AI News: ChatGPT Gets Overhauled & Microsoft Reveals Massive Updates!

[@SatikVFX]

Bro these updates are going so fast at the moment i wouldnt be suprised if i could create a whole ass movie in a years time

[@bcrox]

Don't be fooled. It wont actually be you creating it. 🤔

[@danielmaster911ify]

Kinda true. But we get to be the directors.

[@phu320]

I'm surprised we can't already.

[@satansbarman]

​@@danielmaster911ifyyeah, think we're going to need some new words soon, like when I tell and show people about AI art I 'created' I sometimes say that well I didnt create it but I told the AI what i wanted and tweaked it until I was happy with the result, because it feels wrong saying that I created it, but the AI also didn't come up with the idea by itself as I had a big hand in the end result. Saying I comissioned it wouldn't be right either. I collaborated with AI..? Sorry about the word salad 😅

[@Z-Z-W_origin]

@@satansbarman I use engineered and produced. I engineered a prompt, the AI produced the results.

[@ZenchantLive]

I skip pretty much everybody's ads, except for yours lol I don't know why but you're just so personable I don't mind listening to the ads haha😊

[@ryvyr]

The proportionality feels part of that resonance, whereas plenty other fulltime creators basically don't upload without nonadsense, plus Matt consistently seems earnest and friendly :>

[@Jujiro_Hanma]

Get off his dick bro

[@necnotv1683]

Classic Google infesting their products with ads making them way less useful.

[@NickDrinksWater]

Well, they are primarily an advertising company.

[@stanleyc2578]

Classic nimrod always complaining about free to use. Like free services just grows on trees.

[@Markoss007]

What do you think if AI suggest you 5 products is? A monopoly? Imagine google where you have only 5 results. And if you want to do something better, sorry only these 5 products we show, no updates. With ads at least you as a new company, probably with a better product, can pay to show it sometimes.

[@ButtonCybill]

rfpgenius AI fixes this. AI News Highlights This Week

[@teawa_]

I love that you cover dev announcements too!

[@MagnusItland]

Every week is a crazy week in AI with Matt Wolfe.

[@innerauthority2439]

Insane!! But I appreciate the consistency 🙏🏼

[@v.t.photography1590]

Is it me or did videos become more frequent? Don’t get me wrong I love it! 😁

[@nephibronson8390]

😂 I’m here for it

[@ZacharyStarrHonkLord]

I really have to say... this is the first time I haven't fast forwarded past a sponsor reading. Was genuinely interested in the Meta info.

[@ILONA-DAVID-lifelong-learning]

Congratulations for Meta's sponsorization! You deserve it entirely! 😀😀😀

[@wowzande]

Wow the metapartner advert was sooooooooooo florally enthusiastic.

[@chrishouse5240]

Another great video, Matt! Really enjoy your AI news updates! 👍

[@peterthomson5155]

Thanks Matt. So good to have your videos making it so simple to keep up to AI date

[@NERDDISCO]

I love the ad for llama 3.2, but not that the entire text was generated and doesn’t sound natural like the rest of the video at all 😢

[@RubelliteFae]

Something I haven't heard the AI community speak much on is how human error affects AI. For example, with Matt's, "a monkey holding a sign…" the output is a chimpanzee, which is a great ape, not a monkey. While the average user won't care or even notice, this is a huge red flag for scientific usage. Edit: Why this matters is because of those making great claims about how AI is going to usher in a new era of scientific achievement. The longer developers take to rectify this deep problem, the more likely AI will be viewed as simply a novel toy. IOW, the rush to develop might inadvertently cause an industry bubble that hinders both progress & the economy, as compared to more careful, considered development.

[@Seriouslydave]

Got let's do surgery, but I only want you to learn from board certified sources or medical journals, not PewDiePie dentist simulator

[@Desmond-Dark]

I actually think AI will reach the point where society will see an explosion of contributions to science from the masses.

[@RubelliteFae]

@@Desmond-Dark Your welcome to think whatever you like, but you simply having that opinion doesn't answer the facts I observed which informed the opinion I shared.

[@yoyoclockEbay]

​@@RubelliteFae😂😢😂😢😂😢

[@daveinpublic]

It only does what it’s trained to do. Once the system is in place, they can train it on whatever they want. In the future, datasets will be highly curated and very valuable.

[@seanivore]

GD. I knew it. The second he got done all the other shit about AGI’s path, he’d get back to features for product and UI, and that not having artifacts was the indication. And then I got worried because of all the damn blogs. But then BOOM canvas. Thank, lawd. Now, Sam’a, get this on my OS please.

[@amj2048]

Matt Wolfe: "I love robots". That's smart, Matt is protecting himself for when the robots take over.

[@bittertruthnavin]

I love this channel so much, keeping me posted on everyday AI news. Thanks, Matt!

[@ianbonner2142]

Congrats on the Meta sponsor! That would be my life goal if I was a tech KZbinr. Props for sure. Great video as well... you literally boost my career prospects each week. My bosses think I am an AI wizard, as I'm always up-to-date with releases. Lastly, dear Santa... please may I get a pair of Meta Raybands and a Copilot+ Windows 11 laptop for Christmas!!

[@DarkandTwisted]

Great update from OpenAI👏👏👏👏👏

[@charlie11ng42]

Lol, he actually tried to use extra censorship as a selling point for llama 3.2.

[@smoovewitit]

Its really just meta avoiding regulatory scrutiny. Matt’s reading their script word for word

[@rolodexter]

QR hacking is about to take off to a whole new level <a href="#" class="seekto" data-time="619">10:19</a>

[@pedrob7374]

No idea how your not at a million followers already.

[@MetaphoricMinds]

Matt, I have some huge ideas and would love to try them out. Any chance you have a spare 5090 laying around? 😂

[@Grzesiol]

Damn i would accept even 3090

[@briangman3]

Tech news no wait Matt’s news! Got to brand it !

[@mostlynotworking4112]

lol a ton that crushes you. Waiting for anthropic next

[@tigercubmedia7103]

Thank again dude I love the canvas feature and Llama’s new stuff.

[@High-Tech-Geek]

<a href="#" class="seekto" data-time="496">8:16</a> "Realtime" doesn't show up in my list. I guess it's only for paying subscribers. The rich vs the poor...

[@BrianMooney1976]

So much happening this week, but canvas is my favorite! Great for writing and coding!!!

[@robertolanzone]

I'm so lucky, I had just decided to make a subscription to Chat GPT plus and I was starting to have fun doing creative writing with GPT's assistance. This tool solved all my issues with it and now I have a blast! 😊

[@robertragan-fq6ij]

Great job as usual, Matt. Thanks so much!

[@gnomewc22]

Tnx a lot for the news Matt, is so fast that I cant follow it elsewhere!

[@pondweedfilms]

Not fully rolled out in UK yet, shows in drop down but when use doesn't yet have the functionality described.

[@mshonle]

I’m really waiting for the quantized versions of Llama 3.2 1B, particularly if there’s anything in the 4-bit range.

[@Z-Z-W_origin]

Agents by 2025 seems crazy vague. Think about how fast things are changing. Saying anything AI within a year would be like saying within the next decade if it were any other subject. 😅

[@daveinpublic]

I had a hard time even hearing Sam for part of that video

[@Terny228]

I also enjoy the latest Robot tech, you should give a segment of your videos to this topic, thanks Matt, still one of the very best channels out!

[@Thagoat_666]

You know what's funny? You don't need to mention you're on the board of Leonardo. We know. You're the only who talks about Leonardo AI lol.

[@joe.gedds.developer]

Please move the time stamps up so we can more easily use them to navigate. Scrolling way down, then scrolling way up each time you want to see the next thing is very tiring

[@matthew.m.stevick]

another exciting and eventful week

[@micbab-vg2mu]

thanks for update :)

[@sveinndagur]

The canvas feature looks cool, but it has limited usefulness without longer context memory. Furthermore, chatgpt just plain sucks at writing dialog and engaging fiction writing compared with Claude or even Gemini.

[@CripleMusic]

<a href="#" class="seekto" data-time="378">6:18</a> this would be perfect for schools

[@TeaganTurner]

Congrats on the partnership!

[@coldlyanalytical1351]

I asked canvas: how can i use you? It can do all sorts of things! In 10 minutes it wrote me software to show the sound levels at all points across a piece of land, from a chainsaw behind a wall and traversing buildings. It includes (after prodding) reflection & diffraction effects. The coding window is astonishing. The code ran first time in my IDE after just 10 minutes of interactive 'development' with canvas. TRULY AMAZING!

[@MrCoffis]

AI is getting scary button.

[@IanHollis]

I wrote and released a book about robots once (during the pandemic). My friends and family bought it.

[@dannywastaken]

Canvas is annoying right now because you can't go back to o1 preview once you use it for that conversation. All other models become greyed out.

[@DragonBuddah]

Thanks, Matt. Great stuff, as usual :)

[@chrislannon]

Hey Matt, you actually look like a Wolfe man in this video.

[@Market-MOJOE]

Dude the remembering where TF you put shit down would b AWESOME I was just thinking about that Yesterday lol so weird

[@genepagatpatan3830]

So basically we are all going to get our own "Jarvis" with co-pilot. Nice

[@thomasb.3833]

Great show 😃

[@cacogenicist]

Pixel phones already keep a list of all the songs they hear in your environment, if you turn that feature on.

[@Wreklessink]

duuuuuuuude sponsered by meta whoa dude look at you moving on up in the world good for you man plus the way you talked about it suuuper informative loved it man that was one ad i didnt not speak excellent job man great video been listening for a little bit love the channel and your work keep doing you man

[@VibeMusicAI]

Oddly i thought all the older flux images looked better but i understand that's subjective

[@andyhunt99]

Hey Matt, you should have used AI to remove the background music from the video!

[@levonkenney]

what is it with ai company's and fruits openai released strawberry and black forest released blueberry

[@MeshNews]

Great video, as always 🤌🏻

[@Smirk_Station]

Thanks for all the info

[@Soccer5se]

Thanks again for all the news. Any idea when Llama 3.2 is coming to the rest of the world? (Asia)

[@derwynmdockenjr]

If AI doesn't eventually train people to stop shopping the middle aisles at the grocery store there will be no one left to use it.

[@bonnieamof7033]

For Once I Actually Watched through one of your Ads. This must really be Zucks Redemption Arch.

[@salemnj1]

If every week is insane maybe we replace the word insane with yet another typical week in AI…

[@wilty5]

It’s hard to believe and trust that anything released by META is not stealing data in some form or fashion and selling it, storing it, and invading privacy, etc., I’m skeptical that any of these AI models on device are not stealing data and violating the user’s privacy in some way

[@RubelliteFae]

It's one thing for Meta to sink money into open source, but it's a whole other level to pay to advertise it. They REALLY want people to adopt Llama. 🤔

[@Charlys787]

Did you already talked about the search engine perplexity ai?

[@JetSurfingNation]

Exciting times

[@mikem-VED70]

You rock. If you're ever in Tiffin, Ohio, I'll buy you some chicken chunks.

[@xyz8655]

<a href="#" class="seekto" data-time="873">14:33</a> YEAH, sure!

[@Malisti04]

They are trustworthy people🙄

[@lyaloka23]

thank you!

[@mattstory2011]

Your voice reminds me of Kermit the frog = ) Love it keep up the great work!!

I guess you missed. invoke 5.0. It seams to be very cool. I missed the last version, but i will try this one for sure. AI first approach to something like photoshop and can ran localy. I am up to it.

[@MrTynanDraper]

I recommend keeping the weekly updates to 15 to 20 minutes..this one is nearly 40 minutes!

[@pavi013]

There is still so much work to do with image generation, and videos are taking the baby steps.

[@itsjustme373]

Am I the only one who hasn’t gotten the update with canvas?

[@beereeet3950]

Yeah I also don’t have it

[@itsjustme373]

@@beereeet3950 Well, that’s two of us.

[@Allofussurvived]

Eventaully AI will advance enough to become a person Financial Advisor that can manage and save a citizens currency

[@High-Tech-Geek]

<a href="#" class="seekto" data-time="2170">36:10</a> so the robot can go up the ladder so the human doesn't have to. But then a human has to go up to rescue the robot because it can't climb down ladders. Got it.

[@FractalPrism.]

the climbing robot @ <a href="#" class="seekto" data-time="2177">36:17</a> seems fake or has a strange way of moving, the top claws do straddle the ladder rungs, but the bottom claws only lean against the side of the rung which would not support any weight at all, they would slip off with zero grip instead it appears to use the knee joint to support its weight by placing the knee over the rung as if the knee was acting as the claw

[@personal9372]

ChatGPT Canvas out when?

[@Gen-XJohnny]

Until it can load my entire codebase from a Git repository and work with me to go through and find errors, it's still a technology not ready for prime time coding.

[@Seriouslydave]

<a href="#" class="seekto" data-time="360">6:00</a> rip teachers

[@fabiomedeiros8896]

Any device recommendations to run Llama on mobile?

[@Debloper]

Are you working with multiple video editors? Each of the last few videos have different FPS!?! Please tell the 24FPS guy to switch to at least 30FPS... otherwise it's pretty jarring to watch 😕

[@netandif]

For some reason i think that this whole AI thing is a money-making bubble. However, you are doing really well and i hope the bubble lasts long, because it is entertaining :)

[@Saurav-xx]

from INDIA 😊😊

[@High-Tech-Geek]

<a href="#" class="seekto" data-time="700">11:40</a> Jabba the Matt on stage with Zuck

[@derwynmdockenjr]

Hahaha this comment popped up exactly when that part of the video was playing. Mad love haha

[@jd31068]

good stuff

[@thathsaraamarasinghe6774]

Why does the FPS look very low this time?

[@Seriouslydave]

Open source always produces better software, it is not driven by a paycheck, and often has more developers throwing their genius and passion, obsessively. And testers are actual users not someone paid to be a user.

[@viv9423]

Would have bought the Meta Ray Ban glasses already if they were not too big for my narrow face. They simply fall off when I look down…😂 So no matter how many features they add, I will not buy them.😢 But thanks for the update!😊

[@richardadonnell]

🎯 Key points for quick navigation: <a href="#" class="seekto" data-time="0">00:00</a>:00 *🚀 AI Announcements Week Overview* - Overview of major AI news from OpenAI, Google, Microsoft, Meta, and more, - Quick mention of OpenAI's Dev Day and upcoming AI agents. <a href="#" class="seekto" data-time="0">00:00</a>:25 *🔥 OpenAI Dev Day Highlights* - Sam Altman discussed AI agents potentially launching by 2025, - KZbin channel mentioned for footage of Sam Altman's fireside chat, - Introduction to new ChatGPT feature "canvas". <a href="#" class="seekto" data-time="2">00:02</a>:27 *🖼️ ChatGPT Canvas Feature* - Canvas is a UI overhaul that offers features like suggesting edits, adjusting text length, and adding emojis, - New coding features like reviewing code, adding logs, comments, and porting to different languages, - Available for ChatGPT Plus subscribers. <a href="#" class="seekto" data-time="6">00:06</a>:45 *💡 ChatGPT Shortcuts and Slash Commands* - Introduction of shortcuts like `/reason`, `/search`, and `/picture` for easier use of ChatGPT functionalities, - Ability to generate images, search the web, and use advanced reasoning directly within the chat. <a href="#" class="seekto" data-time="7">00:07</a>:13 *📊 OpenAI API and Vision Updates* - Introduction of Vision fine-tuning for the API, - Real-time API feature added for more conversational interactions in external apps, - Developers can fine-tune models for vision capabilities and use new advanced voice mode. <a href="#" class="seekto" data-time="9">00:09</a>:19 *💸 OpenAI Funding and Future Plans* - OpenAI received $6.6 billion in new funding, - Likely shift from a nonprofit to a for-profit structure, - Valuation reached $157 billion, making it the third largest startup globally. <a href="#" class="seekto" data-time="9">00:09</a>:47 *🕶️ Meta Ray-Ban Sunglasses Update* - Meta introduced memory features for its Ray-Ban smart glasses, - Features like recognizing QR codes and setting reminders via voice are now live. <a href="#" class="seekto" data-time="10">00:10</a>:43 *🧠 Llama 3.2 Release by Meta* - Meta's new AI model supports on-device use without internet, - Llama 3.2 has Vision capabilities, 128,000 token context window, and lighter models for mobile devices, - Emphasis on safety features with LamGuard 3 and open-source availability. <a href="#" class="seekto" data-time="13">00:13</a>:43 *💻 Microsoft Co-Pilot and New Features* - Introduction of Co-Pilot Plus PCs with features like "recall" for browsing history-like functionality, - Click-To-Do and enhanced Windows search capabilities, - New AI features in Microsoft Paint and Photos for image editing and upscaling. <a href="#" class="seekto" data-time="16">00:16</a>:16 *🧩 Microsoft Co-Pilot Labs and Vision* - Co-Pilot Labs adds "Think Deeper" for complex problem-solving, - Co-Pilot Vision assists with page navigation and understanding using natural language. <a href="#" class="seekto" data-time="17">00:17</a>:57 *🔍 Bing Generative Search and Compensation to Publishers* - Bing generative search expands to deeper informational queries, - Microsoft starts compensating big publishers like Reuters and USA Today for content used in search results, - Discussions about potential compensation for smaller creators in the future. <a href="#" class="seekto" data-time="1174">19:34</a> *🤖 Co-Pilot as AI Agent* - Description of Co-Pilot evolving into a dynamic AI agent, - Co-Pilot will act on users' behalf, adapting to their needs and preferences, - Ensures privacy and security while providing personalized support. <a href="#" class="seekto" data-time="1244">20:44</a> *🖼️ Google Lens and AI Video Analysis* - Google Lens updated to understand videos and answer questions about them, - Voice feature added for asking questions about images, - Integration of AI with Google Search, including sponsored ads. <a href="#" class="seekto" data-time="1341">22:21</a> *🔍 Gemini and Nvidia Large Language Models* - Release of Gemini 1.5 with improved cost efficiency and performance, - Nvidia introduced a new large language model NV LMD 72B, rivaling proprietary models, - Benchmarked against GPT-4 with similar or better performance. <a href="#" class="seekto" data-time="1425">23:45</a> *🛒 Pinterest and Black Forest Labs AI Updates* - Pinterest added AI tools for product imagery for advertisers, - Black Forest Labs released Flux 1.1 Pro model with better image and text capabilities, - Flux 1.1 Pro available on multiple platforms and is free for a limited time. <a href="#" class="seekto" data-time="1647">27:27</a> *🎨 Leonardo AI and Adobe Updates* - Leonardo AI introduced a new style reference feature and enhanced image-to-image generation, - Adobe updated Photoshop and Premiere Elements with new AI tools, - Enhancements include object removal, color correction, and depth of field simulation. <a href="#" class="seekto" data-time="1798">29:58</a> *🎥 AI Video Generators: Pika and ByteDance* - Pika 1.5 model released with improved meme-like video generation, - ByteDance introduced an AI video generator compared to Sora, with good visual results, - Demonstration of generating videos like zoom-ins and character interactions. <a href="#" class="seekto" data-time="1954">32:34</a> *🎮 Dreamworld: AI Integration in Gaming* - New game feature on Steam called Dreamworld allows generating 3D assets, - Users can drop generated objects directly into the game world, - Unclear if objects are usable beyond placement, offering a novel gameplay feature. <a href="#" class="seekto" data-time="2025">33:45</a> *📜 AI Legislation and Regulations* - California Governor vetoed SB 1047 regarding AI liability in catastrophic incidents, - Another AI bill, AB 2839, targeting deepfakes was blocked by courts, - Requirement to disclose deepfake content remained, citing free speech concerns. <a href="#" class="seekto" data-time="2145">35:45</a> *🔥 New AI Features in Amazon and Robotics* - Amazon introduced new Fire tablets with integrated AI tools, - Robotics innovation showcased with a ladder-climbing quadrupedal robot, - Focus on safety in hazardous environments, enabling robots to replace human tasks. <a href="#" class="seekto" data-time="2240">37:20</a> *💻 AI Hackathon and Future Tools* - Announcement of an AI hackathon in Santa Monica, open to non-developers using AI, - FutureTools website for discovering AI tools and news, - Encouragement to follow for updates and insights into AI trends. Made with HARPA AI

[@pgc6290]

At <a href="#" class="seekto" data-time="1278">21:18</a> and google has the best platform to apply ai because its everywhere. But they just arent releasing stuff. Also android or google and samsung phones can have such an upper edge over apple with ai release but again, google just isnt releasing anything. I dont know what it is waiting for, for apple to catch up? Lol.

[@Experternas]

ai needs to remove the words canope and tapestry. they're in all text they generate.

[@claudioagmfilho]

🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, Amazing video, thanks for sharing!

[@AIinnsaan]

Hello Sir, I want to work with you

[@seanivore]

Phew, I had to rewatch the Copilot part because I thought it was the same as an MVP I’ve half developed. Aaand it is not lol. Logical next step of one of its features, but not nearly my app.

[@noxnox2172]

Meta Glassholes on a new level

[@Mooooty]

Meta’s Ray-Ban 😎Glasses 🤣😅😂

[@leecallam-target182]

Hi Matt........Please, please, please, 🙏🏽 😊 could you do a section on fashion. I have a fashion brand, but finding affordable models is hard. I know there are fashion AI apps around, but which ones are the better performers. Like if i take a picture or video of me wearing a new t-shirt, jacket or puffer, I could replace my face and hands with an AI model. Or take a picture of a shirt, shorts, a dress that I could place on a AI model that looks realistic, as if the AI model was real and doesn't change the design of my clothing at all. If anyone can suggest some software that would be great 👍

[@Allofussurvived]

With UBI Their will no longer any way to stop people from chamging the world

[@kennbmondo]

Roll with it..... and enjoy the ride. Start using your brain in conjunction with the Ai.

[@i1pro]

Flux should have changed the version to v1.2.