Daniel Bourke finishes Udacity's AI Nanodegree with a speech-recognition model that trains in 1-2 hours instead of 3-6 weeks. After submitting, resubmitting, and getting reviewer feedback within 24 hours, he moves on to fast.ai courses in 4-6 weeks and invites viewers to join him on Telegram.
Chapters
Guess what I'm about to do? Thanks for tuning in. Welcome back to another episode of Learning Intelligence. I just finished the code for the final project of the AI Nanodegree. Check this out. It's not even, it's not even that exciting actually. What did you say? It's not even that exciting actually. So it's not even that exciting actually. It's not even that exciting. Look, Dan, I'm from the future. I'm the future you, and I'm telling you this is exciting. Now get back to the video. Don't tell the viewers that what you're learning isn't exciting, because this stuff, this stuff is exciting, all right? I'm from 2048. I know what it's like. I know what this stuff evolves into, all right? Now this stuff is exciting. Back to the video. So what it was was an end-to-end deep learning speech recognizer. So this was the input True transcription: Out in the woods stood a nice little fir tree. What's even fir? Anyone know what fir is? And then this is the prediction my model made: Oh, I though sto in miss le for see. Not too great, but there's a reason for that. A video I showed in the past, in one of the past episodes, I think it was episode 31, was a talk by Adam Coates from Baidu who showed a state-of-the-art end-to-end speech, uh, speech-to-text model. That model would take 3 to 6 weeks of training. See, training the state-of-the-art model would take 3 to 6 weeks on a single GPU, whereas mine spent, spent only about an hour to 2 hours maximum training. So that's, you can see where the difference is there, and mine trained on a lot less data than what the full state-of-the-art model would do. But I'll show you, let's do a live submission. Well, not live, this is gonna be a couple of days old, but you get the point. We're gonna submit it together. So here we go, the first submission. I've included some questions here. I wasn't sure if my comparison of the models is up to scratch, just asking for some feedback from the reviewer. And we're gonna go 3, 2, 1, submit in before. The browser timed out. Oh yeah! Wait, where'd it go? Woohoo! Great work, your project has been submitted. Review will be back in 24 hours. Excellent, I can't wait. That's, as you can probably tell, my screen is orange. It's definitely time for me to go to bed. It's 9:05. I try to stop all screen usage by 9:00 PM so I can go to bed, be asleep by about 10:00, wake up between 6:00 and 7:00-ish. That's when I like to wake up in the morning. But if you want to see my code for the project, it's on my GitHub. Check it out. I'll leave a link in the description so you can check it out. I'll check back in once I get some feedback from the reviewer. Fingers crossed we're going to pass. This is the last project. Woohoo! Okay, so there was a small issue with the first project I submitted. I forgot to submit the HTML version of the Jupyter Notebook. But I fixed that up and look what's here. So where is it? Yeah, there we go. Issue reviewing a project. That's what happened this morning. And then look what just arrived. Your project has been reviewed. So let's check this out. We're both going to do it live. You ready? Oh, oh, let's get this full screen. What do you reckon? What do you reckon? Yes! I haven't even read everything yet. All green ticks. But I'm concentrating more on— I need to run. I'm so puffed. I'm even sweating a little bit just after that quick run. You know, you run Spartan Race and you do one lap of the hill and it's all over, it's all over. We finished! We finished the final project of the AI Nanodegree! I'm so stoked! I haven't even read what the reviewer gave me, some feedback. Your comparisons are fine, but please read these, some links here. I'll include them in the description so you can read them too. Congratulations on finishing the AI Nanodegree VUI capstone project. I'm so stoked! I can't believe it. Now I don't know what to do next. I think I'll have to refer back to my AI master's degree. I think fast.ai is up. Have you guys heard of that before? Would you like to do it with me? Let's, maybe we could get that going somehow. We could all, we could all do it together. I don't know how it would work, but maybe I'd livestream me doing it. Is that an idea? Leave a comment below if you'd like to see that because I could give that a go. I don't know how, how well it would work. I haven't even looked into the course yet. Okay, so we've been checking out the fast.ai curriculum and it seems like there's two main courses. One for like an entry level, like an introduction to deep learning, and the next one builds upon that. I want to do the whole thing, right? And it's 7 lessons in each part, 2-hour videos for each. So that totals 14 total videos. And the info of the course says that you should dedicate about 10 hours per week across the 14 weeks. But I want to do it faster because the course is, is, it's fast.ai after all, right? So, We're all pretty good learners, right? So, I want to, I want to get the course done in maybe 4 weeks, maybe, maybe 5, maybe 6, I'm not sure. So, I'm going to be starting in 4 days after this video comes out. So, that believe that's going to be June 27th or 28th, one of those days. But if you want to do it along with me, I figured live streaming is probably not the best idea because of the videos and the notebooks and whatnot. So, I'll make some, some, I'll document some of my, my journey throughout the, throughout the course anyway. If you'd like to join in on a group chat, I'll start one up. Message me on Telegram and then I'll send you the invite link to wherever the group chat is. My Telegram is t.me/@MrDBurke or just Mr. D Burke, I'm not sure. I'll put the link below anyway. We're going to learn some AI fast. But I'm so excited to finish the AI Nanodegree. I've just emailed support to send me my graduation details so I can get that shiny little certificate, maybe hang it on the wall or something like that. I got the sign out message. It's been here the whole time. Have you? Did you see it? Did you see it? It's there. Keep learning, team. I'll see you next week. What's cracking, bro? Hi, bro. Guess what I just finished? What? I just finished the final project of the Udacity Artificial Intelligence Nanodegree. That's a mouthful. I know what that means. Yeah, yeah, yeah. So, um— You proud of me? Yeah, very proud. Give me some accolades.
