Vollständiges Transkript von „Google Takes on OpenAI in EPIC Showdown!“. Klicke auf einen Zeitstempel, um das Video an dieser Stelle zu öffnen. Du brauchst ein anderes Video? Nutze den YouTube-Transkript-Generator.
0:00 Hello there. I'm Mike with this week's new and trending AI tools you cannot afford to miss. And this is no green screen behind me. I'm currently in Phuket, Google and OpenAI, both trying to outshine one and the other. It's pretty crazy, and there have been a plethora of releases from both. We looked at Google Gemini 2.0 last week, and the real time interactions were pretty insane. If you haven't seen that video, go back and watch it. But now let's look at the updates for this week. We'll take a look at OpenAI at the end of this video. But Google have really been crushing it. Gemini Advanced now lets you upload your whole code folder so that you can get full context on a project maybe you've been working on for a while. This is great if you want to update things, make your code better, implement new features where the AI knows everything you're talking about.
0:48 I think this is going to be an incredible new feature for those Gemini advanced subscribers. Now I'm presently in Google AI studio and I'm going to switch the model to the new one they've just released this week. It's called Gemini 2.0 Flash Thinking Experimental. It's presently free as well and comparable to ChatGPT o1. And for this I've got an experiment here in my text editor. I've got a code base, and it is a tool to help YouTubers respond to their comments with the help of AI. Obviously comment Genie and I want to make the code more efficient, make it more readable, follow best practices, fix any bugs, and I've got a place to insert the code. And I'm asking it for lots of great feedback. Now I've printed out all of the code here in this file. I'm going to copy it into the prompt that I've designed with the help of AI.
1:33 Copy all that code, paste it into the new experimental thinking model here, and you'll see it's going to count through how many tokens I'm using here. I'm well within my limits within this small coding project. And let's run it and see what comes back. Now we can expand to view the thoughts, and immediately it's thinking through quite a lot of stuff around my project. first of all, after it finished thinking and it did all of this thinking, which is pretty crazy, it's telling me about database query optimization, how to cache contents, how to do things like break down. I've got one main Python file at the moment, and it's telling me, actually, I can break this out into authorization API services models, database config and so on. So I don't keep all my code in one Python file, which I'm well aware of. Well, the model picked up on that really well. And then I can go ahead and bring this back into cursor,
2:20 update my project and make something pretty cool. Gemini 2.0 flash thinking experimental is a really cool release from Google, and I'm looking forward to playing with that more. I also wanted to go ahead and try notebook out for this and actually chat with my code base. So I'm copying that whole code base, going to copy text and pasting it all in here, inserting. Let's generate a deep dive podcast about my code. Okay. And once the audio overview is generated, I can now enter interactive mode. Hey everyone and welcome back. We're diving into the world of what AI in YouTube. Oh. Sounds interesting. Yeah. It's, it's this web app called Comment Genie. I can imagine that'd be pretty useful, especially for, you know, bigger channels. hey, our listener wants to join in. What's up? Hey, could you tell me some ways this app could be improved? Some really cool features that could be added to it.
3:07 One thing that jumps out is the potential for deeper analytics. Oh, you mean like beyond just sentiment analysis? Yeah. Imagine if Comet Genie could identify trends in viewer questions or recurring topics of interest. So like it could help creators find gaps in their content and like generate ideas for future videos? yeah. Can you help me come up with a catchy tagline for Common Genie? What about Comment Genie - Never miss a comment again. Oh, that's a good one. It speaks directly to a pain point that creators have and a little bit more playful. Comment Genie - Make commenting fun again. That's incredible. Let's make commenting fun again. You can see how this is basically a mastermind with two excellent podcast hosts, and I look forward to playing with interactive audio overviews from NotebookLM much, much more.
3:54 Now let's move on to Google Whisk. first of all you can start off by entering the subject. Now you can either generate an image or of course you can upload an image if you have one. Of course I have an image of myself that I can upload. This could be fantastic for making thumbnails. Then for the scene again I can enter text for generative AI or I can upload an image. And of course I've taken tons of images here in Thailand where I am right now. And then again for style we can enter a prompt or upload an image. So I'm going to just type in Photo Realistic Boom. Look at that. We've got our three stars our subject, our scene and our style. We can also add additional details or just click the arrow to generate. Similar, but nothing like me. Really? It's kept consistency with the scene there, and it definitely is photo realistic. We can roll the dice to get more ideas.
4:41 Maybe we'll stick with me and we'll go for this beach pier scene. And then for the style. Let's go for this wonderful Japanese style drawing. And again. Here we go. We have our very own Japanese style portrait. Again, close to me. Not exactly me, but you know what? Not looking too bad. If we like, we can click to refine. And for the instructions, I've typed character is wearing a tropical shirt. Just like that. It's done. Looks pretty cool, right? Now in the premium section of the Creator Magic Community. As I mentioned at the start of the video, I'm doing an AI advent calendar with the new AI idea inside five minutes every single day. It's been pretty incredible. We've spoken about things such as home automation, AI, image generation, and of course, AI assisted coding. Well, today is the 20th of December. Just four days to go now until Christmas.
5:26 It's pretty crazy. Doesn't feel like it out here, by the way. And I'm going to share with you on my YouTube channel. Now, one of the AI advent calendar updates. And it's actually today's update, which is pretty incredible. I've been playing with Replit and the Replit agent, which you probably know is a pretty hot tool, and it's getting lots of updates. I created a really cool Google Maps Finder app in just seconds using the agent, and it actually allows me to find awesome places wherever I am. So while I've been traveling through Southeast Asia and Australia, it's been incredible for typing in my precise coordinates and finding great places to eat there. Let's watch the video from the premium section of my community and see how it works. So today I wanted to look at Replit, but more specifically so I love Replit. It's amazing. You can create stuff in the cloud.
6:13 Their AI agent Replit agent is amazing, and I've got a course inside the premium community you can watch on how to use it to create stuff. I used it to start creating me something called Map Finder Plus and I've hooked it up to the Google Maps API and I'm actually quite happy with it. I'm actually going to run what I made. So this was all made using Replit agent. I said make me something that can search Google Maps and find me the best places or things to do in an area. So it's like a command line tool that I'm running here. Let's see if we can make that bigger. And it asks me for latitude and longitude. So if I actually go here, this is part of this is basically central, a downtown Kuala Lumpur in Malaysia, which is, because I'm traveling this month, that's where I'll be right about now, by the way. And if I right click and grab, the latitude
6:59 and longitude, it's actually also up in the URL bar there as well. Let's grab the latitude, which is that, and put it into this little, app that I made. And then we'll grab the longitude and we'll place that in zip and to the radius, in meters. What should we do? Should we keep it tight? Let's keep it relatively tight. Let's do just in the radius of 100m or a kilometer. Let's make it half a kilometer. 500m. Oh, what do I want? Within 500m of central Kuala Lumpur? Find me good restaurants. Number one. ZIP. Look at that. That's so cool. And it's literally found me the best restaurants in that specific area of Kuala Lumpur. Within half a kilometer of independence Square in Kuala Lumpur.
7:47 So it's not searching the whole of KL. If I'm in that specific region, I can just fire up my app and it uses the Google apps. Sorry, the Google Maps API, to search and and I built this all out just using the Replit agent. So pretty insane, right? So if I want to get a really good feed and what it does, by the way, it's not like Google Maps where you search and it algorithmically shows you places. This is literally scanning the whole of the map in that, half kilometer circular radius. And it's it's using an algorithm here. So I've asked the agent to do this, find me places with good ratings, with a good number of reviews, and run a weighting algorithm. So don't show me a place with, like, you know, two reviews, but they're both five stars. Obviously. I don't want to see that. Show me places with lots of reviews and high ratings.
8:34 So algorithmically speaking, this is the very best place to see. I wonder if I can copy that and find it in my map. If I'm in that, if I'm near Independence Square in Kuala Lumpur, I should go and eat at this restaurant. So let's see. There it is. That is where I should have dinner. And I'll tell you what. Dinner looks pretty good. So there you go. Yeah. It's nice, isn't it? That's that's very nice. So I can use this for whatever I want. Now, if I want, I can go ahead and, And I can do that, that search again. So we can actually go ahead. I'll do this one more time just to show you. So let's, let's run that again into that latitude, and then we'll go ahead and do this again
9:23 and enter the longitude right here. And this time we'll go for the, 500m again. And shall we look for, let's look for a nice hotel, stay near Independence Square. Boom. The Cosmo Hotel, Kuala Lumpur, that is the place. And then followed by the big M hotel. So it's just a bit of fun. It's just showing that if you can find a good API, like Google Maps API. They do charge, but I think they give you quite a generous amount of free credits every month. You just have to be careful not to go over that amount of free credits, because it can get quite expensive, quite quick with the Google Maps API, but for a few quick searches, if you're hitting a new city or a new location and you want to know the best hotels or the best restaurants, this is a really cool app. And feel free to discuss it down below. If if you find it handy. If you think this is interesting
10:09 and you'd like it, maybe we can do something more with this. I can, kind of release the code or talk about it in a future video, but hopefully that helps you get building with Replit agent. Use something like the Google Maps API and create your next amazing app that goes to the moon, and maybe even put a payment gateway on it and, and start making some money, because that's really why we're here. We're builders using no code tools, using AI to make incredible stuff. So there you go. Replit agent. And recently they also added Replit assistant, which is the smaller scale updates and changes to your app. And it costs a lot less than using the Replit agent, which is now capped, but you get a generous $25 a month if you're on the call plan, like me to use on Replit agent design. So go out there, create, make amazing things and share them in the comments. Or of course, in the creative magic community there is a free section
10:57 anyone can join, or you can join premium and get all the AI advent calendar updates and in the new year 2025, we'll be doing live streams, interactions and a lot more there. Hope to see you inside! Founder member pricing is still available if you join up before the end of the year. So as you'll see from this video, genuinely, I'm very excited about Google announcements. But let's not forget open AI. And here are some of my favorite announcements so far on the 12 days of open AI, starting with day one, obviously o1 and ChatGPT Pro. Big upgrades and huge for people who want the most advanced reasoning models available. I didn't find this announcement very interesting on day two, but I guess if you need reinforcement fine tuning then this is for you. Then the Sora! We finally got access.
11:43 I had to play with the Sora Turbo version, and I have to say it was a little disappointed with the results. And then Google came along and released Veo 2 and I was like, wow, Day four canvas. I've had a little play with this, but frankly I'm enjoying using cursor too much day five is an excellent announcement for those with the latest iPhone. I don't have it, but ChatGPT just a click away is super exciting. Okay, day six Advanced Voice with video in Santa mode. I got to tell you, this is by far my favorite announcement. It is incredible. I've been using advanced voice mode all the time while I've been traveling, and it's helped me out incredibly. I can point my camera at things around me, particularly when I'm in a tourist site or location. I visited Angkor Wat temples. I've been in Malaysia, I've been in Thailand,
12:30 and I've just pointed my camera at things and said, what is this? Even animals crawling around and immediately advanced voice with video identifies it. you're looking at a gecko. Possibly a Tokay gecko. It was a bit tough when I was trying to do a real time translation at a bridge in Cambodia, and I asked it to speak in Cambodian, but it was defaulted to Santa’s voice. So it started speaking in Cambodian as Santa. The phrase where is the toilet? Translates to bunker no nuh. Ho ho ho ho! Day seven projects in ChatGPT. I can see how it's going to be super cool to have folders where we can organize knowledge bases and consistently chat with them. Then there’s search. I've already had to play with this. I think it's great, but I'm sticking with Perplexity right now. On day nine, they announced their latest
13:16 o1 Pro model in the API if you're on tier five. So if you spent more than $1,000 with open AI and obviously it'll be rolling out to more tiers soon, I think this is also incredible for developers who want to use the full power of o1 Pro. Day ten the phone number I thought this was really very gimmicky. I won't be using this at all, nor will I be using the international ability to message via WhatsApp. But I did think it was a fun announcement. And day eleven working with apps. Well, I've already tested out the ChatGPT desktop app working with Xcode, and you can go and watch that video. It was really cool, by the way, and I think this is an incredible feature. But the question is, what is coming up for the day 12 announcement? A lot of speculation around this. And as I release this video, it hasn't quite been announced, but
14:04 but Sam Altman definitely dropped a hint. And if we have a look at X, he said, ho ho ho. See you tomorrow. Find one clue. Should have said oh oh oh. And already X is saying under the explore trending topics that open AI's o3 model. The next leap in AI reasoning could be on the way. Well, in a few hours from the release of this video, we'll know what it is, but I'm looking forward to all the announcements whether they’re Google or OpenAI. They're definitely expanding and enhancing where we are with AI right now. Oh, and by the way, YouTube is showing a video on your screen right now that you should watch next. Hey, it's Mike here, just with a few closing thoughts. Thank you so much for choosing to follow me. I only started this channel at the end of March this year
14:50 and look at where we got to now. So grateful for you joining me here and in the community. And 2025 is going to be even more epic. So I'm going to take a little rest now for the Christmas break. And I'll be back fresh in 2025. A very merry Christmas and a happy New Year.
