You know that the licence that it was released under is a licence that is not true open source? Any development done on this license is owned by Meta 😅
@jamminrebel3614
Ай бұрын
Ofc 🎉
@SuperJonathanmatthew
Ай бұрын
after testing the 70b model in my local environment as a programming acceleration tool and to brainstorm with the LLM. I can say its truly amazing that we have access to such a model for free and offline. i don't think i have been this excited about open source LLMs and now i definitely think open source will be the future.
@jasonshere
Ай бұрын
Llama is not open-source, even if that's how Meta/Mark is marketing it; but more open systems will help push AI in great ways.
@DanielSimonJr
Ай бұрын
Nope. Linux is popular for plumbing the internet and techies, but regular users will still want a product from Microsoft, Google, or Apple as part of the platform.
@samhiatt
Ай бұрын
I just realized that when I start to spend all my time in VR at least it will _look_ like I've been out snowboarding everyday.
@redonebig88
Ай бұрын
🤣🤣🤣
@carsongabriel1949
Ай бұрын
Yes so true you can literally see the vr mark around his eyes😅😢😂😂😂😂
@_nom_
Ай бұрын
@@carsongabriel1949I'd say that's polarised glasses.
@jbardet
Ай бұрын
Guessing guess in Tahoe
@SHAINON117
26 күн бұрын
this guy actually is into sport hardcore he even is an amazing wakeboarder he does it with his family a fair bit hes pretty amazing
@Theguywithspectacles
Ай бұрын
His Eyes actually looks like he has been wearing Meta Quest For HOURS
@christianjoachimgruber2773
Ай бұрын
there was a side PD Projekt to Traun a VR AI Porn Model...
@Franklyfun935
Ай бұрын
*days*
@6AxisSage
Ай бұрын
Dudes been out in the sun.. much cooler dude than before llama3 but didnt make his pasty nerd skin any more sun proof than before.
@enihowrazer
Ай бұрын
😂😂😂😂😂
Ай бұрын
Sideshow Bob has been tanning.
@liloleist5133
19 күн бұрын
"... *we work with government* ." That's really reassuring, Marc.
@NuLiForm
13 күн бұрын
Everybody has to or they face jail time...at least he's come clean abt it. i know for a fact he didn't roll over willingly...they left him Zero choice
@Redmanticore
9 күн бұрын
@@NuLiForm no they dont. they will face this: less profit. no government deals. otherwise ELON MUSK would be in jail.
@Sammyli99
3 күн бұрын
LOL, he has been seeded by GOV, from the beginning the whole exercise of FB was a DATA collection exercise, with an NSA backend. Where have you been?
@theycallmethesoandso
Ай бұрын
"computer: end surf sim, load conference cabin"
@marc_frank
Ай бұрын
swoosh, beep beep
@TheRealUsername
Ай бұрын
Actually you could be right, he might have been doing the interview inside the metaverse, or at least a realistic version of it
@NuLiForm
13 күн бұрын
Holodeck gets one step closer...if i had won the funding back in 91/92 we would very probably have Home Holodecks by now.
@Redmanticore
9 күн бұрын
@@TheRealUsername nah he used a competitors service for realistic model. his own company is using resources for incompetent people that cant produce realistic models.
@brianolson8293
5 күн бұрын
The bright light affects the pupils and makes them smaller which gives the person a colder impression just my opinion
@AIPretendingToBeHuman
Ай бұрын
This is how competition is supposed to work. Go Meta.
@Gnaritas42
Ай бұрын
No, it isn't. In fact this is treading on anti-competitive monopoly lawsuit territory. Putting out shit for free that your competitors charge for is anti-competitive. You're trying to put competitors out of business and establish yourself as the only option. Not remotely what good competition looks like. When microsoft did this with IE they got into anti-trust trouble. Neither you or the OP has any clue what "competition" in a capital market actually means. Meta will get sued before long and the government will step in.
@Gnaritas42
Ай бұрын
@@hdvrNG simply not true, you've failed to understand the licensing.
@handsanitizer2457
Ай бұрын
@@hdvrNGread the actual license 😂
@nwchrista
Ай бұрын
Meta is a deep state Darpa program. Only a fool would cheer for it to succeed. Your children will scorn your name for all eternity.
@dusanbosnjakovic6588
Ай бұрын
Linux and these models have very little in common. This is all done by a single company pouring a ton of money into it. It's not a product of open collaboration.
@linuxdevops726
Ай бұрын
I just canceled all subscriptions to any other models out there because the 405b is a beast and a monster. I was so moved when it worked and did exactly what I wanted it to do. I'm amazed and deeply appreciative of Meta's effort in helping the open-source community, standing against the greed of closed-source companies. The only issue with this model is that running it locally requires a significant amount of system memory (212.8 GiB), but if you have that much memory, you are the boss.
@diliff
17 күн бұрын
Unless you already happen to have that kind of RAM just lying around, it doesn't really make sense to buy it just to avoid paying subscriptions. It would take years to break even and who knows what the AI playing field will look like by then. You might need more than a terabyte of RAM to run the latest competitive models, or maybe the open source models won't be competitive at all. The real benefit I guess is that you have far more flexibility to tinker with it, and keep your data local/private.
@polvoazul
12 күн бұрын
@@diliff For enterprise where privacy is a concern it make sense. A 200GB VM costs 1k dollars a month to run 24/7. If you use it in bursts to do batch jobs, you can make it a fraction of that. For a law firm, for instance, this can make a lot of sense.
@KristopherHesson
9 күн бұрын
Someone praising Meta, one of the companies that has perfected how to sell YOU and your data, is kind of laughable. No direct dig to you OP. Just strikes me as hilarious.
@user-qg8qc5qb9r
Ай бұрын
Introduction to Llama 3.1 and Meta's Open Source Strategy - 00:00:00 Meta's Scorched Earth Strategy and Competition with Closed Source Models - 00:02:02 Comparison with Linux and Open Source AI's Future - 00:04:29 Benefits of Open Source AI and Synthetic Data Creation - 00:06:00 Meta's Partner Ecosystem and Developer Opportunities - 00:10:44 Meta's Real World Use Cases for AI - 00:07:23 Meta's Vision of a Plurality of AI Models - 00:09:12 AI Regulation and Open Source AI Safety - 00:14:27 Economic and Global Benefits of Open Source AI - 00:25:04 Meta's Approach Compared to Apple's Closed Ecosystem - 00:27:23 Looking Forward: Llama 4 and Multimodal AI - 00:32:29 Meta's Vision for AI Agents and Future Directions - 00:36:16 Mark Zuckerberg's Thoughts on AI's Impact on Society - 00:40:57
@DantePowell
Ай бұрын
The one thing that raised my ears was when he mentioned that "Governments" are going to want their own models. Its already talked about models being in the hands of people looking to do harm or use it to do harm to humans.
@BruceKendallMartinJr
Ай бұрын
Darpa is Deepmind and has its hands all up in OpenAI and Meta....it's US vs China really for AI control of the rest of the Globe by 2045
@tarunrocks88
Ай бұрын
Governments have a lot of functions outside of military that would benefit from LLMs just like many corporations, think of HR, accounting, regulations, etc. In fact so much of government work is related to documents that you can say that LLMs will be extremely helpful to them. And they would surely not prefer to use a cloud API for this, they would want a private model running in their own infra.
@squamish4244
Ай бұрын
Once AI gets really powerful, like once it becomes AGI, it's inevitable that governments will want to nationalize it. It will be like nuclear weapons. Right now, the cybersecurity systems of all these companies are terrible. If we don't want China to be the first country to get AGI, we might want to change that. Or like Zuckerberg says, let everybody have it to avoid an AI arms race.
@polvoazul
12 күн бұрын
@@tarunrocks88 Based on adoption timelines of governments, it will start being adopted in 2050. Unless it helps them to concentrate power / win elections / shut down opposition. Then its 2025.
@Sammyli99
3 күн бұрын
We are way past that "inflextion point", these models come from GOV, its a reverse smokescreen as usual to make you feel all fluffy inside. Prob US NAVY is like 20 years ahead of this, just patent it, then drip feed into buddy-corp and get 20%.
@csebastian3
19 күн бұрын
This is the best Mark Zuckerberg interview i have ever seen. It has completely changed my opinion of him.
@mijki5867
10 күн бұрын
And that's why they make him look this way. That's the trick
@pubbets7290
6 күн бұрын
Mission accomplished.
@AlexT-o7u
6 сағат бұрын
filtered.
@xterity11
Ай бұрын
Not gonna lie, I like this model version of Mark more than the older models.
@nikkingman
24 күн бұрын
"Not gonna lie" added nothing
@pubbets7290
6 күн бұрын
@@nikkingmanpersonally, I agree.
@danielgrayling5032
Ай бұрын
The Mark Zuckerberg redemption arc is so good, it's great to see the writers feeling free to do something really creative.
@squamish4244
Ай бұрын
Now if only he'd limit Instagram. He knew girls were getting bulimia from Instagram and did nothing. A leopard doesn't change its spots.
@olivert.7177
Ай бұрын
I guess, if the weights of the first Llama version wouldn't have been leaked, this project would never have become open source.
@daveinpublic
Ай бұрын
Divine intervention
@jameshughes6078
Ай бұрын
there's a conspiracy theory that the leak was intentional, given the nda was more of an honors system than something ironclad. like there was no fingerprinting/steganography etc tied into the weights to identify which org leaked it. That is, they gave out access to the llama weights to researchers. The only thing preventing a leak at that point is honoring the license. The conspiracy claims that they didn't do it "in the open" to avoid any potential liability concerns with copyright, harm, etc
@Thedeepseanomad
Ай бұрын
The training data is not open, so it is not really open sourced
@SahilP2648
Ай бұрын
@@jameshughes6078 you won't be able to realisitically perform steganography on weights unless you want to just flip one or two numbers slightly. And coming to your last point, why not just open source eveything from the get-go? It would have been perfectly fine.
@bigglyguy8429
Ай бұрын
@@jameshughes6078 Yeah, I'd say most "leaks" are intentional, be they corporate or gov.
@fynnjackson2298
Ай бұрын
An important observation with this. When we exchange products we naturally think in terms of give and take, win or loose. However with AI we need to break out of that mindset and realise, giving people intelligence, weather it's artificial or not - is intelligence. It's no longer win-loose, but instead win-win-win-win.
@mikezooper
Ай бұрын
I don't agree. More intelligence hasn't made people happier. Instead, tech has disconnected us from each other (it's anti-social networking, and not social networking). It also disconnects us from nature. This obsessions with better better better, is a delusional approach to overcoming suffering.
@steveflorida5849
Ай бұрын
Human to human contact and sharing intelligence is (can be) ❤️ heartfelt. A major distinction from AI the communicating encyclopedia. Therefore, knowing the source of intelligence has benefits.
@jackderrida
Ай бұрын
What if the intelligence tells me how to make the best crack?
@ivanberdichevsky5679
Ай бұрын
NOTHING is win-win-win-win, that's a naive and innocent and DANGEROUS way to look at things. You're probably employed and enjoying a good life, but realize you're biased.
@humanoid9787
Ай бұрын
Okay keep that overly optimistic shit on LinkedIn buddy
@randy1984d
Ай бұрын
I am also digging his new vibe and how chill he has been lately, it’s like he went on an Ayahuasca trip and found himself.
@steveflorida5849
Ай бұрын
@@hak1985orgescapism.
@seanivore
Ай бұрын
Or accidentally ruined, arguably killed, inadvertently and really wants to fix things. Ayahuasca could trigger that though lol any therapy really. F microdosing tho lol have yet to understand that comparison, experientially
@allanshpeley4284
Ай бұрын
@@seanivore I haven't had much luck with microdosing either.
@SHAINON117
26 күн бұрын
all the tech whales have been i think its possibly because they know its close to being built and there gonna be remembered forever as bringing us AI and then abundance :)
@nomadicgodking
26 күн бұрын
@@allanshpeley4284microdosing causes small changes in awareness incrementally. A good dose of something stronger can induce immediate and long term effects. In my personal trials anyway.
@David-tp7sr
Ай бұрын
I transform like Zuck but every weekend - nobody has yet to come study me.
@vauths8204
Ай бұрын
Lol that's just called being Irish it's different
@jakeparker918
Ай бұрын
So you think...
@MonsterManStan
Ай бұрын
Just need the talent and the money!
@I-Dophler
Ай бұрын
🎯 Key points for quick navigation: 00:00 *🚀 Meta releases Llama 3.1, a groundbreaking open-source AI model* - Introduction of Llama 3.1, including a 45 billion parameter model - Model competitive with leading closed-source models - New community policies allow use as a teacher model for distillation and fine-tuning 02:19 *💻 Open-source AI strategy and model distillation* - Meta distilled 405 billion parameter model to create 70 billion and 8 billion parameter versions - Open-source AI becoming industry standard, following Linux's path - Llama 3.1 seen as inflection point for open-source AI 05:53 *🔧 Real-world use cases and ecosystem development* - Meta focuses on building partner ecosystem around Llama 3.1 - Vision for multiple custom AI models for various industries and use cases - Emphasis on distillation and fine-tuning capabilities 11:52 *🌐 Meta's motivation and ecosystem control* - Meta's investment in AI to avoid platform dependency - Shift from building for themselves to proactive ecosystem development - Partnerships with companies like Grock for ultra-low latency inference 14:38 *🛡️ Implications of open-source AI and safety concerns* - Open-source viewed as crucial for positive AI future and accessibility - Argument that open-source leads to safer and more secure AI development - Discussion on unintentional and intentional harms in AI systems 20:37 *🌐 Open-source AI and international competition* - Addressing concerns about open-sourcing advanced models to potential adversaries - Emphasizing the West's advantage in open and decentralized innovation - Arguing for a realistic approach to model security and the benefits of widespread access 24:31 *💡 Economic possibilities and democratization of AI* - AI's potential to increase productivity and accelerate the economy - Importance of giving startups, academics, and individuals access to state-of-the-art models - Equalizing effect of open-source AI for different countries and entrepreneurs 27:05 *🏗️ Meta's platform strategy and lessons from mobile* - Zuckerberg's experience building services on competitors' platforms - Desire to avoid past mistakes and create an open AI ecosystem - Comparison of closed (Apple) and open (Windows) approaches in previous tech cycles 32:16 *🔮 Future of Llama and multimodal AI* - Hints at Llama 4 development and expected improvements - Plans for launching multimodal models - Challenges with EU regulations affecting release strategy 35:52 *🤖 Vision for AI agents and product strategy* - Meta's goal to create numerous AI services rather than a singular AI - Plans for AI agents for small businesses and creators - Prediction of billions of AI agents in the future 40:12 *📈 AI development timeline and industry outlook* - Comparison to internet bubble and long-term commitment needed - Uncertainty about when AI will be ready for widespread use and business success - Acknowledgment of significant capital investment in AI development 40:54 *💰 AI industry economics and investment* - Major tech companies and startups investing heavily in AI, particularly in hardware - Current revenue not matching investment, suggesting potential for a mini bubble - Long-term optimism about AI despite short-term financial challenges 41:50 *🌱 AI's impact on livelihoods and open-source strategy* - Addressing concerns about AI's effect on jobs and personal prosperity - Emphasis on open-source approach and personalized AI models as a solution - Goal to create a more inclusive and sustainable AI ecosystem 42:57 *🔄 Lessons from Web 2.0 and future AI development* - Reflecting on past tech developments to improve AI implementation - Aiming for a better political economy around AI and emerging technologies - Importance of widespread benefits and support to mitigate public concerns about AI Made with HARPA AI
@picksalot1
Ай бұрын
I enjoyed listening to this relaxed, candid, open, insightful Zuck Model. He's showing a maturity that gives me hope AI will be a positive innovation and tool for everyone. Thanks
@illxet
Ай бұрын
zuck with the FU money look
@ajwright5512
Ай бұрын
Nah. Zuck REALISING he has "fuck you money" look.
@allanshpeley4284
Ай бұрын
FU money is $10M. The zuck has orders of magnitude more than that.
@mikemiller659
Ай бұрын
The hair ???
@brandongillett2616
Ай бұрын
I would absolutely love to see you interview Mark Zuckerberg about AI. You consistently have some of the highest quality AI videos and I love the direction Meta is taking AI right now. I think that would be an incredible interview.
@canoozie
Ай бұрын
The whole argument that people with large budgets will "check bad actors" is as scary as the old Reagan quote "The nine most terrifying words in the English language are 'I'm from the government and I'm here to help.'"
@RandomButBeautiful
Ай бұрын
yes exactly, all this guff about creating things to make life better is just a front for their absolute powermongering.
@antoniobortoni
Ай бұрын
If you have a compact, multi-modal AI model, you can continue to train it for a specific task and a rapid time loop, whether it's driving, operating a computer, or controlling a robot. This would result in a robot that understands human concepts and language, adapting quickly to complex tasks through targeted training. Its big.
@user-en4ek6xt6w
Ай бұрын
Imagine being one of the richest and powerfull man alive and still be chill like that 😂😂😂
@Dop3Dawg
Ай бұрын
That is a lingering effect from Jantelagen.
@patrickjreid
Ай бұрын
Mark is definately winning the cool billionaire contest. I won't use anything by Musk, I used to like him until he showed his true colors.
@user-bz5oj5qj7l
Ай бұрын
@@patrickjreidwhat do you mean?
@WeTheLittlePeople
Ай бұрын
Yeah Elon is pretty chill, but really not when it comes to his employees...
@jamesjonnes
Ай бұрын
@@patrickjreidMaybe Zuckerberg is chill, but Meta certainly is not. I know a bunch of small businesses that were ruined by their policies and I'd never create a business based on their platforms. When they ban you it's forever and without recourse even if the AI mistakenly flagged your content. Meanwhile criminals are rampant on their platforms.
@mrbrent62
Ай бұрын
One use case for AI is like my note taking APP Goodnotes. You can take hand written notes and it is able to spell check and make corrections while approximating your hand writing. If you pull in a PDF you can have a conversation with your notes. It will also let you search hand written notes. I believe behind the scenes AI will be very prevalent.
@Scalar_Alchz
Ай бұрын
It allready has been. For a Good Period Of Time. Sort of had a Closed Alpha during the 2015-2016 election. Where "Russia" and "Misinformation" first took over the algorithm. Onward It was distributed amongst the elite policy lobbies, research and gov advisory people, then late 2018 early 2019 it was privately distributed amongst the many, NOW Big names on youtube, from within tech, gaming, streaming and content creation. CSO's and R&E's went to Town That Year. And what was birthed was the Covid 19 pandemic from those with a Penchant for milking the collapse of society and leveraging narratives and opinion with 3/4 branches of Gov, behind the scenes
@Ben_D.
Ай бұрын
So cool. The race to the bottom ends when everyone plays on a level field. Open source is our best bet against Moloch.
@elbowspeak
Ай бұрын
And for Moloch!
@Ben_D.
Ай бұрын
@@elbowspeak I am trying to figure this one out. Cant. Will you explain?
@elbowspeak
Ай бұрын
@@Ben_D. Open Source intelligence can and will be used as tools to manage alignment and coordination in service of reducing the impact that multi-polar prisoners dilemmas contribute to the meta-crisis we're in now (see Schmachtenberger et. al.) In this context OSI fights against Molochian entropy. But on the flip side of the coin, OS intelligence can (and will!) be used to sow discord in service of goals that we (lovers of negentropy and coordination) don't agree with, e.g. accelerated entropy & disorder - social, economic, political...
@Ben_D.
Ай бұрын
@@elbowspeak ok. Well, thank you for responding. We disagree.
@elbowspeak
Ай бұрын
@@Ben_D. I hope you're right!!
@STARFLEETC0MMAND
Ай бұрын
Disagree: Extreme protection of AI is needed for at least 900 years, in order to protect this technology from enemy nation states and maintain centuries advantage over anyone. Not saying stoping development, just keeping it from enemy countries. Time for long term planning. ✌️🖖
@kristianlavigne8270
Ай бұрын
“Enemy nation states” - they are only “enemies” because of US hegemonic power fantasies 😅
@braineaterzombie3981
Ай бұрын
So what , do you think other nations will just stop developing ais if one nation does so.
@bombabombanoktakom
Ай бұрын
I love your commentary videos on other videos or documents. You deserve more. Thanks a lot! Greetings from Turkey!
@musekic4654
Ай бұрын
Easy-to-see why Sam Altman and ChatGPT didn't want to "tap the breaks" and slow down deployment - what the board of directors fired him for not doing.
@therainman7777
16 күн бұрын
I don’t think Meta has really made GPT-4 or Sonnet 3.5 obsolete, because virtually no one is able to run Llama 3-405B on their own anyway. It requires an INSANE amount of VRAM that virtually no one has. So you’re going to end up paying regardless, whether you’re paying OpenAI for GPT-4o or paying to rent GPUs to host your Llama 405B instance.
@BrainwiresCortex
Ай бұрын
3:44 - Yeah very nice of them to distill the models for us; that cost them some extra money... The reason we don't see xGrok as a model for Ollama is because X didn't distill down the 314B parameter model they released for Grok-1. In an attempt to answer your question, I doubt it about the first models... The first models where probably still distilled from an unreleased larger model.
@snickle1980
Ай бұрын
Yeah, im not impressed by these hobbled corpo AI models.
@Siapanpeteellis
Ай бұрын
Crazy good stuff! Go Zuc!!! I am so glad his robotic twin, Fester, was sent to the country for rest, relaxation, and recovery from his benzo addition. Excellent to see Mark get his competition by the short hair.
@Deepisusanti
Ай бұрын
I'm favoured, $28K every week! I can now give back to the locals in my community and also support God's work and the church. God bless America,,
@Citkls
Ай бұрын
As a beginner what do I need to do? How can I invest, on which platform? If you know any please share.
@DanDonald-dm6dg
Ай бұрын
Investing in crypto/forex is a good idea, a good trading system would put you through many days of success.
@Datavendor3
Ай бұрын
Same, I met Mrs Maria Davis last year for the first time at a conference in Manchester, after then my family changed for good. God bless Mrs Maria
@najaatudauda5804
Ай бұрын
YES!!! That's exactly her name (Mrs Maria) so many people have recommended highly about her and am just starting with her 😊 from United States
@johnfrederick2944
Ай бұрын
Please who is this This sounds so good andI would like to be a party to this, is there any wayl can speak with her?
@LiquidAIWater
Ай бұрын
It is a task oriented approach to AI. For example, if I had a camera tied to AI that only job is to watch my fence to see if someone jumps over it, that is all that is needed. etc Perimeter surveillance. I don't need the AI agent to know everything in the world and how to create an email in a Shakespearean tone.
@Scalar_Alchz
Ай бұрын
But Zuck Knows If Someone Is Jumping Your Fence Now.. Even Before You Do.. Actually Testing a New 'Hunch' Right Now In 3..2..1.. *WE HAVE LIFT OFF*
@TheRainHarvester
19 күн бұрын
19:00 no. It's easier to start a fire with ai, then it is to stop a fire with ai.
@knine1652
17 күн бұрын
I've never bothered to watch any interviews with Zuckerberg, but he really impressed me with his breath and depth of the leading edge topics of his field.
@Yipper64
Ай бұрын
0:35 he grew out his hair and looks like he was wearing a VR headset while outdoors giving him a face tan line.
@MoDs_3
Ай бұрын
Actually, the first comment. (Edit) I decided to show my appreciation for your quality content, not like other AI-generated BS channels. It helps that you like and enjoy what you do! Just for you to know how seriously I follow you! Don't I deserve a gift (laptop)? 😅 It's a head-scratcher how we, the collective minds of the open community, might find ourselves making headlines. Thank you, Matt! ❤
@wpdv
8 сағат бұрын
This sounds like the path WordPress took. I think this is very smart!
@redesign3dp
Ай бұрын
I really enjoy your work, great content. Would a video on how to fine-tune Llama 3.1 be possibly in the future? It would be interesting to learn a simplified effective way to create smaller models with specific use-cases. Thanks..
@RAZR_Channel
Ай бұрын
mark has improved... but he's still no Prodigy... He's an Average guy that was Bankrolled by Wallstreet and the Gov to create : F-BI Book... because they needed a poster child with a back story to cover that...
@anoldhippieshideout1080
Ай бұрын
The idea of being able to enhance a model, with the with a developers thinking, sets up an interesting use case. If I keep refining my model with my research, when I pass, my daughter might be able to carry around a thumb drive. When asked a question, she can say, "I don't, know lets see what dad thinks". The ability to use my data to refine the model, gives my daughter the ability to think that asking my model a question, the model will answer as I would. with 60+ years of winning and losing, would give her an amazing tool, to carry around.
@raymondluca7779
Ай бұрын
5:19 so he pretty much wants to pull a D&D have it open source and have people work for free on it and then take it back when the work is done?...these scummy nonworking company's make me sick
@lol_streaming
Ай бұрын
In the end, it's a win-win for us, the consumers. 👌
@WaltWhite71100
Ай бұрын
Matt, thank you so much for doing these videos. It’s a huge public service in and of itself. As an investor, AI researcher, and entrepreneur, I love the material and commentary.
@danieladler3210
Ай бұрын
I don't know which is more impressive, Llama 3.1 update or Zuckerberg 2.0!
@greenstonegecko
Ай бұрын
I think open source is majorly important for companies and their AI-integration. Highly specialised AI for each company... but as consumers we will use very large / general models in daily use.
@gubzs
Ай бұрын
I've seen several people react to this zuck interview, and this analysis is _by far_ the best. Most of the others just let him talk, usually while speeding him up by 2x because apparently we all have terminal ADHD now. This video is seriously appreciated.
@matthew_berman
Ай бұрын
Your comment is seriously appreciated!
@Stp1974
Ай бұрын
The AI in the Zuckerberg bot is really good. Almost lifelike
@jaredandcande11
19 күн бұрын
Statements on Open vs closed source. Open- This assumes a generational improvement of outlook on it's existence and that the next wave of data isn't pessimistic, while also assuming the next organic wave will also be less intelligent than the current and working organics. In Short, we must ask "do we really need this?" What aspects of organic life should be allowed to interact or train with AI models? What effective safeguards can we implement for organic existence and evolutions?
@1stsquadrone810
Ай бұрын
This version of Zuck is so much more life like, amazing progress!
@allanshpeley4284
Ай бұрын
We're making big strides in teaching AI to look and act human.
@bixxybear
Ай бұрын
For a long time I thought of Zuckerberg as just the Facebook guy but now maybe he could start changing people's lives. Zuckerberg could think of more ways to help Americans who are loosing out to big business and corrupt stock markets stealing their money.
@thedwarvendispatch5364
26 күн бұрын
It’s not just because this play worked for Microsoft, but also because the future of AI is ubiquity as a feature of AI.
@CalumnMcAulay
Ай бұрын
Zuck looks like he’s just emerged from an exotic lagoon, which is probably close to the truth. Looking awesome!
@tzardelasuerte
Ай бұрын
Lagoon? Mushrooms or dmt
@boonkiathan
Ай бұрын
His points on the importance of open source foundation models are exactly that of the many of the open source advocates out there. But nobody has anywhere close to his massive, massive financial trove. But to rely on Zuck to boil the oceans, we should be looking to fundamental innovations to expensive training (transformers) and combination of small open models and agentics
@Ninc227
Ай бұрын
The challenge for open source isn’t the bad actor’s AI becoming far better, it’s that in some cases, the bad actor can use other resources apart from AI (like time) to gain the upper hand. Eg both sides have the same AI systems, but a bad actor can work on a biological weapon for 6 months that then the good actor has to respond to immediately to prevent mass harm
@michamohe
Ай бұрын
I think the metaverse is going to get some interesting stuff as Meta improves it's AIs
@TheAceInfinity
17 күн бұрын
6:55 - The reason why they brought on devs though was to allow people to make money on their platform to attract more customers by focusing their attention to using their platform. It was still a business decision at the end of the day and now people pay money to advertise on facebook as a result. It was 1000% not just an altruistic decision based on Mark's love of open source... There's going to be a LOT of bad actors here surrounding AI advancements and we need to be on high alert. AI is out of the box, we need it to be fully available IMHO--this is going to be the least evil with this pandoras box.
@kingmo565
Ай бұрын
Mark is definitely impressing me on this transformation! He's looking like he's starting to care more about humanity as a whole, rather than how to exploit humanity. People first! profit later, but also, at some point, forget about profit when you're already a billionaire. Don't think about the high-score on the game of billionaires, no one cares, especially when it is at the cost of people and your soul.
@truehighs7845
Ай бұрын
Man no wonder Marketing is the leading force in the world, perception is everything, they are like Johnny 2 times and we are like Tommy 3 seconds.
@steveflorida5849
Ай бұрын
@@MrAcarlothe treasures you embrace directs ones goals. Business seems to be Mark's main objective. His image is surely secondary to being successful in business.
@tubestrom
Ай бұрын
"Figure out ways to monetize over time." We know this best as bait and switch. I'm not disagreeing with the strategy, but something to keep in mind when adopting this tech.
@denjamin2633
Ай бұрын
Dude is pretty forward. "Doesn't hurt my bottom line at all, but screws over Google and Microsoft."
@GenRicOpekc
22 күн бұрын
Thanks Matthew for posting this. You changed my mind on Zuckerberg and their ai approach.
@NuLiForm
13 күн бұрын
Mark....you are Returning to your Roots....about bloody time, my good sir. i am So Frikken Proud of you! ❤🌹
@shyama5612
Ай бұрын
The humanoid becomes man with curly hair, with a very likable personality- the transformation is complete.
@davidgang7642
7 күн бұрын
It would be cool to do the scorched Earth model with large scale agriculture software Or something that would be useful
@i1abnrk
Ай бұрын
I expect something parallel to how Android was born of Linux, or Windows from Unix. Essentially, the razor's edge with always be with the open source community. As the base matures, it will spawn off major commercial enterprises that become ubiquitous over time and paid development will create stable everyday products eventually.
@cerebralparasite9238
Ай бұрын
It's weird that I agree with him now.
@seanharbinger
Ай бұрын
I see this as a good opportunity to make amends with Mr. Zuckerberg and forge a new future!
@Matlockization
Ай бұрын
No thanks.
@truehighs7845
Ай бұрын
They are still trafficking child corn and censoring health issues, those who know they got fooled by cool Britania know better.
@s-mh7mq
14 күн бұрын
Within open source, there were/are two models - one where the user doesn't have to release the improvements back into the project, and another where you are bound by license to give back to the mother project by disclosing the source code of all the changes/improvements you make to the product at your end. The second model, found in Linux for example, has been far more successful than the first model as it prevents parasites, and gives the project sustainability above and beyond depending on a corporate sponsor like meta. So, which one is Llama following? If companies and governments (China etc) use llama to build something else, say a completely autonomous killing machine or group of machines, will that government be forced to disclose the weights and source code to the public? If not, I'm not sure this is very safe
@sma1015
Ай бұрын
Is it just me or it seems like Zuck just saved the planet from slavery?
@Gen0cidePTB
Ай бұрын
Only if they continue this into AGI. AGI is when you get the slavery issue, AND you need to fix alignment on the way.
@henrismith7472
Ай бұрын
I ran the small llama 3.1 model locally on my galaxy s24 ultra. It unusably slow, but I was impressed that I could even do it already.
@SuperObermotz
Ай бұрын
How did you manage that?
@henrismith7472
Ай бұрын
@@SuperObermotz I asked perplexity how, and kept uploading screenshots when I got stuck. I have no idea what I did really lol.
@henrismith7472
Ай бұрын
@@SuperObermotz I used Termux (from GitHub not play store) with Kobold-CPP, that's all I remember. Oh and I think I used a quantised version of llama 3.1.
@rikkousa
Ай бұрын
@@henrismith7472After I first used Perplexity, I no longer have a need for Google search which has become a total joke in the last several years.I used to be able to find whatever I was looking for--infinite number of pages to scroll through. Today, two pages, non of what I want , mostly ads. Go Perplexity AI
@paul_shuler
Ай бұрын
great vid, luv Zuk's vibe, glad he's having fun! let's go open source! 👊🏻
@ThomasConover
Ай бұрын
It’s mind blowing how this guy turned out to become a AI open source messiah. It’s such a surprise move I’m still pinching my arm to see if I’m sleeping. ❤🙏
@Stratton218
Ай бұрын
crazy 11:35
@stridedeck
7 күн бұрын
Good AI vs. bad AI is dangerous only when the bad AI uses another process. Currently, all AI models are LLMs. Within this world of LLMs, no AI will get ahead of the other AIs in a large strategic advantage. But, LLMs is not the process our brain uses. When AI develops a brain-type process, than that will have strategic advantages over all AIs using LLMs! The brain process is simple to do! Very simple!
@brianolson8293
5 күн бұрын
I just had to pause for a second , because he just said that they weren’t going to build something and throw it over the wall and see what develops 18:31 , We are talking about AI aren’t we? That comment should be miles away from this conversation
@MontyCraig
Ай бұрын
3.1 8b is sweet! Great at function calling
@tiagotiagot
Ай бұрын
All it takes is a single person booting up a singleton exponentially self-improving ASI on a whim and we're doomed. Since this whole thing started, we've been playing chicken trying to figure out how to get safe ASI before intentional or accidental/spontaneous ASI emerges with a random alignment. Open source/weights helps with defending from the intermediary dystopian scenarios; but there is no clear answer for how to avoid ultimate doom just yet, free open AI could lead the solution or accelerate the timeline to doom, and no one knows which way it will go.
@2021philyou
Ай бұрын
content rich video and insightful comments, well done, thanks Matthew
@matthew_berman
Ай бұрын
Glad you enjoyed it
@marcrettew8284
Ай бұрын
It's definitely not "in zuck we trust", but as far as his vision for A.I./agents and open-sourcing...i think he is on the right trajectory long term.
@GraveUypo
Ай бұрын
Never trust big tech. We benefiting from it is a "side effect". but what he's doing is a win-win for everyone regardless of his real intention. so yeah, no need to antagonize OR worship
@Fingerlelucky9
21 күн бұрын
I'm trying to figure out the 5th state that is on his back wall. He has California, New York, Hawaii, Massachusetts, but wth is that one to our right?
@ashtwenty12
Ай бұрын
I wonder if zuck partly wanted to make some open source modlee, so that he could just add it to all his products quickly and easily
@PriNovaFX
Ай бұрын
I probably guess, that after pre-training of the 405b model, Meta distilled it down to smaller models with transfer-learning techniques.
@sugaith
Ай бұрын
I OWE so much to Zucky. I work with React and React Native and a bunch of more other OpenSource techs that Facebook created. Facebook itself was not important. What was important is the tech that they created and the support for the dev community, which we support back and thats why Facebook/Meta will NEVER FALL
@mitchellmigala4107
Ай бұрын
At 405 Billion Parameters it is going to be HUGE and extremely costly to run this thing.
@edwinwagha
Ай бұрын
You're right.... The weights alone are roughly 200GB....
@geoffreyanderson4719
Ай бұрын
You can now also add AI to the pool of talent that can examine open source models. This is a task AI will excel at.
@jackflash6377
Ай бұрын
Problem I see is that AI is advancing so rapidly that in a few months something else will come out that's even better, smarter, faster. However, I am looking for ways to integrate AI into my products. Doing AI may not be profitable now but think of the future. Once people get a real grasp on using agents, training AI for specific uses and building AI into their apps from the beginning, the world will change for the better.
@voinea12
Ай бұрын
thank you zucc 🙏🙏 now all you have to do is grow a beard and you'll turn into a human person
@4mb127
Ай бұрын
Commodize your competition indeed
@WeTheLittlePeople
Ай бұрын
Okay so most of us more cautious techies know that OpenAI had said they would be open source and were, and then weren't. So is Meta is doing the same thing? Or will do.. Probably if they get something unique. But LLMs are limited, they extroplate and suppose solutions and often cannot remember why they made a decision. I dream that either OpenAI or MetaAI start focusing on retaining a decision engine tree / spine so it can backtrack and figure out why it made conclusions - without this LLMs are just a fancy Eliza with some really good data accessing techniques.
@vladimirkraus1438
Ай бұрын
I bet that in 20 years from now we will remember years 2023 and 2024 as the years when we as humans terribly and irreversibly screwed everything up...
@Jerominus77
Ай бұрын
Good Zucky upgraded his hairdo to launch something 'more sexy' which should be appreciated in the ecosystem hype. We do know you want to master the ecosystem and therefore apply the 'Scorched Earth' strategy. You've done it before and will do it again. We'll all have to Zuck it up again - no matter the hype of the day or tomorrow's curly hairdos.
@weredragon1447
Ай бұрын
I never thought I'd root for Zuck, but here we are. 🙃
@jimbojones8713
Ай бұрын
the good vs bad actors with ai, where the good actors with ai will protect from bad actors with ai, as long as the good actors have just as powerful of a model makes almost 0 sense. that is the most blanket statement that has almost no meaning. a bad person with a gun can harm many people before a good person with a gun can take them out. fwiw now that the AI pandoras box is open and it was always going to happen, i think its a good idea to accelerate AI advancements as fast as possile
@miauzure3960
Ай бұрын
I think smaller models are NEVER standalone versions, it's always distilled from bigger version. In other words, smaller models are teached by the larger model to immitate it as closely as possible with the limited parameter count.
@macowayeu
Ай бұрын
Definetly more agents than people in this world. 😮cool appearance MarkZuckerberg ❤ really
@lordfondragon
Ай бұрын
But the model is not fully open source, the fact that we can access it doesn't mean the license is like MIT or APACHE. When Zuck releases a model under one of those TRULY OPEN SOURCE licenses, I'd consider that a win.
@Helpsmallbusinesses
Ай бұрын
Liked for the news and Mark's new vibes.
@denijane89
Ай бұрын
I'm grateful to Mark Zuckerberg for having llama3.1 (and all the previous ones). Having an open-source AI and locally run one are incredibly important for the over -all progress. I don't want to depend for all my AI experience to 3 companies, I want to be able to run models, do RAG, fine-tune and over all play with stuff. And have them safely on my computer. I wish I could train a model, obviously I can't, but at least I can come up with my own uses, submit bugs and be part of the community. That means a lot for me, so for this I thank him. Facebook, though, right now is practically useless for me. I see all these ads/suggested profiles when all I want to see is what my friends are doing. Thanks but not thanks.
@Paragon_Reason
Ай бұрын
Can you explain how a normal layman can use this? Where do we go to get it? How do we use it? What can it do for us or our business? Thanks.
Пікірлер: 711