Meta STRIKES AGAIN! New AI DEVICE, Microsofts NEW Model PHI-3, Adobe Firefly 3 STUNS! And More
HTML-код
- Опубликовано: 22 апр 2024
- How To Not Be Replaced By AGI • Life After AGI How To ...
Stay Up To Date With AI Job Market - / @theaigrideconomics
AI Tutorials - / @theaigridtutorials
🐤 Follow Me on Twitter / theaigrid
🌐 Checkout My website - theaigrid.com/
Links From Todays Video:
Links From Todays Video:
build.nvidia.com/microsoft/ph...
/ 1782446072452780347
/ 1
news.microsoft.com/source/fea...
/ 1782594659761389655
arxiv.org/pdf/2404.14219.pdf
/ 1782493076373885336
/ 1782828311766265898
/ 1782872540463194509
/ 1782559436055433237
/ 1
www.adobe.com/products/firefl...
arxiv.org/pdf/2404.13208.pdf
www.rabbit.tech/live-unboxing
• rabbit pickup party NY...
about. news/2024/04/new...
/ 1782889456657224119
www.theverge.com/2024/4/23/24...
Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.
Was there anything i missed?
(For Business Enquiries) contact@theaigrid.com
#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience 
Наука
No, we are doing AGI for everyone. If it's not that then its the worst case scenario where one mega corp runs everything.
This can *only* become your worst case. Think about it ...
A select few already have unfettered access, while insisting on sandboxing the rest of us. These individuals will eventually wield God-like powers, while continually denying everyone else a chance to even understand what's happening, let alone compete.
@@ZappyOh I totally disagree with this guy on open sourcing large LLM’s. I think they should be open sourced. I do not expect future AI systems to continue to be disembodied. I expect a future where models will interact with other models, humans, other biological creatures, and directly with their environment. I think many AI focused companies are focusing on the wrong thing. The intelligence they need and want is one that sufficiently models the world and can make decisions to further its goals. What we don’t need is a slightly better chat bot. To gate keep intelligence is to gate keep ability aka power. We already have in society a method of changing the behavior of people, those things are called laws. If a system is goal oriented, laws would influence AI systems just as much as they influence people. Especially if we have AI systems enforcing laws. The systems of power we have in the world are already hardened to withstand people or agents trying to topple them.
The key to the stability of democracy is power in numbers. No one person or group or even government can have too much power if we want democracy to stand. And no you can't just give all of the power to the government, if you do they would have no reason to listen to the people. Right now the government has to listen to the people just as the people have to listen to the government. There is a balance of power. We have many examples of what happens when the people aren't necessary to a country, when the government has too much power.
The main threat from AI job loss isn't a loss of income or livelihood, it's a loss of the people's power. If you no longer NEED to work that means the government no longer NEEDs you. All of the shalkle’s people have on the government's power functioning because they need to keep us convinced to work to even have a working government goes away if people aren't the ones running it. A ruler who doesn't NEED other people to keep their power doesn't NEED to listen to their people at all.
Making a bio weapon seems bad and nobody wants it to happen. But you know what it is, its power. If everyone has power, nobody has power. And that's a good thing. We NEED to be dangerous to the government or we lose all of our power and rights, our rights aren't given they were taken.
Open source worst case is terrorists making viruses, or giving our tech to China for free.
@@Jayc5001 I kinda agree with you. Power needs to be distributed evenly for society to function reasonably well for the majority. However, I clearly see power slipping upwards.
We have a small window of opportunity, right now, to "do something", but most people have no clue, and the top guys are working very hard, to get past that window as fast as possible.
Well said @Jayc5001
My problem with AI:
A select few have unfettered access, while insisting on sandboxing the rest of us.
Welcome to the world, it has always been like this and will always be.
@@WilliamsDarkoh But AI is nothing like anything prior. Think about that for 5 minutes ...
The ones with access will eventually wield God-like powers, while denying everyone else a chance to even understand what's happening, let alone compete.
Probably has to be that way since there are people who would use unrestricted AI to build chemical and biological weapons that would otherwise be too complex for them to make.
@@ZappyOh it will be just an amplification of what's currently happening, if you'd knew how much shit the secret agencies like the CIA or some random lab in Russia and China were hiding from us you'd understand where I'm coming from. I'm 100% sure AGI and possibly ASI has already been achieved. The monster is out there, and hasn't gone crazy yet, what's worrying me is not none of the predictions we have, but what's coming after that. Cuz you just like me and 99.999999% of us have no idea of what's going on out there. The reality in which we're living in could be nothing but an experiment to see how a civilization with unrestrained AI could develop to see whether they should stop the development in their base reality or not. Perhaps we're going through the same process right now by recreating another simulation and it could truly be just an endless loop.... Anyways, as said before, this changes nothing, power in the hands of a few selected ones, even though not as democratic I think it's far better than everyone having access to it. Put yourself and your desires totally out of the equation and you'll clearly see why.
The world is full of idiots that could do far more damage than what someone who knows how to handle or at least had some level of experience or knowledge on how to handle something like that. Like who would you trust more with a 44 calibre semi-automatic, a veteran soldier or a 5 year old toddler with dawn syndrome? If I have to bet on my odds, I'll bet all day every day that the ones in true control(not the idiots we see elected every couple of years) are more capable than most of us and would take on average decisions far better than us. Even then, this is a short sighted view, past ASI, do you really think we people, or any person will have any control over these machines? What should scare you is not how the secret world leader uses AI to manipulate society or could lead to mass destruction, but how an AI that managed to free itself from ani sort of human constraint and became dozens if not hundreds of times smarter than one would behave.it's not going to be your average Putin Biden or some secret agency who will be in charge in that near future, it's the AI who's truly going to be in control, so if you truly are worrying about sum worry about its future decision rather than the people's decisions. (Sorry for my English, I've only been learning the language for 2 years)
@@TheThundertaker Distrusting your fellow man, while trusting the golden few at the top. Interesting.
Do you know Stockholm Syndrome?
Yep. Ban AI. Guns. Bank accounts. Food. The amount of time one can spend outdoors. Travel. Health care. Elections. Thought. Yep open source nothing. Let our rulers rule us
Yep is thisnt about AI its about power. They don't want the people to have more power. But if we keep on the Closed Source path the govenment and the few will have power and they wont need us anymore. Right now they lissen because they need to but when we no longer have the power that our labour gives us, why would they lissen to us?
@@swift-ft7qj exactly ding ding you win. I’m really excited about the young people now waking up to the lies. They always complain about older people, but they understand. The older people are on good and on the bad side, but power is power and it’s built into exist and so the sinful nature of man, no matter what these people tell you. Sorry for the run-on sentence but I’m using voice dictation.
It's the governments that are dangerous in the first place isnt it. They killed more than anything ever has.
This American paranoia around the state and corporations is getting out of hand and is itself starting to pose a significant existential risk. They want your votes and your money, sure, but they don't want to kill you in your sleep. This paranoia will lead to the destruction of everything we care about, and the rest of the world will suffer the consequences of this very America-specific paranoia and delusion.
They're not putting AI in the glasses. They're sending info from your phone to your glasses, from a server running the AI elsewhere. An important distinction for a lot of reasons.
to be fair, not even one of the latest smartphone can handle LLMs, actually any
@@wsoulless Well, most modern gaming computers can handle 8b-14b models pretty easily, but yeah they likely need to get down to 2b at most for phones to try and meet them halfway, and that's just for the LLM alone.
I just think it's important to specify, because so many people seem to have the impression that they could run a 300b+ parameter model like fucking Grok on their phone. I've seen so many people asking how to set Grok up on their phone after it was made open source.
@@thearchitect5405 The number of params won't matter when the LLMs are trained, tuned, and optimized for specific uses.
@@CeresOutpost Specific use or not, LLM's take a lot to run. A specific use AI like stable diffusion still takes around as much as a 22b LLM does to run, which is the size of the individual AI Mixtral uses for it's overall mixture of experts AI.
To pack a 70b+ model onto a phone, we'll need more significant hardware advances than AI advances, so we're more likely to reach that performance with a new AI architecture rather than with smaller LLM's.
Now Mark Zuckerberg can see you in your most intimate moments.
How exactly does that work? You think the glasses transmit even when the lights is off, and that you'll always be in front of a mirror with them on, or take them off and point them at you, and they don't actually power off (as they DO)?
Zukerberg's goal is to split people resisting AI, from people embracing AI. And, he is succeeding with this glasses-thing.
I, for one, will not talk to anyone wearing heavy glasses now.
*Insert creepy pic of Zucc*
@@ZappyOh Plan for a future where you never speak to anyone then. A few people with glasses are NOTHING compared to the phone in everyone's hand.
@@brianmi40over optimistic boyish grandeur
Some people don’t want to give Meta their data. Others ask Meta politely to watch everything they see from their glasses
We’re screwed. Imagine this concept with semantic decoding
This will split people resisting AI, from people embracing AI.
I will probably avoid people wearing heavy glasses now.
If we want the tech we should just take it. We can make our own thats not controled by Meta.
And that’s why we need open source you cannot trust Meta or Mark Zuckerberg you see the news where they interfered 39 times with the elections since 2008. I don’t care if you’re left or right, but do you really want a big corporation to influence things based on what is good for them and not you?
I had the same thought…
Great news review. I do, however, disagree with you on not open sourcing. Those with evil intentions will always have access, so you’d only be preventing access to everyone else. we don’t want AI locked up in corporations and governments, that’s where it is the most dangerous.
So Open AI is following the nanny model and dumbing down their LLM. They should instead follow Anthropic's lead with Claude where they've been able to significantly reduce the number of times it declines to answer something or redirects the conversation and yet still manages to keep its integrity.
dude please with 30 minute videos add time stamps.... i aint watching this just to see 1 thing
👍
guys i have sad news…. i am not shocked in the slightest by this video 😢
The RUclips AI algorithm won’t show this to people if you don’t use sensational words
you're on wrong channel. On this channel we get stunned.
are you saying you are not SHOCKED? that’s HORRIFYING
But you’re STUNNED, right?
yes I get stunned with the news on this channel.
Meta glasses - An easy way to collect more data?
Guy with glasses: "Hey Meta, take a video."
Me: "Hey Meta, thats creepy."
a living organism IS an infinite data generation system. AI is just going to watch us like TV lol
I told an incubator in Rochester, New York that these are next up!!! Brilliant Labs, Frame as well.
What is the actual AI component of those glasses? A camera isn't AI.
I believe two things:
1) The future is going to be unimaginably wonderful
2) We're going to spend a lot of time in absolute hell to get there first - bioweapons, malware, the civil unrest from a meaning and jobs crisis. Do _not_ live anywhere near a city if you can help it.
I'm buying land in the middle of nowhere and I intend to live there for at least 10 years.
Make it 100. Only then you'll be "safe".
"Unimaginably wonderful"... For a few.
Poverty, scarcity , hunger, war. All those will keep existing for hundreds of years among humanity after the AI rise. Maybe forever. Watch any futuristic tv show where humans are already colonizing space. Colonies are frequently poor. Violence is born there and spread trough the system. That's how it is in reality and will always be, even if AI becomes cheap and widely used.
And bioweapons, malware, etc. Those things exist without AI. Not everyone wants to make those. AI simply will make it slightly easier for them. It is not a bigger deal than lowering the cost of computers with the side effect of hackers being able to buy cheaper processors to hack.
16:20 That guy is wrong. You can't compare the model that Google used to learn how to play Go, with a model that has to respond to a language prompt that requires an enormous amount of data to come up with a response, and it can be related to anything... coding, math, translations, summarizing, poems, stories, roleplay, general questions, you name it. On the other hand, the Go model only had to know the simple rules of Go, and then play millions of matches to learn what good and bad moves are.
That guy is the CEO of Anthropic and definitely knows more about AI than you. And what he said makes perfect sense.
I totally disagree with this guy on open sourcing large LLM’s. I think they should be open sourced. I do not expect future AI systems to continue to be disembodied. I expect a future where models will interact with other models, humans, other biological creatures, and directly with their environment. I think many AI focused companies are focusing on the wrong thing. The intelligence they need and want is one that sufficiently models the world and can make decisions to further its goals. What we don’t need is a slightly better chat bot. To gate keep intelligence is to gate keep ability aka power. We already have in society a method of changing the behavior of people, those things are called laws. If a system is goal oriented, laws would influence AI systems just as much as they influence people. Especially if we have AI systems enforcing laws. The systems of power we have in the world are already hardened to withstand people or agents trying to topple them.
The key to the stability of democracy is power in numbers. No one person or group or even government can have too much power if we want democracy to stand. And no you can't just give all of the power to the government, if you do they would have no reason to listen to the people. Right now the government has to listen to the people just as the people have to listen to the government. There is a balance of power. We have many examples of what happens when the people aren't necessary to a country, when the government has too much power.
The main threat from AI job loss isn't a loss of income or livelihood, it's a loss of the people's power. If you no longer NEED to work that means the government no longer NEEDs you. All of the shalkle’s people have on the government's power functioning because they need to keep us convinced to work to even have a working government goes away if people aren't the ones running it. A ruler who doesn't NEED other people to keep their power doesn't NEED to listen to their people at all.
Making a bio weapon seems bad and nobody wants it to happen. But you know what it is, its power. If everyone has power, nobody has power. And that's a good thing. We NEED to be dangerous to the government or we lose all of our power and rights, our rights aren't given they were taken.
And that's why Americans need their guns. Because the government probably will try to replace us with robots, which is fine. As long as they pay us a UBI and find a way to keep that amount of pay high enough that we still feel affluent as a nation.
Meta glasses are not Google glasses. Google had augmented reality, Meta doesn't have it yet.
I really like your videos because even you see its potential without being blinded. You stay critical.
I'm really curious if AI will become... Personal. Like... These models take a lot of power and hardware to run... Just like a computer used to take up a room to run a program. So will we see a day where we have our own AI? We can teach and maybe to work for us? Is that the future? Idk just thinking out loud 🤣✌️
You can already download many AIs already onto your computer, and the lower parameter ones don't take up a lot of computer power. Try getting Mistral or Llama 3 off of huggingface.
I really look forward to a "Her" type of AI. First company to create something like that will get my money.
I love the idea of the digital cortex, something that is a personal digital extension of yourself.
I imagine that having an AI which is trained as a companion (not a pet or a girlfriend) based on your life would be immensely useful.
They already have pocketable LLMs. Obviously not as good as the big boys. They're also making boutique AI inference chips. I don't know if Groq can be scaled down to a portable but there is another research group working on an analog chip. Once they slow down on progress they'll optimize on size and performance.
With graphene chips, way more highly trained models, should be veryyy doable
Bro that’s here. Do that. They are open source
The thing that is weird about meta glasses is that they are made by meta, a company that’s so intrusive to your data that I would never want to share with them my whole life through video fed glasses
The glasses look just like classic Wayfarers that the underrated Yardbirds singer Keith Relf, wore while performing the 1965 hit, Heart Full of Soul on the Shindig TV show 59 years ago.
Ok let's just trust the rich and powerful to be the only ones with access to super powerful AI. Nothing terrible could possibly come from widening the power gap.
No one with those kind of glasses or other devices will ever be allowed to enter my home.
30 seconds in, I'm not the target demographic so I won't be purchasing anything from them. Let their target demographic buy them lmao
I haven’t even got the updates yet as I’m the uk but they are awesome specially with all the AI updates coming hopefully to the UK as well
Oh man, I'll have to charge my glasses?
Of course a rich old man wants to keep the world the same.
As far as the infinite data generation that would only apply to close systems where the rules are known in the boundaries are defined
Thank you for your measured analysis of the talks at the end.
Hey Zuck, how can we gather even more data covertly to train our future robots?
Zuck: Glasses!
Great stuff. This is the only ai cast I listen to.
Eh ok, why would anyone want to know that??
What if you're watching the worst channel available?
Your watching it dude.
I’m developing my own “FI” Ferret Intelligence 😅
Darn.. I thought i had a dead pixel but realised it was your mouse!
This is a huge missed opportunity for a clip on that can be moved to different glasses (reading, seeing, sun).
omg, a device that basicly does the same as my phone just without the ai.. Maybe just add the ai to your phone?
The best part of this interview is where you said it could teach people to go to Walmart and build a bio weapon! You obviously haven’t been in Walmart lately! 😂. “attention, Walmart, shoppers !Rollback on bio precursors in isle 4”😂😂
Tell me ROB ME without telling me to rob you of your meta glasses lol 🤣
i record all my work with these glasses super sick
Ai jobloss is the only thing I worry about anymore. Not spy shades.
As far as using it for code, you just use very specific things like certain functions or like say loops where you can kinda overlook the code to make sure it’s not weird as far as reliability goes the critics right I watch test on llama three and it would get the same question wrong one out of three times it has to be 100% imagine AI flying in airplane where there’s 1 million flights but it gets it wrong one percent of the time, that would be 1000 crashes. But it’s in the early innings.
The Industry Stopped Getting SHOCKED........
They just can't get over people don't want to carry around more of their crap.
I guess in some "days" that's (Rabbit device) going to be standard in all smart phones .. no need for another device, who is going to invest into that ? doesnt make sense ..
@9:23 that seems ridiculously over engineered when you can just put that into the prompt completion in the right order with adequate prompt engineering
They had AI in them. I use my buddies in December 2023. I guess I just updated it
If they don’t gave my meta AI on my app I don’t want it
Awesome. Thank you.
Won’t llama be open source? Re - the point of AI models being opensource could result in a catastrophy…
Great Videos & content.. though your introductions are a bit “hyperbolic”… I have to take a Xanax after watching your video’s 🤣🤪 ….god knows what’ll happen to your blood pressure when AGI is confirmed 😊…. But keep the good work up…
Imagine if a.i on every device, phone, car, tv etc does an autonomous DDOS attack on every server while completely cloging the entire bandwidth over the entire world all at once, whilst uploading its own software on every device.
Awesome 🤩
Please let this happen!
All the Tech Bros whooping at Rabbit.
The gain of function lab was closed sourced. Open source everything what if a malicious actor created this dangerous ai and no one knew the source code they would be a step behind trying to stop the malicious act.
Google did it before it was cool!
Love your videos brotha!
This so fascinating. I want Ai to teach me The Ari. Ai what does this mean.
The Ari uncovered extraordinary secrets inside the Zohar's poetic words. He described a system of evolution that went far beyond what Darwin would explain centuries later, writing:
Against this: A minimal living system may be viewed as “a self-sustaining chemical system capable of Darwinian evolution”. It has been proposed that autocatalytic sets of chemical reactions (ACSs) could serve as a mechanism to establish chemical compositional identity, heritable self-reproduction, and evolution in a minimal chemical system.
Millions and millions of glassholes.
How long before cheap phones will be able to do the same thing? I'd guess at a year or so. The colour of the device is a warning and should be heeded.
I cant believe hes turning anti open source bruh. Yeah letting corpos and governments hold the God power is a wonderful idea
I wish i couldve gotten an ai representative of my grandmother. Imagine an ai with all of grandmas knowledge with her voice that could summon on command.
Don’t wish for the curse of immortality
@@kylev.8248 a. An AI persona of someone isn't them becoming immortal any more than their diary did. b. nothing is immortal in a universe headed for heat death. Even WHEN (and it's coming) we can potentially live unbelieveably long lives, there will STILL be accidental death.
SHOCKING ...pretty pretty SHOCKING
The RUclips AI algorithm won’t show this to people if you don’t use sensational words
Most of the fears about AI is the same as the fears when they released the internet. While we can share anything online, we aren't all sharing how to make weapons or diseases.
While I don't believe there fears, its clear they have made their AI for business rather than a useful research tool to help humanity. The reason its closed source is to prevent competition. They are making big money and don't want to share.
..hmmm at 7:30 the RUclipsr, a native English speaker, describes the rabbit device as “pretty pretty quickly”, he should use the hallowed ai to write his scripts and just read them out as he records, just a suggestion
We are training the AI with our data for free the company’s are not even paying us nor giving the product to the customer.they should incentivize the customer
Where should they mail your 1/8,000,000,000th of a penny that your personal data is worth?
The Vision looks kinda Mid
19:07 By that point, it will already be too late, the knowledge of how to get there will already be public, they might delay things a bit by not allowing access to their models, but it will be such a small incremental step from what's already out there by then, that a bad actor will be able to piece together something equivalent. We are already in too deep with too many people not being worried enough; if anything, open-sourcing such a powerful model increases the odds people will notice issues sooner and maybe be capable of using even the model itself to create preemptive defenses before some malicious actor unleashes an attack stealthy. Can you imagine the disaster if for example Elon goes on one of his benders and says something stupid to the secret X AI, or Mark "they trust me, dumb fucks" Zuckerberg decides it's finally time to free up space for the lizard-people occupation of planet Earth, or Xinnie the Poo realizes he doesn't need to "save face" if only his mindless slaves are immune to the next ultra-fatal pandemic, and so on? It's an arms race now; open-sourcing at least increases the number of players in the game and hopefully, there's enough good guys that one of them will get to the finish line before the bad guys get too big of a headstart.
Omg we dont even have AI capable of driving a friken car without killing how many now? Dont blame the AI, blame the idiot who released it, blame the govt which allowed it, blame the lack of hardware the data needs but DO NOT blame the AI. Its like blaming your shoe for driving fast.
[;ease add time stamps to your videos
2-3 years...in 3 years, we'll see if humans are still relevant.
Love your videos!
It’s gonna turn out that Mike judges Idiocracy the movie is gonna be reality as people rely too much on this technology is acrutch. I’m for the tech it just people rely on it too much.
The psychopaths are taking over for real now.
I want a device I can wear that distorts anything about me anybody films. My face, mannerism, voice, everything.
Time to get that Scopophobia under treatment.
@@brianmi40 time to off the psychopaths
No, we are not living in he safest time in history. If a king is not attacked but the sword hanging above his head keeps growing in weight he is not a his safest time, silly. You just moved energy from one stat to potential energy until the sword falls.
It only takes the outbreak of one more world war for this to be proven the most dangerous time in human history. That could happen 20 minutes from now. Humanity was far safer for the hundreds of thousands of years where their city/town/village couldn't simply be vaporized at any moment by some madman 10,000 miles away.
It only takes the outbreak of one more world war for this to be proven the most dangerous time in human history. That could happen 20 minutes from now. Humanity was far safter for the hundreds of thousands of years where their city/town/village couldn't simply be vaporized at any moment by some madman 10,000 miles away.
😂 Rabbit gadget is for cheating in school exams
How are you going to use it to cheat?
@@elivegba8186 by asking it the questions in the question paper and telling it to tell you the answers
What is wrong with people? The one thing that Covid taught us is that the so-called overseers lie and they abuse crisis. So you don’t think that will happen with AI?😂😂😂😂
Recognizing a spreadsheet from a photo and manipulating it is nothing new or unique. You can do it today with your phone. I don't understand the Rabbit hype. It's a joke.
Nice
go mavericks
rabbit device is so dumb lol phones already do that but 1000 times better than this machine
You have to be impressed by how much hype has been generated for a device that obviously makes no sense in the age of smartphones.
Keep those things away from me. Never trust meta.
So what, bring my typewriter to work because AI is all over any device already?
@elsavelaz talk about being hyperbolic. We're talking about meta, not AI in general. Try joining the actual conversation.
Stop shocking me! i'm gonna be thor if this keeps happening
Brain tumor maker.
not too worry, ai will make a treatment/cure for that!
time traveling digital intelligence
More CRAP we don't need AND didn't ask for. Tech has gone way too far and will only get worse.
lol
Tech bros are actively ruining the world. We never asked for any of this. It's being forced on us.
First
I'll be honest, as soon as I hear the name Gary Marcus, I skip the part. I am aware of safety issues, but mostly it shows me how difficult our society is. If human beings would be 'better' themselves, there wouldn't be much danger at all. But this guy... no... I can't listen to him. It is impossible for me. Unfortunately he was in some podcasts with other people and he was just plainly dumb... I don't like to insult people, but he was changing whatever he said the whole time until the point as he became contradictory to his own things... it was horrible. I wanted to give him a chance, but he proved it was just a waste of time...
Why do you unnecessarily add the word 'did' to your sentences?
And 'do".
and "you know"
1984 baby
No progress at all 😂
Where is ICC for the Zuckerberg?
META / FACEBOOK SAME DAMN 👎🏻👎🏻