 It should be live. Yeah, I find the more I prepare for these live streams, the worse they are. So I might just go for it and I can't here. Yep, the microphones seem to be okay, so it should be fine. We're also on Twitch now, which I have to monitor from not here, because that's what everyone sees. I went to monitor it from the creator dashboard. It's a live button and this is where the chat is. Just test the chat and the send button is called chat. Should be called the send. Oops, the source code of this page should have the live stream in here as well. So we don't actually need it. We have YouTube analytics. It's actually showing you way too much stuff I can handle. We're having some errors before the firewall was blocking something and there's still more errors. I might just ignore them for now. Ignore, ignore. They're very yellow and red and scary. GPT-4 which is timed out. Hello. Yep. Yeah, this gives us this another hour which I don't have of a timeout for GPT-4. So I'll have to use default model 3.5. How can I help you with biomedical engineering today? Yeah, okay. I told it to behave like a biomedical engineer. Okay. Are you a real biomedical engineer? Also, any chance you can indicate? I keep asking for this like all the time and it never actually works. I'm asking for an indication when the GPT-4 timeout ends. So when can I start using it again? There's no indicator on the page. I don't think you can do anything about this. Yep. I'm not a real biomedical engineer, but I can certainly help you with questions related to biomedical engineering. As for the GPT-3.5 model and timeouts, there isn't a specific indicator on the page to let you know when you can start using it again. The timeout period can vary depending on the usage and demand on the model. You can usually try again after a few minutes if you encounter a timeout, but it's not something I can control. Just give it a little time and you should be able to use it again. All right. So we might be primarily using GitHub Copilot for this session. So let's get it started quickly. We don't have much time. We go from there. So what are we doing last time? So we have this Flask application. I'm using Flask NumPy SciPy SKFuzzy. It's the fuzzy logic toolbox for Python. I'm still not sure how well this works. There was something similar for JavaScript as well, but it wasn't updated in the last like six, seven years or something. So let's run this one quickly. Right. So we have this game. If you haven't tried it out already, please go try it out. It's scrolling ECG on the screen and you have to click on the abnormal waveforms over there. I'll explain that in a sec. It actually should appear in here, but it doesn't for some reason. So it's the kind of stuff we have to resolve. It's technical boring stuff. It still has to be done. Anyway, so hopefully you know how to play this game already. Let me know if I need to go over this again. And currently we're trying to create a bot that will play this game for us. And then I mean, you could compete with the bot or something like that. We're using it with fuzzy logic, but why is the bot still going? Yep. Yep. It's still listening. I think it's Sure thing. Let's get started with GitHub Copilot for your biomedical engineering project. If you've been using fast API alongside SK 5 z logic toolbox for Python, that's a good choice. As for JavaScript tools, it's essential to stay up to date for compatibility. If you have any specific questions or need assistance with any part of your project, feel free to ask and I'll do my best to help you out. Time is of the essence. So let's dive in. Okay. Okay. Short. Short responses are great. Don't mind you trying to be funny. I mean, not funny, just smart, casual, funny. Not like ha ha ha funny. You know what I mean? Now, yeah, so we are having a few issues with the tool will be trying, did you want to generate the code or should I get the GitHub Copilot to do it? We're having some things not this playing on to the front panel HTML. Yep. Got it. I'll keep it smart casual. If you want to generate code, GitHub Copilot can certainly help, but feel free to let me know where you need assistance. And we can tackle the issues with your front end HTML or any other problems you're facing in your project. Just point me in the right direction and we'll work on it together. Okay, so let's let's actually do both a workspace. And what was the problem? So the problem was the problem was the output of the fuzzy logic algorithm. The output of the fuzzy logic algorithm is not displayed on the front panel HTML page, even though we have all the components for it. Clear? Yeah, talking to two boards. It's a bit hard. Still trying to figure out how Copilot is actually deciding what code to look at supposedly when you put this workspace. It will look at all the files or not. I'm still testing that bit. Right, so we have the fuzzy logic analysis. We're returning results we must do. Yeah, returning result a front end. There's no link. She can just click on like the blue one there. It's a bit annoying. But this will be our game logic in Ajax request. Do we have anything Ajax? No, why not? I'll show you my Ajax call to serve to get the result of fuzzy logic algorithm and then display it on the HTML page. Okay, why aren't we doing this already? Let's show we have a fuzzy logic also. No, we do not. Okay, this is good. I have a patch request. And to have any get get done by D. No, that's not it. So I suspect it's giving me as something generic. So based on the previous response, how to update this code, I do not currently have an Ajax request, but I do have some get from HTML using fetch API. Right. That's right. And then display those are not clickable. It's pretty annoying. Should be click a function. Okay, we're getting the L and Y D. And that should be okay. Check the IDs. Those IDs do not match. The IDs do not match. That's a click modified using copilot. But then do we always need that workspace? Can we show can we make sure the IDs match in JavaScript to what we have in this HTML should have been obvious, but it's not actually looking at the code. New HD and JavaScript, you have row frequency, normal frequency, row peaks. We do want to display essentially all the output of the fuzzy logic algorithm. So we know how the decision is made. Can you change? Okay. Yeah. So if you put this workspace, it's determining what the structure deciding which information to collect supposedly from all the files. But then seem to be either hallucinating or I just don't understand how it works. Display all the output fuzzy logic algorithm. You need to ensure the service and blah, blah. So unless it's a clickable link that doesn't actually work out and click on it, why? Right. Have to wait for it to finish. Okay. So in part show that the analysis data function, this is the wrong analysis data, all the output. Wait a second. This is all wrong. Is it actually looking at the code or what is it doing? So we have analysis data is currently not the function. We have it mentioned six times just a variable. This is wrong. We have process data, response data return. While we're returning it as Jason, not convinced it's actually looking at the whole code. Are we still timed out from GPT or are we? Yes. Until another 20 minutes. Yeah, 20 minutes. Shall we see how it seems like you ended DFG, which doesn't provide any can show me see if GPT 3.5 can write code. I need a product context for this thing. And obviously the GitHub co-pilot wasn't works with a function. Can you explain the structure of the whole project? Yeah, we have the waveforms J stupidity 3.5 over times up. Then we have this format document. Let's see if it's able to resolve basic stuff, generate contents, code, okay. We have the Python static, we have game logic, waveforms templates, index HTML test by which we're not currently using. So we're trying to display the feature values for frequency peaks amplitude and the abnormality likelihood. Essentially, what we have in here, control shift, see it is meant. Okay, I'll ask it as a question. Is this meant to be displayed on the front? Okay, GPT 3.5, that's a question. Compatible to the front bar, ProStat object contains the role of the frequency peaks, blah blah blah, and the output will defuse the logic. Yes, that's correct. You can display this data in your HTML using JavaScript. Yep. Why aren't these clickable stuff? Okay, results. I need an example. I need you to tell me how to do it. If I ask to co-pilot peaks, you end up getting based on the shit discussion. What nerve is it doing, showing in there? Well, at least this is a bad list, it's showing the actual output of the fuzzy logic. The amplitude is not changing. It should be one of those, or maybe this one has this high amplitude, but there is one waveform, prerecorded waveform that has a, not prerecorded, it's a simulated, so it's all synthetic data. But one of them is different, not this one. Yeah, the peaks should not be normalized. No, the amplitude is changing. But yeah, get abnormality and defined. Yeah, we really had some prompts for this thing to improve. Okay, the peaks should not be normalized. Okay, again, the peaks should not be normalized. I don't want just to not show them, but I also don't want the fuzzy logic algorithm to use the normalized peaks. X should not be normalized, I don't want just not to show them, but also I don't want the fuzzy logic algorithm to use the normalized peaks. But suggesting me just to hide them, but obviously that won't do it. Based on symbols, so it's still generating, why can't click on it? Save that. No, it's not generating, one should be able to click on those, I don't know, it's not working. So fuzzy logic algorithm is in a normalized positive peaks, the normalized function, process data, analyze waveform, fuzzy logic analysis, this one. How is it getting it wrong? There is no function called the fuzzy logic algorithm. There is a function called fuzzy logic analysis. So is this, we'll make sure this is actually using the not non normalized peaks because we don't need to normalize them. Okay, I just changed the same thing, process data. Yeah, this one is taking the normalized stuff. So positive peaks, negative peaks should be a row. Wait a second, so it's taking the whole positive peaks and that has a sub component of row and normalized. One question is, when we do the fuzzy logic analysis, since the API for the fuzzy logic analysis, okay, I'm confused by this one. How do we make sure that for the fuzzy logic analysis, we use the row, a negative and positive peaks, not the normalized version. No, no, no, no, it's not good. It's not good. The frequency is wrong as well. The frequency would not work anyway, would it? Yeah, ideally we probably should be, yes, so if we're looking at amplitude, we're looking at a number of negative and positive peaks, we should be looking at the distances between the different ECG components, the PQRST waves, and those should be within a certain range. Currently, we have abnormality undefined is, we have an abnormality value, it should be, that's what should be displayed. And the thing is, again, if we select workspace and the era, it should look at all the files in theory, so it's using, use six references. Pretty sure we have abnormality defined, okay, I think it's a case sensitivity, might be a case sensitivity issue. No, that one is a case, just a text in the printout, then everywhere else. That should be fine. Okay, let's go html again, the ID in fuzzy logic, in the game logic, that line there, yeah, that's a good idea. Where's the return of this thing, fuzzy logic analysis, called result, just change it like this, oops, we have an abnormality, no property of data, can you solve this for me? No, okay, we only have like 10 minutes or so, right, it should be, get to this, right, this is better. Okay, that's working, it should be 0.83, super weird. Okay, can I get read of that, just play it all the time. Okay, might have to enter this next time. Anyway, hopefully, so this tool is already on the website, you can go check it out, it's in the ECG game, you can play this game yourself, there's a video of me getting up to 1000 points, you can see if you can beat that or reach that or do any better. It becomes, it starts going faster and faster, I think every 50 points, it will increase the speed and hopefully sometime soon we'll have another tool that will show how a machine learning algorithm and explainable AI in the form of fuzzy logic will be playing this game. So you could compete with both labeling ECG data, questions, questions, YouTube, Twitch, LinkedIn, you can still put them under the video and I'll see you next time, bye, bye.