 I don't know, just done it in a second, still checking the sound and everything else. I'll share my screen in a second as well. And yes, this is my real face. Just in case you were wondering, I don't know why the lips are not moving properly, something to do with the light in the room as well. Okay, sound is okay, so I don't have to want it to myself to speak clearer as well. That helps, doesn't it? Just share this on LinkedIn, just in case someone wants to join in to know how the thing works. It posts LinkedIn posts, it's not clear which link it's actually selecting. Okay, we have to change the order of the links, I think. So yes, you can provide live feedback on a YouTube stream, just rephrase that. If you've already checked up the data, send a blog somebodykills.com, see which link it's no feature. Yeah, now it's the YouTube channel. We can post that, close this case. So it should be live, sound should be okay. Let me know if otherwise. So as you already know, we are working on this website called bodykills.com. This is a bunch of blogs. And this one is the most recent one on the ethical concerns around neural implants. And yes, this was about 100 prompts in Chagabt for the text, another 50 or so for the images. Primarily this one, yeah, no, it's not for the light-hearted, but it is a fake image. But yes, it's made to be, to look more realistic image of neural implant being expanded. We actually have for a change relevant or not quite, so might as well close that ad. So read this blog, provide your feedback. Yes, it was mainly written by Chagabt, but yeah, as I said, about 100 prompts back and forth. Hey, we also have this EG2 music converter. Should work, might be loud when loading from memory channel, one of the channel. It's loading channel 7 by default, but actually I think channel 14 sounded more cheery. So this is a seizure on that channel. It can actually load the whole thing. It's about 100 seconds of seizure on 16 different channels. So they look, as you can see, quite different, differently on the different channels. A default window sizes, 10 seconds. You can scroll through it, it will convert the delta, theta, alpha and beta band power into a corresponding musical note. We'll play four of them at once, like a musical chord. They tried out, let me know what you think. Do you have any comment suggestions to improve it? Please do let me know, can do it in one of those live streams or yeah, during a YouTube live stream will be better because I can actually, I won't actually forget to address your comment, but hopefully it's relevant as well to the tool. Get a lot of non-relevant stuff. Now, yes, there's a bunch of other tools that you can check out, hopefully you already did, but again, if you did, don't forget to provide your feedback. And currently we'll be focusing on the last two on this list, phasologic neural network. So primarily we will have some examples that we already had in the past, if you open the tool currently, it's not working. It used to work. There's a video of what it meant to do or how it used to work in the past. The idea is, we want to grab this image uploaded to chgbt, so it has a reference of what it should be doing. Can we do it now? Do we grab that window, print, screen, straight into chgbt? So what did this used to do? Now, does it control C or just print screen? Control C, close that. We already did this before. So this is what the tool meant to be doing. It's an example of a phasologic detector, waveform detector. We select an input waveform. It's actually essentially an envelope of the time series waveform. This could be GECG or anything else. You can add more examples later. We adjust the noise and then the phasologic is doing classification based on different parameters. It will go over these parameters in a sec. Essentially, we can do it now. Essentially looking at the membership of symmetry and the peak, essentially the number of peak, the sharpness of the signal. So if it's a triangle, it will be sharper than a square. And symmetry, based on symmetry, could detect the difference between, say, a square and a left triangle, because the square will be symmetric. Left triangle would not. So left to right triangle will be kind of similar to an envelope of spike during a seizure or a QRS complex during ECG. Let's see what's this. Here we have an input wave selector, noise amplitude control, output wave selector. Yeah, it's actually not a dropdown, it's just an output in square. Yeah, so that in that example is the false detection, because the noise is too large. The input is a hexagon, but it's detecting a square. And this percent detection error is a very high statistic display graph. Yes, education purposes, waveform, evaluate, performance of a fuzzy logic-based detection system, and that's right. So essentially what we're doing now is converting this tool into Python. We have both GitHub Copilot and GPT-4 trying to help us with the task. Currently, this is what we have running, it's giving us a note. Yeah, there's a bunch of problems, so the noise works okay. The text is out of work, we can fix it later. We won't worry about it right now. Yeah, let's see if we get a bunch of errors, get an unexpected token, JSON file is not valid, shell script line 59, that's where the error is coming from. And let's see if we get an error in the back end as well. Length of x and fuzzy membership function must be identical. Let's clear this quickly, restart. Right, so the post, we only have one post requested at the moment. Scroll to analyze, it's giving us this error. Okay, we'll try and get both this error in Python or JavaScript. Python, I have to select the whole thing, they pop this in and do the same with GPT-4. There's a mismatch in length, the possible fix doesn't, it doesn't look like shit, I don't kind of trust a GitHub co-pilot. The responses from OpenAI website are normally more comprehensive, surprisingly didn't give, didn't suggest the code, it did not suggest the code. Session, and aggregates fuzzy membership function, right in the fix, session error, it's finding aggregated bit. Yeah, okay, we're happy to try it. Yeah, we had this problem before, we had the plotly loading for quite some time. So we know why detection error is wrong now, so it's showing a square and there are no errors. Okay, that's not cool, so that means the output is always a below one third. By the way, this is the only template at the moment, so essentially the current configuration the output that is looking at the symmetry, as always let's print this out against this clear terminal. Yeah, the output is super low, at least it's changing. That's a good sign, it's always a very low number. This is the tree you would expect for the left triangle, just do a sinusoid, essentially only square works. Input waveform only works when we choose square as an input for all other inputs, it goes to the default option, which is the sinusoid. How do we fix that? Yep, should be talking like that all the time, then probably YouTube will think I'm a robot as well. Maybe I am, I don't know, I don't know. Put wave, default case, it's always going to that else, a default case, generate signal. The same question, oh github copart, we've given github copart, we actually have to select the code that you wanted to look at. Can that function generate signal? Okay, we have that logic already, and yes, else the default option is the sinusoid. Yeah, we have to make sure the signal there matches, it should seem to be the general function, the function is a string comparison to determine the type of waveform to generate, and if the input doesn't match square, left triangle, etc., the function falls to yes, this could be the case sensitive or extra, blah, blah, blah. Can you fix that? Is it case sensitive? fx, go on an else, return signal. So replacing that, that should not be case sensitive anymore, doesn't work. So it's actually not in the code what you are suggesting to convert to low case and move white space, you can print wave, input wave again to the conversion, left square square, left trapezoid, left trapezoid, okay, but there is no match. It's a sail, trapezoid, okay, go away, more troubleshooting. It's always this little annoying things, input wave, left triangle, whatever that is, it's definitely not a sinusoid, that's for sure, so only square. Oh, come on, yeah, I'm not sure, it's a, it's a head the better, the luck of the new chat with gpd, it's cool for the whole being the html file, shell script, we're doing it because we know we can hold more context, shell script and the item code, check the sums of cases, sensitive issue, verifying stream matching, getting gpd4 got stuck, the co-pilot theory when you do workspace should be able to take whatever it needs from all the files, fix these strings to match, yeah, so that workspace never, never works, you just do it manually, square, so not actually the parameters that we, the square can't get rid of that one, don't actually have it, there's something wrong with that to order that, so I have square trapezoid and this is the default, the signal length is 100 dot the points, okay, the wave is of left trapezoid, should have square working okay, plotly is loading forever, it's the left trapezoid, okay I'll fix that a bit, we have right trapezoid, we have left triangle, right triangle, hexagon and sign as the default, we really have to fix that plotly loading forever, and left triangle, right triangle, it's making sense, hexagon, okay, she insist, what they should look like, a hexagon, that's what a hexagon should look like, and it's the envelope of the signal, you're also having some sort of timing, a time issue, the x-axis of it might be wrong, that's what the hexagon should look like, like this I don't know, something like this, exactly like this, no it's the same thing as before, no it's not defaulting anymore, that's okay, yeah I don't think I interrupted this image correctly, in the image there is a red hexagon, can you provide a function that will generate a similar envelope of the signal as the output wave, yep, yeah it now picked up the 200 from the image on the x-axis, 100 should be fine, a fixed 100 signal length should be okay, we also don't need a separate function for it, just the piecewise should be signal hexagon wave, and if nothing, if no strings matches it does the same, but I should not see the size of it anymore, rising edge, first vertex, top flat line with 30 is 100, dividing the thing into six, and let's see, the hexagon again, and so it's dividing the thing into one, two, three, four, five, six, yeah I know what the problem is, this is meant to be two waveforms instead of one, okay let's solve it, this is what I get for the hexagon, how to fix, yep it's still dividing into six regions, the first region two minus here, rise to the first horizontal top, down to the middle, back to the start, go from here, should be better than what we currently have, how many errors, now those errors should be okay, yeah that's wrong, you can't get it right, no it's the waveform you imagine provided, it has to be a square wave instead of hexagonal, just to maybe initiate the yes to fix the generation, adjust the conditions, this is still incorrect, a candidate would check, do we need to separate waveforms for the top and the bottom parts of the envelope, yep, yeah that's right, we need to find it with two flat sections and four stems like this, this is the same, this would not work with and happy to try it out, spending too much time on this already, yeah I want to get to the fuzzy logic bit, we essentially need two waveforms for the top and the bottom shape of the envelope, yep okay keep repeating the same thing but it doesn't work, this doesn't look like an envelope of hexagonal shape, do we need to split it to two waveforms, first it looks the same, this would just change this slope, it's playing dumb, the square is a square, even the whole sign or something, I know it's not a sign, it's a sign, it's clipping what's left, it's making sense, it's having trouble and it's close, keep up, go power doesn't give us anything useful, it's too deep, it's wise, it doesn't go right, we have the hexagonal shape, a print screen, generate wave hexagon that looks like this, a good basic geometry does it look like a hexagon as in the original image provided, yep that's the hexagon that you generated looks like the image I provided, where the hexagon is built out of two waveforms for separate for the top and bottom envelopes, yep it's interesting how its logic works, let's say yes sometimes it's like an advanced developer, in this case it's playing a bit silly, more like a, so the buff should start from zero, one five by five, that's a separate top and bottom envelopes, yeah none of them are correct, just do the top, control f5 it, let's run it again, let me just do one envelope, just the top part, yep one two three four five should have, let's divide into six parts, no five parts, yeah it's just wasting time, but that's making sense, the output doesn't look like a hexagon, it's obviously a contradiction between Python and the JavaScript code, it's JavaScript in Python, which code generating the waveforms, is it to JavaScript or the Python, yep okay, so what's this button doing, are the waveforms being generated by Python or JavaScript, yep are the waveforms being generated by JavaScript or Python, the waveforms are being generated in Python, how do we update this JavaScript?