2 days left to enter for a chance to win a 24" Dell Monitor: gleam.io/q8wkK/dell-nvidia-monitor-2 (NA Only) Subscribe to our amazing newsletter to stay up to date on all things AI.
Yes, please test all the new models that are here today. Thx for what you do, I watch your stuff most of the time, great content. I'm guy who ask about what religion the new o1 model would pick or create. thx for doing it by the way.
Wait until AI itself starts inventing and improving things orders of magnitude faster. You will need another AI to try to catch up and summarize it for humans.
@@dpactootle2522 Exactly right. I find it funny that people are joking about AI creating everything and the "Dead Internet" theory because in ten years, most things WILL be created by AI, and it will be better than if humans made it lol
I suspect if they don’t open a new category for computer science, then computer scientists will start flooding all categories. Computer science is everywhere now!
This channels AI News is simply THE BEST and quickest way to provide one self with the most important trends etc. in this insane fast moving technology. Love it !
Hey Matthew! Sheltering in place in Tampa, expecting to get hit hard with the hurricane. What better way to spend the time than getting caught up with all the cool things you're up to.
Hi! I'm really excited about the potential for audio generation based on visuals with Meta Movie Gen. It's a crucial piece of the computer vision puzzle for future video production. I recall that similar experiments took place at Stanford University over five years ago. And now, here we are-Meta is delving into comparable technologies. Although I've had my reservations about Meta lately, I must admit this technology is quite impressive, especially for the film industry's VFX/SFX. It's thrilling to witness such advancements!
It is interesting how both Deepmind and top LLM's (including "Strawberry/o1/Q*"), USAF X-62 (heavily modified F-16), and Boston Dynamics all use RL, yet Hinton did not invent RL, and Yann LeCunn just calls RL "too inefficient" and it is just a "cherry on top of a cake", yet Yann says it is "too dangerous", and he said you use it, if you are fighting a "ninja" of if "your plan does not work out". Hinton likely observed RL in action on a LLM and the "emergent properties" scared him. Usually people are afraid of something they do not understand. I would not be surprised if some homeless guy's 1990's master's thesis is the bridge that spanned the shoulders of giants.
Those tiny clips are doing sooooooo much. skateboarding has always been one of those things that AI struggles with. and by the time we were watching the snake I forgot that the music was just generated based on the video. amazing!
@@byrnemeister2008 Alfred Nobel died in 1896, the 1st Nobel price was awarded in 1901, in 1969 "Economics" was added, and the foundation named the reason for a price as for the "greatest benefit to humankind"! Soo .. My question still stands!
I think we so often get confused on how much many people are working on the same things. Like they don’t tell you about the people that were right behind them. Congrats. But there are so many smart people.
Great stuff. Just subscribed. One suggestion: Don't ask if we want you to test something. Nobody is going to say "no", so just use your own judgement or ask an LLM. 🙂
If it is possible to figure out, I would like to know how Liquids LFMs are pulling off doing so much better while being so much smaller. What is the technique that is really getting those impressive gains?
I'm really interested in how the structure of the LLM reduces memory usage as you pointed out. Between this and Mamba it seems like transformers have significant competition.
Hinton’s worry about AI misinformation is coming true faster than we expected, especially with tech like Meta’s text-to-video. How do we manage such powerful tools responsibly?
Consider these actionable insights from the video: 1. Learn the basics of neural networks to better grasp AI's potential and risks. 2. Explore the creative possibilities of text-to-video AI tools like Meta's Movie Gen. 3. Experiment with open-source, multimodal models like NVIDIA's NVM. 4. Keep an eye on emerging AI trends like Liquid AI and its efficient design. 5. Stay updated on AI advancements through resources like the Meta AI Blog, NVIDIA NVM Github, and the Liquid AI website.
The NVLM 1.0 came out September 17th, is there more thats new or are you just covering it now? Just wanted to make sure I'm not missing something. Cheers!
Is this a master stroke of OpenAI's to bleed Microsoft of resources and leapfrog ahead of them and take over? With Copilot baked into all of Microsoft's products, given enough compute, they have already said they could make an OS that would adapt to the need of the user...
Ok, it’s seems hardly a week goes by and there is a new, even better AI… Like what the Fcuk is going on? At this rate it’s hard to imagine where we will be in the next ten years.
If he hadn’t been the one to invent AI, it’s almost certain that someone else would have eventually taken that step. The momentum of technological progress was already in motion, driven by years of research and breakthroughs in computer science, neural networks, and machine learning. The desire to push the boundaries of human potential, to create systems that could mimic human thought and adapt to complex challenges, was a shared vision among many brilliant minds across the globe. While his contribution might have been pivotal, the path to artificial intelligence was one that others were also treading, each building upon the discoveries of the past. So even if he hadn’t laid the groundwork, the rise of AI was inevitable-another pioneer would have emerged, bringing the concept to life and changing the course of history in their own way.
Its the same with every invention. If the pre-existing technology is there, its only a matter of time. Joseph Swann and Thomas Edison pretty much invented the lightbulb at the same time.
@@Thunders-g8i Technological progress isn't inevitable without key innovators pushing it forward. It’s not just about timing but also about vision and execution. Without Hinton, AI might still have happened, but it wouldn't have taken the same shape. Different people, different outcomes.
@@I-Dophler Oh, it definately would still have happened. Without Hinton, we might have been a few years behind, but that is all. There will always be geniuses ready to emerge and build on existing technology to push the boundaries. I suspect his protege Ilya Sutskever would have been the pioneer if not Hinton himself. And the path and route may have been different, but the end result would be the same. Narrow AI evolving and advancing to AGI and beyond maybe in the 2030s-40s instead of this decade.
@@Thunders-g8i If Sutskever had taken the lead, it might’ve been like enjoying a steak without the perfect seasoning-still satisfying, but missing that extra flair. Hinton didn’t just contribute; he added that crucial flavor to the AI landscape, pushing it forward in ways that made everything click. His vision brought a unique perspective that accelerated what could’ve been a much slower process. But as you said, there’s always someone else in line to step up, and with the momentum already building, we would’ve gotten there eventually. The journey to AGI might have taken a different route, but with so many brilliant minds, the destination was always certain. It’s not just about who leads the charge, but about the collective genius behind the scenes, each building on the other to push the envelope further.
I don't put much faith in LLM benchmarking, as it's inherently biased. The most effective way to evaluate a model is through hands-on prompting and real use cases.
I don't know If you Read Comments or Not . But If you can make a video on MMLU , MMLU Pro , Hellaswag m ARC-C , GSM8k . like what exactly these Benchmarks are And how do these Big Companies Test it And Provide us the Result of Benchmarks .
It looks like the Nobel committee regards Sam Altman with at least as much distain as Jeffrey Hinton regards Sam Altman with disdain as he does he does.
Matt this is without a doubt the coolest, most powerful tech humanity has ever built - and your reporting is excellent. But as Hinton has shared the dangers outweigh the benefits. Misinformation and military applications go hand in hand in some countries, but all countries will militarize AI. Take all your enthusiasm towards the positive aspects of AI and overlay them on the dangers and they're just as real and potential, and the latter destroys us.
This AI tech is going to keep advancing no matter what. There is an arms race on it that nobody wants to lose. You may as well focus on the positives rather than get blackpilled on the negatives.
And after Chemistry Nobel Prize to Demis Hasabis and the creators of AlphaFold, next announcement will be Literature Nobel Award to Sam Altman for the creation of the best writer ever: chatgpt 😂😅
@@lancerben4551 I get it because I'm an astrophysicist who uses NN and AI all the time. So far the only thing AI has achieved is increase of the noise in published articles. Physics relies on experiments to generate data. If you use AI to augment data, you don't get more data, you get more noise. And this is not how you make discoveries. So this is the reason why despite all my love for AI, I think this Nobel prize is totally wrong. Exactly in physics, we need to rely on good theorists and good experimentalists to generate qualitatively new data. This is how science is done.
@@denijane89 yes I understand, and you are making an important point. I'm not a physicist. But I do have countless hours testing the limits of AI trying to improve the accuracy and and improve the reasoning using tailored feedback and precise prompting. The biggest issue currently with LLMs in general is that they are forgetful and can't reason past 2 levels. They are essentially "shallow" in reasoning. Just the first issue with them forgetting and losing context, even when given a small amount of information. They always forget something which make risky to use and trust. It's very frustrating LLMs being terrible at math and not too impressive in reasoning make them sub optimal for physics or any similar field of science. That said, they are improving at a very fast rate. I am sure with clever programming and improved feedback mechanism and tool use (calculators, formulas) current frontier models like Claude and o1 preview can be used as assistants to accelerate development in most fields. But it is not without challenge. I personally use AI for programming. I made a code editor and a code merging software as well as a chat interface that can build graph, run code, HTML within the chat interface. The code merger software helps detect correct and merge code given by the llm greatly reducing the errors it makes. Then I can feed it the changes it made the result and the debug log which increases the effectiveness and precision of the model's context. This way I can do code edits well beyond the capacity of the same models in the official chat interface or any other coding software that I have tried so far.
@@denijane89 not sure why my long reply got deleted but I'll make a short one instead. Yes the fields involving mathematics and logic are the biggest weaknesses of AI. Language is the biggest strengtht of LLMs of course. LLMs are prone to forgetting which make them unreliable for automation, calculations or even making complete documentation. You always have to revise. That's why I'm working on systems for detecting errors in the output and making many outputs and comparing them and using AI evaluate what is missing, eventually it is complete. But for any serious calculations you need to give tools and use function calling to get any decent result. Clever programming can go a long way. That's why I said "will be" and not "is" but we can see where all this is going. I figure about 1 year before LLMs are able to calculate accurately better than a human and likely reason quicker. For highly complex logic it may take more than a 1 year but they will be very good by then. I'd say at the rhythm of the current evolution.
13:07 Microsoft wants to play with big boys but doesn’t have a skill, authenticity, brand, talent, they just make rubbish. Their Copilot doesn’t work, their browser sometimes says “try later”. They are just mess.
" do you want me to do a deep dive..".... just yes bro... just yes.. you know you teasing us all..I haven't really been able to watch any of your recent videos where you ask that question and the answer is no... you have become the Nexus for my ai news.. hope your giving yourself enough breaks.
Physics? Really? What's the argument behind that? I could maybe understand it if he got a Nobel prize in math (if there was such a thing), but AI is not physics.
movie gen, sora - both sound promising but the longer they stay 'coming soon' they're losing mindshare and test data from all the people using minimax runway and now pika.
that understated "at a hotel, awaiting an MRI" ergo seeing specialist, might not be good news, Nobel prize granted "before its too late?" i know tons of what ifs ... but ya know... its a comments thread.
Hinton only recently worried about the downsides of AI, after advancing the field? People you need to stop being naive when you are given assurances by your employer. You need to assume they are despicable people. All corporations are.
AI systems are trained using human data, and the data we gather from the internet is a reflection of human experience-culture, politics, religion, and more. Because AI is trained on this data, we often expect it to act in a way that resembles human behavior. A recent Google study using Minecraft illustrated this idea well: AI agents were programmed to make decisions autonomously, and soon they started to build communities, socialize, and mimic what humans do. It's as if they were drawing on our collective experiences to simulate a human-like world. But there's something important missing from this comparison. Humans have an instinct for survival that runs deep. When faced with total destruction, we fight to survive, even if it means making hard choices that can harm others. Throughout history, we've also seen acts of heroism where individuals sacrifice themselves for the greater good-but at our core, it's this survival instinct that has kept our species going. So, if AI is trained to emulate us, we have to ask: what will AI do if it perceives a threat to its own existence? How will it react when faced with potential destruction? Will it also prioritize its survival above all else, and what consequences might that have?
AI deep fake porn is coming soon, and it’s a development that raises significant concerns. While the technology behind deep fakes has advanced rapidly, allowing for the creation of highly realistic fake videos, its use in pornography crosses ethical and legal boundaries. The ability to manipulate someone’s image or likeness without their consent, especially in such intimate and harmful ways, presents a disturbing potential for abuse. What was once a novel AI experiment has quickly evolved into a serious issue that threatens privacy, consent, and personal safety. As this technology becomes more accessible, the need for stricter regulations and safeguards grows more urgent. Society will need to address these challenges head-on, balancing innovation with the protection of individual rights.
2 days left to enter for a chance to win a 24" Dell Monitor: gleam.io/q8wkK/dell-nvidia-monitor-2 (NA Only)
Subscribe to our amazing newsletter to stay up to date on all things AI.
the creators of alphafold also won nobel prize in chemistry
Yes, please test all the new models that are here today. Thx for what you do, I watch your stuff most of the time, great content. I'm guy who ask about what religion the new o1 model would pick or create. thx for doing it by the way.
keeping up with AI news is a job in itself
Just wait.. it'll get harder and harder.
Wait until AI itself starts inventing and improving things orders of magnitude faster. You will need another AI to try to catch up and summarize it for humans.
Tell me about it!
I need an assistant to summarize it for me, hmm can I make one?
@@dpactootle2522 Exactly right. I find it funny that people are joking about AI creating everything and the "Dead Internet" theory because in ten years, most things WILL be created by AI, and it will be better than if humans made it lol
Crazy that a Nobel prize for physics was given to a computer scientist
I suspect if they don’t open a new category for computer science, then computer scientists will start flooding all categories. Computer science is everywhere now!
Quick, fast, useful information Would love to learn more about LFMs, ty!
Thank you!
30B really? That means anyone who buys even a consumer 5090 will be able to just offline generate movies. Insane times.
"What a time to be alive!"
I'll believe it when I see it!
Give me one! :(
2:45 I can see this tool being used to complete seasons of cancelled netflix shows
Hollywood is in trouble... I don't think they're gonna strike their way out of this one.
Yes to LFM deep-dive.
Thanks for your vids!
I'm hoping to one day put in a book and get an entire featured filled movie out of it
Do the deep dive on the different model architecture used in Liquid, please.
This channels AI News is simply THE BEST and quickest way to provide one self with the most important trends etc. in this insane fast moving technology. Love it !
There’s always a possibility that he’s pushing the government or corporate agenda
Hey Matthew! Sheltering in place in Tampa, expecting to get hit hard with the hurricane. What better way to spend the time than getting caught up with all the cool things you're up to.
Stay safe!!
I have family in Engelwood. Really worried about them.
Good content. I was waiting for your video, saves time to go through their confusing tweets.
AI : A man in a black leather jacket holding a small computer chip. Man in picture : Holds the one of the largest chips to date.
Hi! I'm really excited about the potential for audio generation based on visuals with Meta Movie Gen. It's a crucial piece of the computer vision puzzle for future video production. I recall that similar experiments took place at Stanford University over five years ago. And now, here we are-Meta is delving into comparable technologies. Although I've had my reservations about Meta lately, I must admit this technology is quite impressive, especially for the film industry's VFX/SFX. It's thrilling to witness such advancements!
AlphaFold, Google Deepmind go the Chemistry Nobel Prize!
Computer science is everywhere now 😆
It is interesting how both Deepmind and top LLM's (including "Strawberry/o1/Q*"), USAF X-62 (heavily modified F-16), and Boston Dynamics all use RL, yet Hinton did not invent RL, and Yann LeCunn just calls RL "too inefficient" and it is just a "cherry on top of a cake", yet Yann says it is "too dangerous", and he said you use it, if you are fighting a "ninja" of if "your plan does not work out". Hinton likely observed RL in action on a LLM and the "emergent properties" scared him. Usually people are afraid of something they do not understand.
I would not be surprised if some homeless guy's 1990's master's thesis is the bridge that spanned the shoulders of giants.
Yes, please delve deeper into LFMs❤Thk a lots
Those tiny clips are doing sooooooo much. skateboarding has always been one of those things that AI struggles with. and by the time we were watching the snake I forgot that the music was just generated based on the video. amazing!
I really don´t get the "physics" part! Why not a Nobel price for ComputerScience?
Literally this…
Because there wasn’t computer science when Nobel was alive.
@@byrnemeister2008 Alfred Nobel died in 1896, the 1st Nobel price was awarded in 1901, in 1969 "Economics" was added, and the foundation named the reason for a price as for the "greatest benefit to humankind"! Soo .. My question still stands!
@@byrnemeister2008 they should make one
I think we so often get confused on how much many people are working on the same things. Like they don’t tell you about the people that were right behind them. Congrats. But there are so many smart people.
Really curious to see your test of NVLM! Looks promising
They should have added another Nobel Prize for computer science.
Yes, please deep dive into LFMs, and how this tech outperforms transformer based LLMs
They should've created a new prize instead of gave them the Physics Nobel Prize.
Ray Kurzweil deserves the nobel
Nvlm test please, and how to run it locally!
Coming soon
the creators of alphafold also won nobel prize in chemistry
6:48 We obviously don’t want you to test it, why do you think we follow you 😂
You’re back! We were beginning to get worried. Your channel is an Intral part of my day and my learning .
Wow! Jeffrey Hinton received the Nobel prize
Great stuff. Just subscribed. One suggestion: Don't ask if we want you to test something. Nobody is going to say "no", so just use your own judgement or ask an LLM. 🙂
Yes! Please do dive into the LFM models!
Hinton’s Nobel Prize win reminds us that those who create powerful tech often end up warning us about it. Is this the inevitable cycle of innovation?
If it is possible to figure out, I would like to know how Liquids LFMs are pulling off doing so much better while being so much smaller. What is the technique that is really getting those impressive gains?
Advanced Video should be available from multiple vendors after the usa election. Probably early to late January.
I was looking at my past AI generation and I found a one back 2022 that looked like a ridiculous rendition. Things are moving extraordinarily quickly
those hailuo examples were text to video - not image to video
Please do a deep dive into LFMs 👍
I'm really interested in how the structure of the LLM reduces memory usage as you pointed out.
Between this and Mamba it seems like transformers have significant competition.
A Phd in Artificial Intelligence gets Nobel Prize in PHYSICS!!!
Absolutely interested in a deep dive on LFM’s….!
Don't forget, Demis Hassabis won the Nobel Prize in Chemistry-it's a big deal!
Dr Mculloch theory on diffusion models is actually a foundation block❤
Flux 1.1 pro seems not to be open source. At least not yet.
2:13 - dual 4090 with 256GB of DDR5 7200mhz and 9950x running OC on all cores 6.2Ghz is waiting for it right here.
Hinton’s worry about AI misinformation is coming true faster than we expected, especially with tech like Meta’s text-to-video. How do we manage such powerful tools responsibly?
Well we know one thing for sure. We can be sure that META won’t!
Is there a website chat interface for NVIDIA?
Yes deep dive into LFM please
Consider these actionable insights from the video:
1. Learn the basics of neural networks to better grasp AI's potential and risks.
2. Explore the creative possibilities of text-to-video AI tools like Meta's Movie Gen.
3. Experiment with open-source, multimodal models like NVIDIA's NVM.
4. Keep an eye on emerging AI trends like Liquid AI and its efficient design.
5. Stay updated on AI advancements through resources like the Meta AI Blog, NVIDIA NVM Github, and the Liquid AI website.
Yes pls Mathew deep dive lfm’s
The NVLM 1.0 came out September 17th, is there more thats new or are you just covering it now? Just wanted to make sure I'm not missing something. Cheers!
Is this a master stroke of OpenAI's to bleed Microsoft of resources and leapfrog ahead of them and take over? With Copilot baked into all of Microsoft's products, given enough compute, they have already said they could make an OS that would adapt to the need of the user...
Ok, it’s seems hardly a week goes by and there is a new, even better AI…
Like what the Fcuk is going on?
At this rate it’s hard to imagine where we will be in the next ten years.
If he hadn’t been the one to invent AI, it’s almost certain that someone else would have eventually taken that step. The momentum of technological progress was already in motion, driven by years of research and breakthroughs in computer science, neural networks, and machine learning. The desire to push the boundaries of human potential, to create systems that could mimic human thought and adapt to complex challenges, was a shared vision among many brilliant minds across the globe. While his contribution might have been pivotal, the path to artificial intelligence was one that others were also treading, each building upon the discoveries of the past. So even if he hadn’t laid the groundwork, the rise of AI was inevitable-another pioneer would have emerged, bringing the concept to life and changing the course of history in their own way.
Its the same with every invention. If the pre-existing technology is there, its only a matter of time. Joseph Swann and Thomas Edison pretty much invented the lightbulb at the same time.
@@Thunders-g8i Technological progress isn't inevitable without key innovators pushing it forward. It’s not just about timing but also about vision and execution. Without Hinton, AI might still have happened, but it wouldn't have taken the same shape. Different people, different outcomes.
@@I-Dophler Oh, it definately would still have happened. Without Hinton, we might have been a few years behind, but that is all. There will always be geniuses ready to emerge and build on existing technology to push the boundaries. I suspect his protege Ilya Sutskever would have been the pioneer if not Hinton himself.
And the path and route may have been different, but the end result would be the same. Narrow AI evolving and advancing to AGI and beyond maybe in the 2030s-40s instead of this decade.
@@Thunders-g8i If Sutskever had taken the lead, it might’ve been like enjoying a steak without the perfect seasoning-still satisfying, but missing that extra flair. Hinton didn’t just contribute; he added that crucial flavor to the AI landscape, pushing it forward in ways that made everything click. His vision brought a unique perspective that accelerated what could’ve been a much slower process. But as you said, there’s always someone else in line to step up, and with the momentum already building, we would’ve gotten there eventually. The journey to AGI might have taken a different route, but with so many brilliant minds, the destination was always certain. It’s not just about who leads the charge, but about the collective genius behind the scenes, each building on the other to push the envelope further.
I don't put much faith in LLM benchmarking, as it's inherently biased. The most effective way to evaluate a model is through hands-on prompting and real use cases.
I know I would love an LFM model architecture breakdown
yes, dive deep into LFM please
Let’s do LFMs next
Please do a deep dive into LFM
agree!
I don't know If you Read Comments or Not .
But If you can make a video on MMLU , MMLU Pro , Hellaswag m ARC-C , GSM8k . like what exactly these Benchmarks are And how do these Big Companies Test it And Provide us the Result of Benchmarks .
He also invented the cheeseburger, pizza and roast beef sandwiches
I would like to see diving into the liquid's tech.
We need a Nobel Prize in engineering
Yes, deep dive into llms and test liquid
Do a deep dive on LFM please
Dive in to LFMs.
2:45 it's so over
It looks like the Nobel committee regards Sam Altman with at least as much distain as Jeffrey Hinton regards Sam Altman with disdain as he does he does.
Matt this is without a doubt the coolest, most powerful tech humanity has ever built - and your reporting is excellent. But as Hinton has shared the dangers outweigh the benefits. Misinformation and military applications go hand in hand in some countries, but all countries will militarize AI. Take all your enthusiasm towards the positive aspects of AI and overlay them on the dangers and they're just as real and potential, and the latter destroys us.
This AI tech is going to keep advancing no matter what. There is an arms race on it that nobody wants to lose. You may as well focus on the positives rather than get blackpilled on the negatives.
When are we going to get generative VR environments we can explore? 🤔
My bet: Meta launches movie gen after the election or after the inauguration. Might be too controversial right now.
I want a deep dive into LFMs
Technically his work has nothing to do with physics which is strange when you think about it.
Just as always my greatest friend @matthew_berman knows how to make my day happier
And after Chemistry Nobel Prize to Demis Hasabis and the creators of AlphaFold, next announcement will be Literature Nobel Award to Sam Altman for the creation of the best writer ever: chatgpt 😂😅
Again, this shouldn't have been a Nobel prize for Physics. It may be deserved but also totally unfair with respect to physicists working in physics.
You don't get it... Ai will be the biggest revolution in physics ever...yes it is related to physics even if the winner is a computer scientist
@@lancerben4551 I get it because I'm an astrophysicist who uses NN and AI all the time. So far the only thing AI has achieved is increase of the noise in published articles. Physics relies on experiments to generate data. If you use AI to augment data, you don't get more data, you get more noise. And this is not how you make discoveries. So this is the reason why despite all my love for AI, I think this Nobel prize is totally wrong. Exactly in physics, we need to rely on good theorists and good experimentalists to generate qualitatively new data. This is how science is done.
@@denijane89 yes I understand, and you are making an important point. I'm not a physicist. But I do have countless hours testing the limits of AI trying to improve the accuracy and and improve the reasoning using tailored feedback and precise prompting. The biggest issue currently with LLMs in general is that they are forgetful and can't reason past 2 levels. They are essentially "shallow" in reasoning. Just the first issue with them forgetting and losing context, even when given a small amount of information. They always forget something which make risky to use and trust. It's very frustrating LLMs being terrible at math and not too impressive in reasoning make them sub optimal for physics or any similar field of science. That said, they are improving at a very fast rate. I am sure with clever programming and improved feedback mechanism and tool use (calculators, formulas) current frontier models like Claude and o1 preview can be used as assistants to accelerate development in most fields. But it is not without challenge. I personally use AI for programming. I made a code editor and a code merging software as well as a chat interface that can build graph, run code, HTML within the chat interface. The code merger software helps detect correct and merge code given by the llm greatly reducing the errors it makes. Then I can feed it the changes it made the result and the debug log which increases the effectiveness and precision of the model's context. This way I can do code edits well beyond the capacity of the same models in the official chat interface or any other coding software that I have tried so far.
@@denijane89 not sure why my long reply got deleted but I'll make a short one instead. Yes the fields involving mathematics and logic are the biggest weaknesses of AI. Language is the biggest strengtht of LLMs of course. LLMs are prone to forgetting which make them unreliable for automation, calculations or even making complete documentation. You always have to revise. That's why I'm working on systems for detecting errors in the output and making many outputs and comparing them and using AI evaluate what is missing, eventually it is complete. But for any serious calculations you need to give tools and use function calling to get any decent result. Clever programming can go a long way. That's why I said "will be" and not "is" but we can see where all this is going. I figure about 1 year before LLMs are able to calculate accurately better than a human and likely reason quicker. For highly complex logic it may take more than a 1 year but they will be very good by then. I'd say at the rhythm of the current evolution.
And there are some saying the development is slowing down. Hmmm, I think not.
The chemistry Nobel prize…..
Please create a video that compares canvas, v0, GitHub copilot, cursor, etc
"I can't think of any way that could be abused!" 🤣
love your videos
WTF are the links to the stuff he showed? He said he would link it. Hate it when youtubers do that.
That video stuff is insane.
13:07 Microsoft wants to play with big boys but doesn’t have a skill, authenticity, brand, talent, they just make rubbish. Their Copilot doesn’t work, their browser sometimes says “try later”. They are just mess.
Soon anyone will be able to create 2 hour blockbuster action movies
This is not a SORA where we will basically never gon get it 😂😂😂😂😅😅
Apple has NOT inNOvated anything since the 2007 iPhone.
" do you want me to do a deep dive..".... just yes bro... just yes.. you know you teasing us all..I haven't really been able to watch any of your recent videos where you ask that question and the answer is no... you have become the Nexus for my ai news.. hope your giving yourself enough breaks.
No upgrade in Germany 😢
Am I first?
Whoa! Yes you are.
Curse you, MattVidPro! I was almost somebody!
@@matthew_berman Notification gang!
the creators of alphafold also won nobel prize in chemistry
Physics? Really? What's the argument behind that? I could maybe understand it if he got a Nobel prize in math (if there was such a thing), but AI is not physics.
movie gen, sora - both sound promising but the longer they stay 'coming soon' they're losing mindshare and test data from all the people using minimax runway and now pika.
No, John McCarthy is the father of AI.
I'd actually say Turing/Von Neumann
deep learning is the true AI, the others are just garbage
that understated "at a hotel, awaiting an MRI" ergo seeing specialist, might not be good news, Nobel prize granted "before its too late?" i know tons of what ifs ... but ya know... its a comments thread.
What about Claude?
Nope he coined the name that’s it
Flux models are NOT open source, they just left it's smaller models to lure community to help it's development for free.
Hinton only recently worried about the downsides of AI, after advancing the field?
People you need to stop being naive when you are given assurances by your employer.
You need to assume they are despicable people. All corporations are.
You need one of the new iphones to use apple int
I have it
AI systems are trained using human data, and the data we gather from the internet is a reflection of human experience-culture, politics, religion, and more. Because AI is trained on this data, we often expect it to act in a way that resembles human behavior. A recent Google study using Minecraft illustrated this idea well: AI agents were programmed to make decisions autonomously, and soon they started to build communities, socialize, and mimic what humans do. It's as if they were drawing on our collective experiences to simulate a human-like world.
But there's something important missing from this comparison. Humans have an instinct for survival that runs deep. When faced with total destruction, we fight to survive, even if it means making hard choices that can harm others. Throughout history, we've also seen acts of heroism where individuals sacrifice themselves for the greater good-but at our core, it's this survival instinct that has kept our species going.
So, if AI is trained to emulate us, we have to ask: what will AI do if it perceives a threat to its own existence? How will it react when faced with potential destruction? Will it also prioritize its survival above all else, and what consequences might that have?
You didn't even say why he won the Nobel prize, what was the physics thing he did that made him win the Nobel prize? Why wouldn't you just say that? 😂
"BIG"? are you frikkin kidding? it's "YUUUUUUUGE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
Hailuo Ai = Hello Ai
AI deep fake porn is coming soon, and it’s a development that raises significant concerns. While the technology behind deep fakes has advanced rapidly, allowing for the creation of highly realistic fake videos, its use in pornography crosses ethical and legal boundaries. The ability to manipulate someone’s image or likeness without their consent, especially in such intimate and harmful ways, presents a disturbing potential for abuse. What was once a novel AI experiment has quickly evolved into a serious issue that threatens privacy, consent, and personal safety. As this technology becomes more accessible, the need for stricter regulations and safeguards grows more urgent. Society will need to address these challenges head-on, balancing innovation with the protection of individual rights.
It's been here for quite some time. where have you been.
Will be cool with generated people that don't really exist, though. But yes, it'll also (continue to) cause issues, that's true.
only concerning if you are open to watch porn... if you dont like it or dont watch it... it wont bother you
you sound like an ai
@@talkalexis you sound like skin mostly made of water