AI is NOT Slowing Down! Open AI Voice Release Date, and Cutting-Edge AI Innovations

[@MattVidPro]

HUGE thanks to Hubspot for Sponsoring today's video! Checkout Breeze AI Copilot Here: clickhubspot.com/duab

[@ThymeHere]

Huzzah

[@LouisGedo]

👋 hi

[@JojoSmacks]

good job matt

[@sychrov81]

Hi, for last 3 year my biggest test is, to try chatgpt to generate python script to generate 3d model in blender...keeps failing..otherwise...in programming..

[@NicVandEmZ]

Do you think it could make your sims look real

[@AIprotopia]

Matt, It's incredible how quickly AI videos have advanced. In such a short time, we've gone from being amazed by the technology to now critiquing details like skin tone and other fine points. The progress is truly remarkable.

[@ThymeHere]

This is actually really cool. Cant believe how far AI has come, Especially in the last few years. I rember when Dall-e 2 was the coolest thing ever and now it’s just unimpressive.

[@nickgirdwood3082]

It was cool, but it didn't do much.

[@teelo523]

It's crazy how fast things feel normal or outdated

[@henrythegreatamerican8136]

Most of the corporate released A.I. stuff has generally been limited by censorship. They want everyone to play nicely, only allowing you to imagine sunshine and lollipops.

[@randymulder9105]

Would it be cool if it showed the entire process...as it made the image? Odd thought Or a robot making art. Seems yucky. Lol

[@kneelesh48]

Dalle-3 is unimpressive now

[@fynnjackson2298]

Imagine taking a 3 year break from the world, living in a cabin and then seeing how far AI has come. just a single week and so much happens, 3-10 years....... like what the even f

[@change2023now]

At this point I believe a permanent break living in a cabin in the woods would be a sensible idea.

[@BionicAnimations]

It's so funny hearing about Advanced Mode coming on the 24th because I have had it for over 2 months now and have been using it daily. I was one of the lucky few who got access early. I hope you guys get it. It's freaking awesome. Sometimes I forget it's not human.🥰

[@INTELLIGENCE_Revolution]

I had it when it first released - just briefly. I’m waiting so patiently to get it back. I don’t think people can appreciate how good it is until they try it!!

[@WhatLurksBeneath]

Just a friendly PSA to stop and smell the roses because this stuff is moving so fast it's amazing!

[@AdamB1_23]

Omg the 24th actually lines up with what Altman said of having a couple weeks more patience. Could this finally be it please lord

[@Saif-G1]

It's already here Go see for your self.

[@AdamB1_23]

@@Saif-G1 it isn’t I just checked, it’s the regular voice mode and I’m a plus subscriber, but it’s also 1am so I’m going to sleep

[@randomn793]

I read it "Antman" 😂

[@saksham3]

Still no advance voice mode 😢

[@Sunyatasattva]

“Versus” Code Studio cracked me up 🤭

[@FinanceLogic]

Very cool that even people who call it versus code studio can build stuff. What a time to be alive.

[@IceeSwirl]

Calling Black Myth Wukong a "weird furry man" is insane LOL

[@abadidibadou5476]

I think he never plays video games haha

[@OMARD55]

@@abadidibadou5476 He does listen here 25:26

[@IndigoGPT]

The hope is that videos like these put pressure on OpenAI to actually deliver instead of just advertising everything for years and never releasing.

[@MattVidPro]

Me too, I don’t think it does though.

[@jonorgames6596]

I tried the chatgpt o1 model preview with university level math, and it is just better than me. It makes proper substitution to solve things I didnt think of. It is insanely powerful. Truly astonishing. AGI in 3 years? Much sooner imho.

[@JAK85.]

thats crazy, but I still think within the next 5 years is more likely, for something to be declared AGI it has to do a lot more than math

[@Saif-G1]

now chemistry and coding are the most difficult for ai.​@@JAK85.

[@antonystringfellow5152]

@@JAK85. 5 years sounds like you're thinking linearly. We humans have great difficulty thinking exponentially and that is what is required here. If you've been following the development of AI models for a few years, you should be aware of the acceleration. Not just the acceleration but the accelerating acceleration. - That's what exponential is. I know we're not close to true AGI yet, that's why I think it'll take 6 months to 2 years.... thinking exponentially.

[@cbnewham5633]

​@@JAK85.And it can't even do that effectively. AGI possibly by 2028/9, but it will require a breakthrough beyond large language models.

[@14supersonic]

​@JAK85. From my understanding, AGI is the ability to do most economically viable work as well, or better than a human. This includes both virtual and physical space. However, if an AI can completely master virtual space, then it wouldn't be that much harder to transfer that capability over into an embodied form, i.e robots. Imo an AI would suffice for AGI if it can master either. o1 seems to already be so intelligent that giving it proper access to the right tools it would be able to master virtual space relatively quickly. GPT was already smarter than most humans with 4. We are on the tipping point, AGI v0.9. I expect that GPT5 will be the first model to unlock AGI, in less than a year from now.

[@Yipper64]

Am I the only one who read the title and assumed that it was OpenAI's voice mode that got a "release date" and thought that meant it was a date for when its guaranteed for all plus users?

[@orthodox_gentleman]

Yes in the beginning of the video it shows 9-24 as the next date….

[@matty.j_1997]

Yes.

[@nate_d376]

So do any of you have any voice capabilities with chatGPT?

[@4.0.4]

NotebookLM really surprised me. It can generate podcast conversations from any set of links, PDFs or other text content, and it's super natural.

[@14supersonic]

I've been using it for research, but you can practically use this in any way. It's mostly just limited by your creativity. It could use some additional features to control the output of the generation though.

[@4.0.4]

@@14supersonic I think they were very narrow on it because it's both a prototype and a proof of concept. They don't want everyone using it yet, and probably don't know if it has a viable business model either. So being just English, casual style, gives them a good idea for how people use it. Do people load entire books? Dozens of small txt files? Websites with images that don't work? Etc.

[@Satrina777]

AI, tech in general grows exponentially, not linearly. Prepare for everything to shift rapidly. You thought the jump from Y2K to 2020 was crazy, you'll be stunned by where we'll be by 2040.

[@Icemanr85]

some examples please to support your statement

[@Icemanr85]

@@Tone_Of_Dials is your mom a man

[@nickgirdwood3082]

I'll probably be dead. 😢

[@nickgirdwood3082]

​@@Icemanr85What statement?

[@HUEHUEUHEPony]

someone has not heard of logistical curves

[@MichaelTrompeta]

Sorry to be that guy, but it’s V-S Code (not versus) and it’s short for Visual Studio Code

[@morena-jackson]

I was looking for this commnet

[@tasteyfoood]

“Weird furry guy” 😅

[@oufukubinta]

Like how the internet is one of the defining inventions of the last century AI is a defining invention of the 21st century

[@AAjax]

The one thing to note on the IAME graphs. The X axis (train-time compute on one graph, test-time compute on the other) is log scale, not linear. So there's actually diminishing returns that aren't being represented.

[@TheAero]

Insane progress is happening!

[@EmperorCaligula_EC]

EU or at least here in Germany we still not have the normal improved voice to due to some stupid regulations. :(

[@LeonGustin]

To think at the rate that all these AI are improving already is only going to exponentially accelerate even more!

[@BrianMosleyUK]

Surely advanced voice mode should be a part of GPT-4.5 at this point? Would be amazing if we get 4.5 in 4 days time.

[@DJ-dh3oe]

Why?

[@BrianMosleyUK]

@@DJ-dh3oe Why put all the effort into adding Advanced Voice mode into GPT-4o when GPT-4.5 is ready to ship? I think the only thing truly slowing them down at the moment is politics.

[@DJ-dh3oe]

@@BrianMosleyUK Advanced voice mode is natively a feature of 4o, they've just been holding back the release of it to fix some problems (scaling, weird behaviour etc.) Advanced voice probably won't be possible in 4.5 anyway, since it will be a larger and slower, so not capable of real-time speech.

[@Ramiassaf55]

@@DJ-dh3oewe got something much better, the O1 model

[@DJTechnosapien]

You are now a coder Matt! If you ask ChatGPT to explain the code to you could learn from a PHD level teacher. I want to start using more AI to help my channel, some of it requires coding, so all the examples u shared are very inspiring! Thank you

[@se777en73120]

Why did the Moshi demo appear to be one-sided? I heard the green audio but not the purple audio.

[@v-sig2389]

I think it is to not confuse people between the ai voice and the real voice.

[@apache937]

cut

[@Ricolaaaaaaaaaaaaaaaaa]

"This weird furry guy" lol that's Wukong Matt! Good vid though lol!

[@HUEHUEUHEPony]

who?

[@Ricolaaaaaaaaaaaaaaaaa]

@@HUEHUEUHEPony Sun Wukong

[@apache937]

@@Ricolaaaaaaaaaaaaaaaaa who?

[@Ricolaaaaaaaaaaaaaaaaa]

@@apache937 look it up

[@friendlyyeen4844]

Have you heard of a multi-prompt 3d environment generator called WonderWorld? I'd like to see you cover it. The paper is called "WonderWorld: Interactive 3D Scene Generation from a Single Image"

[@Gilster85]

Thank you for starting off with the news we are all most interested in!

[@nemonomen3340]

Almost interested in.

[@JPeoplzOIII]

Who else thought he was about to say "Im a massive fan of SHENMUE" lol

[@missoats8731]

Me

[@Bottleshare]

Hey man! It's awesome when you cover anything open source! Anyway you can cover Bittensor or anything from any of the subnets? Unless it's your thing, ignore the digital currency tied to it for now. Bittensor is decentralized open source AI on the Blockchain. All subnets talk to each other, so when there's a breakthrough in one, there's a breakthrough in all. It's what humanity needs to keep up with centralized, closed source AI like OpenAI, Microsoft, etc.

[@ShadyTalesyt]

Eleven labs has amazing tts ai voices.

[@markmuller7962]

Incredibile stuff btw, I can't believe we're getting used to AI building a software or a physics simulator in one go. Also the video to video is groundbreaking, we can expect an explosion of alternative/remastered videogames footage on youtube. Everything else also crazy stuff, AI is at full throttle

[@ChannelWaugh]

Great roundup Matty. Perhaps it’s time to use some AI to generate some better options for decorating your background. I mean there is nothing wrong with the living at home with mom vibe. But seems like a good ai related project. Whether it is an ai artificial background that evolves with ai improvements OR just designed by ai and physically built, seems like it would be a good project

[@MattVidPro]

That’s honestly a really good idea. Appreciate the comment!

[@V4VestA]

The Future is yet to come. Wondrous and MARVELOUS will be the things we see!

[@claudioestevez61]

22:30 This "weird furry guy" is Sun Wukong (aka Monkey King) the protagonist of a Chinese novel called 'Journey to the West' and this specific image is of a video game poster called "Black Myth: Wukong" that portrays the novel. The novel has been read by hundreds of millions of people and the video game has sold tens of millions of copies. I haven't read the novel or played this game, but I think you just might have angered a lot of people.

[@not_riley]

also the basis for the creation of Dragon Ball and Goku

[@RM-xr8lq]

OP is type of person to call hercules from mythology as "dwayne the rock johnson" 😂

[@neilmiller4611]

Wow! I learned a lot from this video. It covered products I haven't seen in other videos. Good job!

[@tomgreen8246]

Moshi is interesting being open source. Nvidia Mistral Nemo r2.1.0, even on my single RTX4090, is incredibly capable of on all things voice and even NLP. It does better than anything I've experienced online; closed or open source. It's closed source too, so potentially you match these up (on a rig far more capable than my own), train the models together and you have a much higher quality product. Oh yeah, Nemo is very scalable and fantastic at training.

[@FusionDeveloper]

1 LLM to reach AGI and ASI would be nice, but simply allowing AI to be module, such as using a calculator, makes more sense.

[@jaredf6205]

I’m just waiting for the day that an AI researcher is using an AI and the AI suggests something and the researcher thinks “huh, I never thought about that” and then implements it to improve the AI. It’s all downhill/uphill from there.

[@Joeherman]

Can you all imagine where ai be in ten years from now? Kling looks awesome and I loved that Gen3 video recreation. Will we be making independent animated movies in our bedrooms? Speaking of which, I actually did make a movie in my bedroom in 2013 and it just got put on Tubi! (which I am pretty excited about). By the way, it is also about Artificial Intelligence 😀 If you like animation and a good scifi story, it would be great if you had a little time to watch it. It's called The Mind Machine. If you do decide to watch it, let me know!

[@dasomen]

As usual top content! You're awesome, amazing workflow, thx!

[@DiceDecides]

I have an idea for developers: create some model that can use an LLM to interact with a video model, image model in a specific way based on a prompt you want and an ability to post a whole book and have it create a film script itself and use it in parts to prompt the video, image models to get the correct visuals based on your description. maybe possible with o1?

[@contentfreeGPT5-py6uv]

You can do it now, try to test with o1 you can free test it, i don pay, i ger everything free. Anyway thanks for that.

[@DiceDecides]

@@contentfreeGPT5-py6uv oh wow I had no idea gpt can create images for free now, since when? haha thanks

[@uncletimo6059]

Meanwhile, Pi AI has THE best human voice out of all AI. By far. And it has been that way for over a year.

[@cbnewham5633]

It's just a shame that they are no longer updating it and are soon going to charge for using it. Pi is the only AI I've used where I've had conversations and forgotten I was talking to a machine.

[@apache937]

i really like the voice from what i heard but idk lazy to even google the site name

[@JM168M]

What a time to be ALIVE 🔥🤟🔥

[@cbnewham5633]

Hold on to your papers fellow scholars.

[@JM168M]

@@cbnewham5633 LOL 😆😆😆

[@TheSuperStroker]

Of course the baby looks a little crispy. Where do you think the saying "bun in the oven" came from?

[@ginocote]

The question I always ask myself is: Why are Kling's videos always in slow motion?

[@christiancarter255]

I was baffled months ago when I heard some people saying that AI would be slowing down and other things.

[@OliNorwell]

Great summary, gonna check out the paid Kling option, though maybe Runway3 is better?

[@cbnewham5633]

My general use of these generators for image to video (I tend not to use text to video) is that Kling is the best and Runway a close second. I recently tested real physics for a handful of generators and Kling was the best for real physics - although there is still a long long way to go. The first physics video is on my channel. I'll be posting up some others soon once I get some spare time.

[@cbnewham5633]

9:37 one thing I've noticed with Kling and other generators is that they tend to flex objects that are in motion that are supposed to be rigid. You can see it here - the gun moves up and down from the running and there is a slight bending of the weapon. It's subtle, but once seen you can't unsee it 😄

[@jaminunit]

That script was sooooo GPT! When you say “buckle up “ you know it was AI generated

[@cbnewham5633]

Maybe 😄 The big give away for OpenAI AIs is the use of the word "intriguing". For some reason that word is used a lot, often combined with blend: "intriguing blend of...".

[@starblaiz1986]

I've been saying it for over a year now - OpenAI is WAAAAAY further ahead than everyone realises. They just like to wait for everyone else to almost catch up before releasing the next step change.

[@cbnewham5633]

Nonsense. If they are that far ahead they'd not be waiting for competitors to catch up. That is not how companies work.

[@eduardomartin8510]

The V2V is impressive.

[@lupusk9productions]

Wow this is insane. it will soon become too hard to tell what is real online anymore. or over the phone, etc.

[@picksalot1]

GuardRails AI should be used for Fact Checking Politicians.

[@garryb.1334]

Sun Wukong really settled down when he met Iron Man.

[@JAK85.]

The first AGI just has to be called "Strawberry" at this point, I know thats o1's codename but I mean like officially

[@JohnSmith762A11B]

Greaat video. Inference scaling baby! 🥂

[@BabyDogOfJustice]

I can't wait for the youtube game movies / tv series being run through this, like Mass Effect or Warhammer 40k etc.

[@Crazy_Truth]

Will advance voice model available for free users!!?

[@BionicAnimations]

I have the Advanced Mode and have had it for over 2 months now (one of the small group users who got it early). I don't think it's coming to free users, at least not for a while now.

[@bakablitz6591]

cant wait for artists to riot after omnigen drops!! so hyped

[@dl8337]

Love this! You have a great delivery!!!

[@IDKHonestly]

Will the new voice mode have access to the memory storage? It will literally be useless to me unless it keeps its memories.

[@westingtyler1]

23:40 I want there to be a colorful accuracy bar on all videos, so as people say things in videos that are true or false, it changes color or goes up and down a slider, bringing up links to credible sources. would be amazing.

[@esimpson2751]

Genuine question for you mat, why do you think Kling or Minimax is better than Sora? from the demo videos we have scene Sora really is the superior model, the downsides are not its performance but how censored and inaccessible it is

[@ThymeHere]

2MI 13 L 4 C 273K S 52 V 20 SEPT 2024

[@JustFeral]

I just want better NPCs in video games. Then a game where you have to play as a cyberpunk noir detective 😂

[@Krommandant]

So when can we see the AI eat and transform every ebook and audiobook and movie in any format in one fell swoop?

[@HUEHUEUHEPony]

in the coming weeks, just buy nvidia stock, and pay subscriptions

[@gpierce6403]

Nice weekly recap

[@richardede9594]

Great video. Thanks for your hard work and keep it up.

[@hellohogo]

now it's here

[@chrisw2823]

Me when im in court and see an AI video of me doing a horrible crime I didnt actually do:

[@DorianRodring]

Will Teams accounts get advanced voice mode?

[@milo8425]

22:30 Lol, calling Sun Wukong "a weird furry guy". 😄 It's like calling Merlin "a weird Gandalf guy".

[@mindbuildingmusic]

Where are you pulling this information from?

[@strangereyes9594]

Does anyone remember this startup that aims to make ASIC chips just for inference? I remember they claimed like a 1000x speedup like the ASIC miner did for bitcoin mining. That means anyone could run a 10t or bigger model on a local machine with ease with minimal energy cost, especially if the chain of thought reasoning is so inference heavy and those chips are tailored exactly for that task. Damn, what company was that?

[@alexmipego]

I don't believe the "real AI" concept is going to slow down… but I've to explain the "real" part first. You see, having a computer chat with you is nice and useful… but if you think about it, that's not really what we humans want or need. We need something that understands us, and can communicate with us… and the trick is that the rest that we want (the truth, reasoning, knowledge) is already something humanity mastered for a while. The point is, an AI that generates videos is impressive… but there'll always be "holes" in it, that's the math of the thing. However, an AI that opens blender, models the 3d scenes, and does the camera animations, and then gives you the files for you to edit… that's something else entirely. LLMs (basically), are just the "universal translator" from Star Trek. Use that to manage inputs/output of proper databases, applications… and things change drastically.

[@digitalatscale]

+💯 on inference scaling

[@SamWilkinsonn]

Thought voice was going to be out ages ago

[@esimpson2751]

if it really gets rolled out on the 24th Ill be content, but I dont think it will

[@SamWilkinsonn]

@@esimpson2751 even if it does, I don’t get the excitement. It’s not that amazing is it? (Or I should say the novelty’s worn off now where the announcement was seen ages ago. When it gets released I won’t feel the same buzz as when it was announced)

[@tuckerbugeater]

@@SamWilkinsonn heroine drip

[@SamWilkinsonn]

@@tuckerbugeater wut

[@cbnewham5633]

So did we all 🙄

[@not_riley]

It’s September 24. No voice update.

[@felixgraphx]

"Versus" code studio 😂

[@cbnewham5633]

Wow. Amazing. Astounding. But too little too late. I'm now using Claude and i don't think the release of AV is going to make me jump back to OpenAI. I tend to dislike companies that make announcements and then don't deliver. It's not far off lying - and that may be ok if your users aren't paying you money, but if they are, then that's a red line for me. I'm not paying you to lie to me and tease me about vapourware; I'm paying to use your product and be given realistic release dates for new features. Who knows, OpenAI may even release the mythical Sora - but given how good other AI video generators are now, it would have to be an order of a magnitude better than the demo reels we have seen. I'm not saying i would never go back to subscribing to OpenAI, but every time they pull a stunt and don't deliver then my view of them becomes dimmer and dimmer.

[@ShannonJosephGlomb]

Omnigen wow ❤❤❤❤❤❤❤ all of this is epic ❤

[@whizadree]

What if you change the date on your pc to 25th heh

[@Agispsi]

Bro just called the monkey king "this weird furry man"... 😮😂

[@Raulikien]

Yup, videogame rendering pipeline is definitely getting replaced by AI. Or at least remasters will be made like that.

[@matgilroy]

Thanks for your updates

[@fynnjackson2298]

the chess game was insane. Soon our only limitations will be our own imaginations.........

[@kevinsm2039]

You know it’s crazy. They already found the AI can be more creative than humans so you can already scratch that out too.

[@MrNobodyX3]

20:25 "Verse Code Studio" 🤣 its just V S Code or Visual Studio Code

[@ozhinz]

lmao fr

[@Lugmillord]

VS doesn't stand for Versus, but Visual Studio ;)

[@randomn793]

20:23 "Versus Code Studio"

[@Airsoftshowoffs]

Great video but your video Description is very messy to find sources fast and easily.

[@needsmoredragons]

Matt, are you reading off an AI-generated script? Your dialogue and delivery don't sound how they normally do when you speak in your videos (which is far more natural and human-like). If I'm right, please get the model to give you far more human-sounding scripts. Your genuineness makes your personality shine through; I've read, created, curated and corrected enough AI-generated material that things don't sound right... "so buckle up folks, because we're about to take a peek past the curtain into the future, how about we start this off right and talk about the big guy in the room OpenAI" just screams AI generated script. I love your material, you are one of the people I actively keep up with for AI news / content, keep it genuine to your exceptional personality mate.

[@MattVidPro]

I can send you unedited video files if you’d like. I never use scripts (except for sponsored segments) I’ve been doing a lot of recording and scrapping more videos lately. It’s all off the dome!

[@HoboGardenerBen]

To me AI has barely begun. It's not a part of normal reality, just online. It hasn't improved my physical interaction with the world. It will be real to me when it helps make some new supermaterial or cleans up the garbage in the ocean or something. All the current applications of AI are creepy and I don't want them in my life. I want it to do stuff in the real world. Main real application I can think of is scary AI drones.

[@AzraelRuin]

😂😂😂 90% of the world is online at this point and it’s only going to become more engrained. How old are you?

[@HoboGardenerBen]

@@AzraelRuin I'm 43. All the most important aspects of your life still happen offline. You can't eat the internet or plug yourself in. Direct interactions with friends have much more bandwith than online interactions. If you're young, that comment just makes me sad because you guys don't even know what you've lost, you grew up under the influence of big tech psyops so it all seems normal. It's not.

[@AzraelRuin]

@@HoboGardenerBen LMAO. Ok boomer.

[@HoboGardenerBen]

@@AzraelRuin Wrong word, that's my parent's generation. But I accept the general old person burn behind it, definitely entering my stiff old man phase, it's fun. I like to go to bed early and get up early now, and birdwatching is a good time lol

[@spectralvalkyrie]

Hey you look like a grown up in the thumb nail 😛 thanks for the updates

[@JoelSapp]

I put in a video request for kling it took so long i forgot about it until this video (2 days later) and the vide is still not ready. WTF.

[@helloworldcsofficial]

Great recap!

[@Yipper64]

17:44 essentially they found another area to optimize, but I think that will be optimized, and then we will level out again until someone else finds some out of the box way to get even more out of it. But essentially we hit a limit in base intelligence, now we optimise that base.

[@missoats8731]

But at one point the AI will be so capable that it can find the next area to optimize by itself

[@Yipper64]

@@missoats8731 maybe. And then hit the ceiling.

[@agnesslovehealz]

We all will get gpt plus users by sunday minus few countries and magne gpt model will get more access on api wevb version and if wanf restrictions and worried should have tiers 18 over its all about money they give access to companies cuz follow the money

[@favesongslist]

Kling is incredibly slow right now, I have been waiting days to complete :(

[@nickgirdwood3082]

How many AI KZbinrs named Matthew are there? Because that sounds like some kind of conspiracy, maaaaan!

[@similarvideo]

That is pretty cool 😎 voice model