4:16 WTF JUST IMAGIN EDUCATION SYSTEM AFTER THIS YOU CAN LITERALLY FEED A CHAPTER AND LET IT GENERATE. A short and crisp podcast covering all the important things and and you and ask questions in between like you would with your friends . ITS DEFINITELY GOING TO CHANGE HOW WE DO THINGS 5 YEARS LATER
@JohnSmith762A11B i think thats a way better world than this one where a person is pushed to go in scools and colleges for 15-20 years 20-25% of their whole life to learn things which are hardly useful after a point and spend whole life doing jobs and trying to maintain work life balance all the time forgotten the whole point of existence . Look at cave mans or pre industrialisation era people didn't they lived a happy life do they had money or luxaries we have today . Money is a man made concept , luxaries is a man made concept Languages are also man made concept so how can you say this is the purpose of life then . I think if these all boring and useless things got auromated or atleast require way less people than it needs And once people addapt to that lifestyle . people can actually focus on learning about nature and doing things which are actually worth it .
@@JohnSmith762A11B thats actually right now. but if they were smart they would use their education and methods of learning to advance in the AI field. nothing lasts forever, humans in general have always adapted to overcome or got left behind.
It's neat, but listening to a podcast merely makes you _feel_ smart about a subject. Only by working through sample problems and writing detailed answers to specific questions do you gain a deep grasp of it... and why bother when you can ask an LLM to do your homework? We've had thousands of great Khan Academy videos for school subjects for over a decade... where's the leap in student abilities? Will Khanmigo powered by ChatGPT actually deliver on the promise?
Matt, as always, thank you for keeping us updated. Have a Merry Christmas, and take a deep breath because you will show us AGI and human robots next year.
I love your videos, but it's getting boring to just show pictures and videos. What does AI do in other areas, like medicine, science, research and so on?
Good question. Nvidia says it uses AI in all aspects of its business, and they seem to be doing rather well. Employment in software is flat even though there's new a torrent of new products and updates, suggesting that AI coding assistants are making a difference. Individual researchers like the math wizard say AI is helping them a lot, but details are sketchy. Businesses may consider the details of how they use AI a competitive advantage they're unwilling to share. Consulting companies make claims about big productivity gains from AI, but they're trying to sell their services. Bottom up, if you're not using LLMs to do your job better, you will be replaced by somebody who is.
@@annonymbruger Did you not see the presentation this morning? The guy said now that O3 scored that high on this test. They are going to change the test and make it more difficult. They are literally changing the goalpost because its lower closer to AGI than anything has ever been. That is amazing and terrifying because AGI is within sight
BS google took the lead for sure, by the Time o3 is out, a lot of the players will have catch up, the same way sora looked great until it’s not Even in the race for top contender anymore
As an amateur video creator, very much looking forward for testing all these new AI video models! Only tested Runway so far. Would be fun to try out Sora when available in EU. Veo2 also seems great! Invideo probably opens up many new possibilities. A bit overwhelming, not to complain. :-)
Gemini 2.0's reasoning capabilities seem like a significant leap forward. It’s intriguing to see how AI models are becoming more interactive and context-aware, this raises the bar for all of us in tech.
Matt, why don't you have a membership subscription here on youtube? I think it could be really interesting creating a private community with all the people that is interested in this field
I was hoping for a dall-e3 or dall-e4 update, but nope. Starting to feel as though OpenAI doesn't really seem to care much about image generating any more.
“Consumers are the real winners”, really…? Must be the Sales rep ‘agent’ speaking. Consumers are giving away their ‘intelligence’ and therefor part of their own identity… it must be Christmas for big tech.
No AI beats Anthropic because no other AI has an artifact window where tou can see your prompts as websites, assets, modules, etc. You see what you ask for if and thats a big if but if you can pet Claude the right way you can create OMG amazing stuff yall. I created a prompt (30) that generated and entire working WebAp builder. Works front to backend and in between.
AI image generation won't be a solved problem until we achieve real character consistency. Also, we need to be able to describe more than one character/element of a scene without those characteristics bleeding over into the other characters. And of course there's always the censorship to deal with, but obviously the AI companies don't consider that a problem.
1:57 not really. objects with many parallel straight lines still look curved, random blobs in the background,... i would consider it almost solved, when you can really exactly describe how to change, refine image and it does exactly that, without any artifacts
Matt, can we get some coverage on how Ai is used to harm people? Currently there is a debate on how Ai is used in HR and healthcare to automate processes that were done by humans. With devastating effects. I would personally appreciate a thought piece about it by an expert. Thank you for considering.
If Google so-called won with the latest round of updates, then why do I have no urge at all to cancel my sub with ChatGpt and move it to Google? Well after some thought the simple anser is most of what Google released is useless to me, they barely released anything practical. 2025 will most likely decide who I stay for the long run, for now ChatGpt has everything I want for now but that can change
We are at a point where I do fear AI chatbots deployed with naferious intentions could pull these tools together to lay waste to the web with dead internet domains filled with an amalgamation of the best of these tools in use together.
nvidea seem to be going the useful way with dedicated engines and now wow with its jetson nano at a reasonable price of $300 (cheaper in china) to run them on
once these things can ask me questions and then debate my answer with its own i will feel better to go and purchase one but as my personalised AI system I have built does that then why pay for a roadster when a bowl of cornflakes will do (in a bowl color of your own choosing)
free is good to try as there are so many choices to make a decision on what suits you best .... like those awful rows of corn flakes in stores and all you want is something to eat, no toy, just some milk and no membership to i got coupons or anything as silly as some aborted penut butter sandwich
@@coldlyanalytical1351very good ,cheap and available. O3 is also good O3 is expensive but because of compute requirements and nvidia is going to solve it with faster gpus and new chip architecture . And it is going to be avalable for all .in fre monthes I think both model have their pros and cons but i think open ai has an upper hand over here . Anyways anyone is up the win will be of AI .
O3 is super good. But it's very expensive, I have herd that for High compute it cost 1000$ per task. Per task not monthly. Low compute verson costs around 10$ per tasks, which is also expensive. I am not trying to talk shit about o3. It really is great. But you still née cheaper model of not so high end taks, Google Flash Thinking fits perfectly in that market.
My Pika generations have been running for hours and none have completed, lol. Can't really test if you can't see results... oy! My Kling generations say the estimated waiting time is "longer than 3 hours", lol. Come on!
You made me an app in which I am a virtual AI 3D avatar teacher in which you can ask me any question and tell me in very simple language in Hinglish language explanation with comedy and a good sense of humour and sarcasm, basically such as explanation like never bored with the teacher, enjoy with study. And there is such a screen on which you have to write and the question to be asked is related to the image or video. If the video is required to explain the question, then the video or else the image means which is correct according to the question. You can see the image or video and The image or video is not copyrighted and for commercial use, why do I have to upload this app on Play Store and add movement in 3D avatar and also add lips movements feel realistic and like something explaining someone movement then you can ask me if you need any files., you need the link or file of the 3D avatar and environment and API key and I provided any more than asked me. just take this app and make it professional please make tutorial for this 😢
With all due respect you very wrong about image generation being able to generate anything. It still has many limitation, try get it to have someone throwing a ball, doing a high kick. Hitting something with a weapon. A train or bus interior that doesn't look like total mess... It still cant do any of those. If you think it can do anything, you need more imagination. Basic portraits and landscapes it can do easily, that's a long way from 'everything'
we have to watch out for the early scams when the likes of microsoft came out with a 'new' version but only changed the menu icons or buttons and merged all their products to have a bit of each in each other (eugh!) not a new skin more like a slap of slime
Chatbot arena is an ABSOLUTE JOKE...gemini is so extremely far behing from chat-gpt...and i would really like some competition there. but for now...its dimensions how smart they look. Gemini 2.0 is something like gpt 2-3.
You listened to my feedback!! thanks so much for putting chapters on the TH-cam video, really appreciate it a lot!!
I don't remember him not using chapters
"What a time to be alive!" Love the 2 minute papers callout
Hello fellow scholar. 😁
@@jtjames79The scholars have found eachother!
4:16 WTF JUST IMAGIN EDUCATION SYSTEM AFTER THIS YOU CAN LITERALLY FEED A CHAPTER AND LET IT GENERATE. A short and crisp podcast covering all the important things and and you and ask questions in between like you would with your friends .
ITS DEFINITELY GOING TO CHANGE HOW WE DO THINGS 5 YEARS LATER
The future: lots of very educated people and no jobs because AI can do them all faster and cheaper.
@JohnSmith762A11B i think thats a way better world than this one where a person is pushed to go in scools and colleges for 15-20 years 20-25% of their whole life to learn things which are hardly useful after a point and spend whole life doing jobs and trying to maintain work life balance all the time forgotten the whole point of existence .
Look at cave mans or pre industrialisation era people didn't they lived a happy life do they had money or luxaries we have today . Money is a man made concept , luxaries is a man made concept
Languages are also man made concept so how can you say this is the purpose of life then .
I think if these all boring and useless things got auromated or atleast require way less people than it needs
And once people addapt to that lifestyle .
people can actually focus on learning about nature and doing things which are actually worth it .
@@JohnSmith762A11B thats actually right now. but if they were smart they would use their education and methods of learning to advance in the AI field. nothing lasts forever, humans in general have always adapted to overcome or got left behind.
AI will be able to turn off caps lock for everyone!
It's neat, but listening to a podcast merely makes you _feel_ smart about a subject. Only by working through sample problems and writing detailed answers to specific questions do you gain a deep grasp of it... and why bother when you can ask an LLM to do your homework?
We've had thousands of great Khan Academy videos for school subjects for over a decade... where's the leap in student abilities? Will Khanmigo powered by ChatGPT actually deliver on the promise?
Matt, as always, thank you for keeping us updated. Have a Merry Christmas, and take a deep breath because you will show us AGI and human robots next year.
Google dropped the ball early but now the giant seems to have woken up
What a time to be alive!
I am addicted to your channel. The best of all update about AI
I love your videos, but it's getting boring to just show pictures and videos. What does AI do in other areas, like medicine, science, research and so on?
Good question. Nvidia says it uses AI in all aspects of its business, and they seem to be doing rather well. Employment in software is flat even though there's new a torrent of new products and updates, suggesting that AI coding assistants are making a difference. Individual researchers like the math wizard say AI is helping them a lot, but details are sketchy. Businesses may consider the details of how they use AI a competitive advantage they're unwilling to share. Consulting companies make claims about big productivity gains from AI, but they're trying to sell their services.
Bottom up, if you're not using LLMs to do your job better, you will be replaced by somebody who is.
Nothing.
@@skierpagefrom another country
@@skierpagetrue
I agree, there's too much focus on video and image gen.
o3 wins the week because it makes people change the goalpost on what AGI is
😂😂😂😂
Bullshit - I haven’t shown anything but nice looking presentations
@@annonymbruger Did you not see the presentation this morning? The guy said now that O3 scored that high on this test. They are going to change the test and make it more difficult. They are literally changing the goalpost because its lower closer to AGI than anything has ever been. That is amazing and terrifying because AGI is within sight
BS google took the lead for sure, by the Time o3 is out, a lot of the players will have catch up, the same way sora looked great until it’s not Even in the race for top contender anymore
@@annonymbruger as they usually do when the competition launches the really interesting, available and therefor usable innovations or applications.
As an amateur video creator, very much looking forward for testing all these new AI video models! Only tested Runway so far. Would be fun to try out Sora when available in EU. Veo2 also seems great! Invideo probably opens up many new possibilities. A bit overwhelming, not to complain. :-)
7:27 I used this model to solve one of the hardest maths and physics questions and it nailed it
Everything looks great and hopefully that continues to be the case when people start using these AI tools routinely.
Google Whisk is only US, and veo 2 is not in EU
😥
Thanks Ursula.
Only a small select group have access to Veo 2
Thank you for all the work you do. I don't know how you manage to cover so much
Gemini 2.0's reasoning capabilities seem like a significant leap forward. It’s intriguing to see how AI models are becoming more interactive and context-aware, this raises the bar for all of us in tech.
Matt, why don't you have a membership subscription here on youtube? I think it could be really interesting creating a private community with all the people that is interested in this field
I was hoping for a dall-e3 or dall-e4 update, but nope. Starting to feel as though OpenAI doesn't really seem to care much about image generating any more.
I realize that Bing's release with PR=16 is a dall-e update, but why hasn't openai mentioned it?
genesis is INSANE! as a physics student who made a lot of simulations, seeing such a general thing run that fast is incredible!
Oh Matt, your videos are amazing, it's delightful to watch, specially this one, i loved the 3d world generators
Open AI wins, hands down. Vio is amazing, passing the ARC benchmark marks a turning point in humanity.
but remember only google flight him in first class and many other youtuber so we got a winner :P so put all google in the front seat lol
My mind is blown... Again!!! Wow! So much to take in it's overwhelming _ Thanks Matt, you're the best Merry Christmas and Happy Hanukkah
Thanks Matt we love everything about your videos, please don’t change a thing! You’re covering everything that interests us.
16:28 The crazy eyes!!!
They look like they're in so much pain! One child's eye rolls a full 180!
I would be a big fan if Google flew me out to play with their new toys. (I'm shallow that way.) :) Thanks for the great work, Matt!
I’ve poured so much effort into this video. I’d love for you to watch it and share your thoughts!
O3 just solved coding. I am confused what to do with my life now.
You not alone😊
trust yourself, that's a opportunity
If it costs a million dollars to solve any problem, you're probably safe for a while.
@@user-on6uf6om7slol it costs less than a human
@@user-on6uf6om7s for a while (few months)
Yo, that join function on Notebook LM just tripped me out so much when the Deep dive AI's started responding to you! 😮
This year has been the biggest AI wave🎉🎉🎉❤
Looks like Google wins the "best toys" competition while Open AI wins with the "serious stuff" competition?
Sam is not having a good month. Google has been constantly dunking on OpenAI
>Sam is not having a good month
Except for ‘Orion o3’.
Amazing week!
Will you be talking about o3 on the live?
“Consumers are the real winners”, really…? Must be the Sales rep ‘agent’ speaking. Consumers are giving away their ‘intelligence’ and therefor part of their own identity… it must be Christmas for big tech.
Thanks for the always great content!
FYI - Imagen is pronounced like the word imagine.
No AI beats Anthropic because no other AI has an artifact window where tou can see your prompts as websites, assets, modules, etc. You see what you ask for if and thats a big if but if you can pet Claude the right way you can create OMG amazing stuff yall. I created a prompt (30) that generated and entire working WebAp builder. Works front to backend and in between.
I really hope Veo never ever bugs out crazy when released to the public
Thank you for the great video!
AI image generation won't be a solved problem until we achieve real character consistency. Also, we need to be able to describe more than one character/element of a scene without those characteristics bleeding over into the other characters. And of course there's always the censorship to deal with, but obviously the AI companies don't consider that a problem.
What a time to be alive ❤
What a time to be alive!!!
1:57 not really. objects with many parallel straight lines still look curved, random blobs in the background,... i would consider it almost solved, when you can really exactly describe how to change, refine image and it does exactly that, without any artifacts
I would buy plushie Matt... drop the merch!
No, we do NOT need a plush Matt toy. What are we going to do a with plush toy of Matt? Put it above a fireplace? Stick in the fridge. No, no, no.
I want a lawyer to interview a podcast about privacy
Matt, can we get some coverage on how Ai is used to harm people? Currently there is a debate on how Ai is used in HR and healthcare to automate processes that were done by humans. With devastating effects. I would personally appreciate a thought piece about it by an expert. Thank you for considering.
Hopefully you gonna make a prediction video for 2025 end of this year ^^ like last year
Hi Matt, would it be possible for you to use screenshots from a 4k screen so it doesn't get so pixelated?
7:50
i’d find the gwern interview and watch
and not only opt in
but get on the hyperstition train as fast as you can
The opt-out is only for third party. Google themselves will always be able train on your videos.
If Google so-called won with the latest round of updates, then why do I have no urge at all to cancel my sub with ChatGpt and move it to Google? Well after some thought the simple anser is most of what Google released is useless to me, they barely released anything practical. 2025 will most likely decide who I stay for the long run, for now ChatGpt has everything I want for now but that can change
I signed up for it a long time ago when you talked about it, but I never got access.
Well, OpenAI released O3. this tug of war is INSANE
Who made the endcap?
OpenAI dropped o3 to safety testers today, and they're kinda claiming AGI. That remains to be seen.
Im new to AI software, which video generator would you recommend to make videos as a beginner, invideo or VidAI?
Yeah, OpenAI's o3 won the week. And the year.
11:29 But it says right there also Other regions can WhatsApp message that number!
We are at a point where I do fear AI chatbots deployed with naferious intentions could pull these tools together to lay waste to the web with dead internet domains filled with an amalgamation of the best of these tools in use together.
Oh, I was so naive again... "Whisk is not available in your country yet".
Just here to say I'm loving sora. Is there anything better?
People are saying that VEO 2 is significantly better.
Google VEO
Veo 2 is better but not widely available, you need to join the waitlist and get invited.
@@sanderhoogeland9161 thank you
Kling 1.6 is S tier❤
Google should pay us if they want to use or videos to train thair AI models.
I’m sorry I love invideo used to have the most expensive package, but the 3.0 as amazing as it is doesn’t make sense $120 for 15 min of generations .😊
never mind you of course showed aistudio lol
So many gifts in this christmas..❤❤
I think we are going see Google Flex in 2025 more than OpenAI
Great microsoft prioritises mac too rather than microsoft lol.
I'm just waiting for the "spark"
Thank you sir
nvidea seem to be going the useful way with dedicated engines and now wow with its jetson nano at a reasonable price of $300 (cheaper in china) to run them on
how many hopper's is musk now shoving in his super complex ???? wow ! grok flash time !!!
once these things can ask me questions and then debate my answer with its own i will feel better to go and purchase one but as my personalised AI system I have built does that then why pay for a roadster when a bowl of cornflakes will do (in a bowl color of your own choosing)
Hmm, penguins in top hats huh.
i see that some good research is taking place with talking to whales and dolphins, maybe most of humans too :)
Gemini and OpenAI bots are battling each other in the comments
Gemini 2.2 flash is the best!!!
With release of o3, this seems so 48 hours ago. You need an AI to fix this 😂
Matt, at 2:30 you overlay the wrong URL for Google Labs' Whisk. It's at /fx/tools/, not /looks/.
Crazy week!
what are your thoughts on 03 model put out by openai..is it a gimmick scam or is it useful in mundane office tasks?
free is good to try as there are so many choices to make a decision on what suits you best .... like those awful rows of corn flakes in stores and all you want is something to eat, no toy, just some milk and no membership to i got coupons or anything as silly as some aborted penut butter sandwich
OpenAI o3 just crushed Gemini 2.0, sorry Google fanbois.
Not it hasn't. One is very good, cheap & available. The other is meant for researchers, is very expensive .. and not available!
@@coldlyanalytical1351 exactly 💯
@@coldlyanalytical1351very good ,cheap and available.
O3 is also good
O3 is expensive but because of compute requirements and nvidia is going to solve it with faster gpus and new chip architecture .
And it is going to be avalable for all .in fre monthes
I think both model have their pros and cons but i think open ai has an upper hand over here .
Anyways anyone is up the win will be of AI .
OK Dummy. Next week it all changes again.
O3 is super good.
But it's very expensive,
I have herd that for High compute it cost 1000$ per task.
Per task not monthly.
Low compute verson costs around 10$ per tasks, which is also expensive.
I am not trying to talk shit about o3.
It really is great.
But you still née cheaper model of not so high end taks,
Google Flash Thinking fits perfectly in that market.
May you live in interesting times. :-)
i would really like some new glasses that show me what is going on around me ..... ? does everybody wear dark glasses ?
The thumbnail of this video makes you look so strange 😅
"super super cool" is more like super super scary. there's going to be so much fakery
TH-cam won't take my thumbs up for you, thus suppressing your likes.
1. Does it make po1n
2. How much
My Pika generations have been running for hours and none have completed, lol. Can't really test if you can't see results... oy!
My Kling generations say the estimated waiting time is "longer than 3 hours", lol. Come on!
Nobody is talking about Grok😢 Elon Musk is probably bad for the algo.
Fax? No, telex.
Remember to 👎 these types of clips because if it spreads then you will end up having more competition while you're trying to reach financial freedom
You made me an app in which I am a virtual AI 3D avatar teacher in which you can ask me any question and tell me in very simple language in Hinglish language explanation with comedy and a good sense of humour and sarcasm, basically such as explanation like never bored with the teacher, enjoy with study. And there is such a screen on which you have to write and the question to be asked is related to the image or video. If the video is required to explain the question, then the video or else the image means which is correct according to the question. You can see the image or video and The image or video is not copyrighted and for commercial use, why do I have to upload this app on Play Store and add movement in 3D avatar and also add lips movements feel realistic and like something explaining someone movement then you can ask me if you need any files., you need the link or file of the 3D avatar and environment and API key and I provided any more than asked me. just take this app and make it professional please make tutorial for this 😢
With all due respect you very wrong about image generation being able to generate anything. It still has many limitation, try get it to have someone throwing a ball, doing a high kick. Hitting something with a weapon. A train or bus interior that doesn't look like total mess... It still cant do any of those. If you think it can do anything, you need more imagination. Basic portraits and landscapes it can do easily, that's a long way from 'everything'
Check strrawberry...
o3
we have to watch out for the early scams when the likes of microsoft came out with a 'new' version but only changed the menu icons or buttons and merged all their products to have a bit of each in each other (eugh!) not a new skin more like a slap of slime
Until veo 2 is available to all it is not the best out there.
👍
Google 2.0 isn't as good as Claude when it comes to writing code.
🤩👍👏👏
Chatbot arena is an ABSOLUTE JOKE...gemini is so extremely far behing from chat-gpt...and i would really like some competition there. but for now...its dimensions how smart they look.
Gemini 2.0 is something like gpt 2-3.
16:32 horrifying!