh if if this scenario to some extent the travels and the economy is generally in a bad place, fire safety engineering is not gonna be in a great place either because the construction world will be a victim of the economic circumstances around us. That that's perhaps even though if we may not get automated, uh this may be a thing that uh causes a little bit of suffering to us and makes our lives harder. Uh I'm I'm a little worried that it's gonna unravel like that. One more thing about the uh civil engineering or or built environment, engineering at large and uh AI is that uh I remember I seen some report uh about the state of digitalization in EU of 20-something key areas of uh economy. And actually, I think construction was like either the last one or among the the last ones with like very, very, very low score of how digitalized it is, and it was worse than agriculture. And and then I thought, oh man, we're so lagging behind, we're so bad in in digitalization. And today I'm like, oh hell yeah, that's good because it's uh if we're not that digital, it's not that easy to train on us. So uh perhaps uh our reluctance to BM2 digitalization is actually something that uh saves us in in the end. I'm not calling to drop the attempts to digitalize the civil engineering. I still see massive opportunities in doing that in a good way, but you know it's kind of a circumstance that our inability to transition into the digital age as an entire industry kind of worked out for us in the end as the AI revolution unravels. So yeah, the this report really stressed me out. I spent like half of the weekend thinking about this and uh you know creating scenarios in my head. I have children, I want the children to live in a world where it's great for them to live, not in a world where people compete for the most basic jobs because all other well-paying jobs have been automated already. So yeah, it was kind of stressful. Probably overacting. Uh, probably it's not as bad as uh the report says, but still I think it was a valuable exercise to read through that and think about uh how the future will look like. Uh, I wonder if you've read it. I wonder what are your thoughts about it. Uh I just really wanted to share this with you. Those two things that happened over the weekend uh that they caused me to view the world of AI in a slightly different way. Uh still seeing the opportunities, but a little bit more frightened about the consequences of what's uh going around. Um that would be it for this podcast episode. I think I've achieved my goal of rambling and speaking out on the stuff that distressed me. Please stay safe online. Please do not send me any money. Really, like seriously, don't. If if I reach to you, especially from a suspicion, number suspicious number, especially from Philippines, please do not uh answer and do not uh do not send me anything in any condition. And uh stay safe online and uh as promised, uh very soon I'll have a proper AI revolution episode with the opportunities, with the possibilities uh with the person who is the undisputed leader of the Adam Zenoser. So uh if you lacked fire science or science in general in this podcast episode, I will make sure that you will get the double dose in the one where we do it properly. Thanks for being here with me. It's it's really great to have uh people that you can speak up with or to and share stuff like that. And uh I hope uh one day we meet at a beer and we can uh take this conversation further. Thanks for being here with me today. Cheers. Bye.