Zuck just released Llama 3 and made history

If you want to learn how to use Llama 3 to build AI Agents, join my community: www.skool.com/new-society (tutorial releasing tomorrow!)
Follow me on Twitter - x.com/DavidOndrej1
Please Subscribe.
Open-source community, we're so back.

[@DavidOndrej]

If you want to learn how to use Llama 3 to build AI Agents, join my community: www.skool.com/new-society (tutorial releasing tomorrow!)

[@troll6595]

When do you think we will get an uncensored model?

[@HoneyMarketingBoard]

I thought Grock had 315 billion perameters. And its free with source code and weights. or am I confused ?

[@HakaiKaien]

@@HoneyMarketingBoard You are correct. But you cannot run it on a personal computer. Llama 3 has 8 b parametres and can run on a potato computer.

[@fliu5282]

naaa. it's a bunch of b.s. because the data as the "knowledge of human" is all woke, conjecture and close to fake, despite the 70 billions of modes per second, it does not matter if the data input by a "human programer" is the woke left. So not going to use. it. To me, it is a non-event. Who gives a fk...naaa. Plus Zuck is pro-china. I mean for example, see if you dare to post anything about Xi Jinping, your account will be "This Violated our Posting Policy and account is suspended". So China is behind this Dolly Lamma AI. Not gonna use it. It does not matter. It's a Pro-china AI..means nothing...to us 77% of Americans.

[@doingtime20]

Zuck is becoming human. I do like the lose hair and the chain, he looks more relaxed.

[@TheExodusLost]

He’s trying

[@future-mindset]

no it's called winning the AI war long therm, and collecting most of data, why free, every business want you

[@Corteum]

Nah, not falling for it. after all that shxt fb did

[@keithlol]

So his AI upgrades are working!

[@noncomplacent]

It's called pr. He's well aware of his bad image. He will now manipulate people into thinking he's normal.

[@PseudoProphet]

Just imagine, there are going to be even better versions of llama3 within a month.

[@DavidOndrej]

Can't wait

[@2BluntsLater]

And then Every week

[@PseudoProphet]

@@2BluntsLater Chinese will be the first to come up with their version of llama first, like always. 😅

[@purplemashine9122]

And then every day, minute, second, millisecond,etc.

[@eric7758]

@@PseudoProphetthen they will claim how innovative they are 😂

[@user-uy4lu9fl3u]

Zuckerberg becoming a great dude and open sourcing was NOT on my 2024 bingo card

[@Corteum]

He's zucking you all in. he's a hacker. it's all social engineering

[@zero_given]

Zuck transformin into the bussin meme I like it

[@Corteum]

Nah. dont like it. he's still a robot. and reemmeber, he doesnt think much of you. "Domb Forks!"

[@IdPreferNot1]

Your enthusiasm is contagious and justified. I cant believe how much OAI is sucking right now. I know OAI is going to have better things, but their approach is revealing that they are not interested in truly advancing things for our benefit.

[@Darkt0mb5]

Mark is very human

[@DavidOndrej]

it's hard making videos on Zuck because I always have to edit out his lizard eyes

[@SAINT_PATRICK]

😂😂😂

[@moderngames8892]

😂😂😂

[@Noqtis]

Zuck the buck can suck a duck.

[@Darkt0mb5]

Hard for me to watch Zuck and not love the guy I used to want to be a herpetologist

[@rebeccamiller8772]

Best masterclasses on all these topics. Thanks David!

[@DavidOndrej]

Glad you think so!

[@GeoMeridium]

This is incredible! A 70B open-source model with better performance than the original GPT-4 or Gemini Pro 1.5.

[@DavidOndrej]

exactly

[@jmg9509]

Hey can you explain to me what this will mean if possible? Because currently I'm using the free GPT 3.5. Thank you =)

[@TheThetruthmaster1]

So itcan run on my own pc

[@daryladhityahenry]

@@jmg9509 For normal people that doesn't have powerful GPU, basically nothing changed. But the 8B version also very very great enough ( based on benchmark ). But if you have quite powerful GPU, that changed everything. You don't need to use GPT 3.5 anymore, you can freely use it like their API... So it's super great ( for anyone who can run it on their own computer ) if the benchmark was true.

[@HakaiKaien]

Oh yeah, the 70B model will be insanely good as agents. Which reminds me, I need a new goddamn machine

[@makesnosense6304]

2:30 I just want to say THANK YOU for mentioning that with regards to models not really being open source! People misuse the term so the model seems good, while in reality it's like calling a png file open source. Having an image you can change and tweak doesn't mean you have the source to create it.

[@MrMwenesi]

Excellent point David regarding the need for open source models.

[@coinspeednews4703]

Hey David, love your content and your style! Please share what hardware you are running a 70b model!

[@testales]

Since you were wondering, you can do realtime image generation with ComfyUI with a fast model and instant update on change enabled.

[@I-Dophler]

Honestly, this seems overly optimistic. While the excitement for Meta's Llama 3 is understandable, it's important to remain critical about such releases. There are potential issues and implications not addressed here, such as ethical concerns and the actual usability across different platforms and countries. Plus, claiming it will replace other models overnight without extensive real-world testing and feedback is premature. Let's not get carried away by the hype and remember that these advancements often come with their own set of challenges and drawbacks.

[@solaawodiya7360]

Exactly ❤. I feel the same

[@DavidOndrej]

what ethical concerns? everyone having access to SOTA models? If anything, not giving people access is unethical.

[@I-Dophler]

@@DavidOndrej I agree. It's great that companies are making this type of technology available to everyone. Increased transparency and wider access will promote more responsible use.

[@I-Dophler]

@@solaawodiya7360 Definitely! This type of access will be a game-changer.

[@Dynoxprakash]

It is good I got access

[@BlaqSheep_TV]

This video gave a great breakdown as to why open source models are important, thank you!

[@freddy29228]

I just watched Matthew Berman's video about Llama 3. Good tests!

[@Krienfresh]

Hey David, I really like ur content. I was curious what experts are in your community. Cheers :)

[@ninanordbo]

yes tbh respect to zuck for this, altho agree the chain is over the top

[@BobbyMasteria]

zuck himself is over the top !

[@jesussaeta8383]

Wow David impresive

[@brianhopson2072]

And next week there will be an even bigger large language model released. The cycle just goes on and on and on.

[@loichommel8197]

Llama 3 8b and 70b is now available on groq!!

[@strzaua123]

Do you know any examples of use cases for this basic 8b model that can be used locally?

[@holdthetruthhostage]

Is it mixture of Exports thats what's needed 8x70b

[@sirens3237]

Still not available in the UK...

[@tomasbusse2410]

Look forward to your video. And it works with a VPN just tested out seems good on first glance.

[@DavidOndrej]

Looking forward to hearing your thoughts Tomas!

[@tonnytops]

David Don't forget "If It's Free, You are the Product".

[@ShivaInu42]

Just in time! We needed a buff to our lyric generating superpowers now that version one of usuable text to music is here

[@rick_amsterdam]

Could you further explain??

[@matthewbond375]

I got the 8B model from Ollama the miniute it hit, and plugged it into a crewAI "research and report" crew. For JUST function calling, it performed very well. But for summarizing, it fed me a bunch of DEI garbage that had nothing to do with my prompts. Still, this can get worked out with fine-tuning, or better prompting on my part. I'm always excited to see new FOS models drop, and kudos to Zuck and Meta for their contributions to free open source AI!

[@luisfonseca9045]

Note they say *MEASURED* in the chart comparing Llama3-8b to other 7b models... Don't know who or how they measured that, but some scores for gemma and mistral are actually way lower than whats shown in huggingface's LLM dashboard.

[@kimaegaii]

Is there a way to use it without having to download the whole thing on your computer and if you don' thave good GPU's?

[@briankgarland]

Good lord....don't hand over even more of your data to Zucker.

[@jeffwads]

I don't see anything ground-breaking in the Meta version offered on their website. It makes some basic mistakes that something like 70b deepseeker gets correct. Example: Suppose I’m in the middle of South Dakota and I’m looking straight down towards the center of Texas. Is Boston to my left or to my right?

[@dearlyzin]

We are still waiting for Claude in Brazil, you can't even create an account yet

[@emanuelmma2]

Nice

[@djnoj3371]

So I just tried it and it links to your Facebook and you can create and edit images and make them into videos

[@ddabo4460]

Love your videos. But dude, how much coffee did you have before making this video , lol

[@archonjubael]

I checked it out and it is super fast.

[@HakaiKaien]

Alright, this is a step in the right direction. It actually gives me hope for a future where we maintain our individuality and sovereignty.

[@vikashshrestha8536]

Not in uk also available......😢😢😢😢

[@makesnosense6304]

6:52 What you do is not just believe what the model is giving you. Keep in mind that a model is just a statistical model where tokens are guessed based on context and other parameters.

[@qaon5748]

where is the demo? i dont wanna see a table

[@conroybogle3713]

You spoke of praising Mark Zuckerberg, you should be praised as well. I am baffled how it is that you have only 410 members in your community. The $77 is more than I can afford right now, but I will for certain be joining. Also, you should say what country you're in. Because doing so would most likely result in faster deployment/adoption for your entire country.

[@jayhu6075]

An important limitation point for us to know is: the Llama materials are restricted in their ability to improve other well-known language models. While developers often aim to refine models for better performance, the licensing terms limit the use of Llama materials, unless specifically using Llama. This highlights the importance of understanding licensing agreements when working with AI models. Overall, the license appears to impose some restrictions on the use of the Llama materials and emphasizes that this model is NOT entirely OPEN source.

[@garypick]

Can Meta detect you in your country if you use a VPN?

[@hskdjs]

llama 3 is already available with groq

[@paul1979uk2000]

Yep and it's insanely fast, the 8b model got 900 tokens lol. Also, it seems to work in Europe as well.

[@rickdobbs1a]

To me it doesn't matter if it's open source or closed source. Things can be manipulated to make you think and do things that you would not normally do. So what we need to do as humans is to remember to use our brains and think for ourselves.

[@TheRealUsername]

That will be crazy af that an open source 400b model catch up with GPT-4 Turbo or Claude 3's performance

[@DavidOndrej]

Indeed.

[@ShaneStevens-pd3nz]

Im using it right now...

[@nhtdmr]

make video how to use it on laptop

[@troll6595]

When will we get an uncensored model?

[@DavidOndrej]

as soon as someone makes it lol

[@troll6595]

@@DavidOndrej When do you think that will be

[@future-mindset]

You seem very smart, can we organize a call?

[@testales]

There are already quantized versions of the 8b and 70b versions on ollama ai. These guys are quick. The q4 70b is probably the best model I could run locally so far. I still can't believe that actually something good is coming out of Facebook. I wonder what the actual reasons are because I doubt that Zuck does this only because he suddenly turned into a nice guy. The llama models are for LLMs what Stable Diffusion is for image generations. So without Meta's and Stablity AI's releases we would probably have next to nothing in terms of AI stuff.

[@Corteum]

Cambridge Analytica!

[@jamisony]

If the best AI model is open source, yet the company with the most user has the power of distribution. What’s the new dystrophia?

[@shimblypibbins]

couldn't you have used a VPN to access it?

[@DavidOndrej]

probably, but constantly switching my locations is a nuisance

[@Bsurfing]

David you need a vpn to by pass all of these regional restrictions.

[@RABWA333]

What does it mean 7B or 8B model?

[@BlackEagle352]

7 and 8 billion parameters.

[@cyborgmetropolis7652]

I see Llama 3 in Ollama.

[@DavidOndrej]

NO WAY

[@HakaiKaien]

Well, that was fast

[@Rollthered]

UI matters more to me, I prefer the UI of ChatGPT and it does good enough. I don't need "the best". I just need somewhat alright.

[@JJGhostHunters]

Great point! ChatGPT 4 is doing everything I want it to. Why not just pay for ChatGPT? It is absolutely worth it.

[@ominousplatypus380]

GIVE ME THE ZUCC

[@Charvak-Atheist]

wow

[@thr0w407]

Gpt 5 shouldn't be long now. Wewt to real competition and +1 for open source!!

[@TheInfoPlace]

You should do snippets of the video and comment on them instead of the jarring pause then comment. Appreciate your videos

[@yumaheymans]

vpn?

[@bobharris5093]

yeah but what can you do with those ? I just spent days on crewai, and autogpt, they're useless

[@DavidOndrej]

I could not disagree more.

[@juancarlosrial7197]

Because you don’t use a VPN to access

[@cucciolo182]

LOL. You could just use a VPN, or you’re just trying to tell something to Mark

[@BruceWayne-jh4wh]

Have you tried using a vpn?

[@teatray75]

Lol on the chain!!🤣

[@PseudoProphet]

Grok is moe, llma will 400b dense model😮😮

[@DavidOndrej]

Exactly. This model is about to go crazy.

[@EpicratesDamus]

Cant you use a vpn lol

[@keepforever726]

couldn't you just use a VPN to spoof like you're in USA to use?

[@rogerc7960]

It's posted on huggingface

[@DavidOndrej]

And Ollama too!

[@DarwinianUniversal]

I've been putting Llama 3 to the test and it doesnt rate that high

[@globalcharts369]

Llama 3 has cutof date in somewhere in 2022, so it not so relevant to current information. Gemini has cut of date October 2023. Which more recent usefully.

[@a.x.e.l8971]

I used a VPN and i had access to it

[@thecooler69]

turns out Meta didnt develop this, Zuck just copied the code over from his own robot brain.

[@DonaldKramer-qo6gr]

Is this $77.00 USD a Month? The video added confusion?

[@DavidOndrej]

Yes Donald, it's $77/mo. Join and see if it's worth it! If you don't find it valuable, I will personally give you the money back.

[@FMSTheWord]

Ithink with Chat GPT i own what it creates, I could be wrong, what do you think? 5. Intellectual Property. a. No trademark licenses are granted under this Agreement, and in connection with the Llama Materials, neither Meta nor Licensee may use any name or mark owned by or associated with the other or any of its affiliates, except as required for reasonable and customary use in describing and redistributing the Llama Materials. b. Subject to Meta’s ownership of Llama Materials and derivatives made by or for Meta, with respect to any derivative works and modifications of the Llama Materials that are made by you, as between you and Meta, you are and will be the owner of such derivative works and modifications.

[@virtualassistantbureau]

They will also mine and spy on your conversations etc lol

[@DavidOndrej]

That makes literally zero sense. The model is open-source, meaning MILLIONS of developers can look at the source code and spot any "spying" going on. Not to mention you can run the model locally, without any internet connection, so I'm not sure how exactly you imagine the data leaving your computer.

[@Hollera8]

It'll probably be as popular as the threads app lol if it's related to IG, FB or Zuck, many people (myself included)... don't care. Mostly because of the level of bias and censorship

[@qwazy0158]

Why don't you use a VPN to access the site(s)?

[@DavidOndrej]

it's a solution, but very inconvenient constantly switching between countries

[@qwazy0158]

@@DavidOndrej I guess. You could open another browser like Opera and use it's built in VPN and just switch tabs between browsers.

[@cornelljohnson7854]

I like to be included please?

[@Instant_Nerf]

H.E.R

[@Instant_Nerf]

Look people always want to time travel! Why isn’t anyone making agents to use Gaussian splatting to 3D reconstruct a city as it is in 2024.. by 2044 you could go back 20 yrs to see how the world looked like back then.. a way to time travel! And so that for every year. Maybe using google maps data like the 360 photos or other satellite data. It is possible .. please someone do it else I will!

[@rikkousa]

It may be the smartest but I want nothing to do with Facebook…or Google.

[@hydoffdhagaweyne1037]

Yes, please don't use it. Meta servers are getting crashed by the shire number of traffic now

[@VanCliefMedia]

Sonnent is like an infant compared to Opus. This is cool but Llama is still way behinf anthropic and open ai. Dont get me wrong. Love the open source. But thisnt isnt game changing in the big picture

[@sindrejakobsen]

just use a vpn bud...

[@torontoyes]

Bro,. VPN

[@HR-Services]

David, you are excited as a child... So you believe Zack has improved and remorse? Grow up.

[@radicalaim]

?

[@markmuller7962]

I mean education is a thing and I wouldn't be shocked if they implement some basic education, values and morals into the most powerful AIs out there. People are nasty enough for the world to not give up the efforts in education, actually the opposite makes more sense in this nasty nasty world we live in. "And who decides what's right and what's wrong?" Morality is a fluid concept that keeps changing overtime and we definitely need very intelligent experts that understand modern values and can keep pace with the times and that can definitely be a challenge. If one is afraid of being brainwashed by an AI... Well, that person values might be weak, fragile, and maybe wrong in the first place

[@pauleharrisiv8736]

1st

[@hope42]

Please David stop your clickbait ChatGPT goodbye. GPT is not going away think how insane you're coming is. I'm considering not subscribing to you now. Do you realize how much money is behind GPT? Don't market clickbait.

[@CarrotStick2165]

I gave up with that rubbish ages ago. And yes I used a paid version. Perplexity beats them all and by using the free version.

[@AroundDurban]

Its average. GPT4 is still better