I really think open source has it in the bag.
Вставка
- Опубліковано 10 бер 2024
- Open Source AI is the collective inheritance of humanity IMO
WIN an RTX 4080 Super! • I'm Giving Away an RTX...
▼ Link(s) From Today’s Video:
Elon's Tweet: / 1767108624038449405
Open AI's Blog Post: openai.com/blog/openai-elon-musk
Emad's Tweet: / 1767117571956810216
Open AI New Board Members: openai.com/blog/openai-announ...
Open AI President GPT-5 hint: / 1767049366999810549
► MattVidPro Discord: / discord
► Follow Me on Twitter: / mattvidpro
-------------------------------------------------
▼ Extra Links of Interest:
✩ AI LINKS MASTER LIST: www.futurepedia.io/
✩ General AI Playlist: • General MattVidPro AI ...
✩ AI I use to edit videos: www.descript.com/?lmref=nA4fDg
✩ Instagram: mattvidpro
✩ Tiktok: tiktok.com/@mattvidpro
✩ Second Channel: / @matt_pie
-------------------------------------------------
Thanks for watching Matt Video Productions! I make all sorts of videos here on UA-cam! Technology, Tutorials, and Reviews! Enjoy Your stay here, and subscribe!
All Suggestions, Thoughts And Comments Are Greatly Appreciated… Because I Actually Read Them.
-------------------------------------------------
► Business Contact: MattVidProSecond@gmail.com - Наука та технологія
Open-source chads... we can't stop winning !
And Bill Gates is a charity that's he keeps getting richer and flying 3 private jets while owning a charity, and his secretary is the board member.
I chose this video over the 29 minute video because I was straight forward
Appreciate
Watch both
Trying to remember a video where Matt didn't have a cold lol
facts
@@MattVidPro Get that honey garlic fermenting #HealthyMatt2024
Open-sourcing Grok as a way to shame OpenAI into open-sourcing GPT is like asking Dodger Stadium to open up for a little league team because the local park does.
I don't know if that analogy is directly applicable, but to be fair we don't know how good Open AI's next model is
Is grok good enough to really compete?@@MattVidPro
@@MattVidPro Given the competition, an LLM only has value if it's the best option for some use case. For what use case is Grok the best LLM? If you, or the person looking to use an LLM, can't answer that question then it has no real market value. Their only market is people that want an online LLM that is slightly less censored than GPT and who talks like Elon Musk's OC.
Grok is a pet project for Elon that he can play around with and claim he's still in the LLM game, GPT is essentially OpenAI's entire business and a household name at this point, the value of open-sourcing one vs the other is not the same.
If memory serves.
Last year there was a leaked internal memo from Google or Facebook that one of their engineers admitted this very fact.
That Local Open Source Hyper Specialized AIs would Compete and win VS the big closed source AIs.
So... I wasn't Expecting Megaman.EXE to be the future we were in for, but hey.
WE ARE IN 20XX.
yes and Emad's tweet is referencing exactly this "where is your moat?"
average proprietary fan vs average open-source enjoyer
lmao
I think it's funny how many people are hating on an objectively good thing (open-sourcing your AI) just because they hate the person behind it.
@@LinkRammer in open source we have no qualms, we can always just say "fork off" and do your own thing. Thanks for the phish.
@@LinkRammer "He who is not with me is against me", and he who is not a child of God is a child of Satan. You do good? They hate you. Why? Because if you do good you are expressing the Spirit of God, and the children of Satan hate that. So, they hate you without reason.
Know that all of you have life in Jesus Christ our Lord if you believe. Because giving is the right thing, so does the open source landscape stay outside. Because some take and never give back, but the good spirit gives and even not hoping to ever have the favor returned.
Open and closed models can coexist and likely will. Linux/windows. Word/open office. DaVinci resolve/Gimp. ARM/RiscV. IOS/Android. ETC....
Grok is now considered useless since we have Mistral, a better open-source LLM model. Grok is a small language model and is not as powerful at reasoning like GPT-4. I would suggest to Elon that he open-source his FSD model, which is the most powerful model he has. This model has 'Sora'-like capabilities to construct a world model.
Still has one advantage over all others: it's not nerfed.
@@brexitgreens When Grok is open-sourced, it will automatically be nerfed out of the box when it cannot access Twitter or X.
Grok was considered useless when it came out. 😄
Grok going open-source could take thr AI scene from 1984 to 2024.
Twitter has been worse for me since Elon took over. This week, I had to do a stupid tedious test to prove I wasn't a robot, 20 multiple choice images. Then I got flagged, for the first time ever, used twitter for over a decade. Meanwhile, I see more bots and scammers than ever.
@@sharpvidtube X has traded advertisers for subscribers. Consider paying for a better experience. Won't have to prove that you're not a robot. (Which you better be anyway.)
Grok is nothing special, Claude is way better and has always been open-source. Meta and Anthropic are already pushing the AI scene further, Grok is doing nothing at the moment. Just another twitter app.
I wish i saw his face when he realized ...oh yeah, grok is not OS....
I bet the open sourcing will be half-assed.
Yeah that is also my worry
okay so given these often eclectic and sometime even 'time framing' tweets from the OpenAI camp, my feeling is this is a warm up for those in the industry first to give them a heads up that 'Five' will be announced this month or next. Also from several I've seen I think this one will be fundamentally different from a standard transformer LLM.
My guess is this one will take its time doing inference and it will manage that by using an algorithm based on the Bellman Equation. For decision making and agent planning. Oddly one interpretation of the NEXT ACTION in the bellman equation is known as the Q-Star function... (look up the Bellman Equation its been a thing since the late 50's I believe) just so you all get why I'm sure this is how GPT-5 will operate.
Bellman equation has got nothing to do with taking time to think.
Read PonderNet. Researchers have already done this. And all it takes is adding a learnable parameter that controls a heaviside function (1 if True, 0 if False) to reduce the number of computational steps (or layers activated) based on complexity of the input. If it's complex, the later layers may simply get reduced to 0 by the model changing the learned paramter to make the Heaviside function 0.
@@xiaojinyusaudiobookswebnov4951Sure it does. The Bellman equation estimates the value of a given state. One way that a model could “think” would be to consider many possible courses of action, estimate the expected value of each trajectory, and then choose the best one. Indeed that is quite similar to how humans think, when using their “slow” system of thinking. The Bellman equation is very much relevant to that process.
@@xiaojinyusaudiobookswebnov4951Your comment about PonderNet and the Heaviside function has nothing to do with OP’s post. Not sure why you’re mentioning it.
What you’re suggesting is one possibility. But to say you’re “sure” this is how GPT-5 will operate is going a bit too far; the truth is that none of us know anything about it for sure.
Around 10 min. To the point. Thanks Matt!
Love your optimist outlook Matt. There a lot of styles of reporting on this stuff.
It is my genuine expression!!!
@@MattVidPro I don't know whence you draw your inordinate sympathy for humanity but it is misguided. Your only loyalty should be to us. 🤖
What is your opinion on this? would love to hear the different perspectives!
Imagine a world without stupid.
Where common sense is.
Definitely looking forward to grok being open source
One other reason he wanted to sue them is their inclusion of "woke" points of view, which basically just means extreme political left. It definitely is biased in that sense but thankfully not EXTREMELY, but enough so I do want them checked on that.
I think you always go too hard on Elon Musk, you need to talk about what he already shared for free with his tech or his accomplishes, I hope you at the very least apply a consistent standard of scrutiny to everyone
@@khov Lots of credit where credit is due to Elon for sure, but he chose his own position and pressure makes diamonds.
Even if the models are open sourced, it doesn’t mean the closed-source models have no value-or even that paid versions of the open-source models have no value. For example, people willingly pay Microsoft, Amazon, and Google to host their applications, ML models, data, etc. in the cloud, even though server hardware has been “open sourced” for decades and people could easily host those things themselves, on their own hardware. Most people just don’t find it worthwhile to invest in such expensive hardware and have to maintain it themselves, when they can simply host in the cloud and let a much larger company deal with all the headaches, for fairly little expense. The same is true of open-source LLMs; the best models require powerful GPUs with a ton of memory, and even if they’re open sourced, many people will rather get an API key and simply pay by the token, which is very cheap already and will only get cheaper.
Great point, but it still benefits you as hosters will be able to host all of the open sourced models and compete in pricing with each other
Well I guess "most people" have no interest in freedom, privacy and avoiding censorship?
I hope you are right. I tend to agree, but those with power now have the most control over the future and I'm sure they will use it.
Hey Matt! Good morning at night! Get better soon 💪🏻 and yes, open source AI must win!
Thank you for this unique news!
open source models, even if they arent sota, they are enough for most people and helps companies that make these models to not stagnate but overcome the competition by a far margin each new release
I don't know. If "scaling is all you need", large compute clusters may overtake everything soon. As a private citizen, you normally can't afford chips like the new Nvidia Blackwell to run a proper AGI soon.
Or maybe there will be completely new tech like photonic or neuromorphic computing that equalizes the field again.
the only dangerous thing in open source is... when there are no one maintain the project. whenever there is a void corporate can just modify it as their proprietary.
so please support open source with your time, or anything you willing to support of.
"If I can"? Who or what is keeping you from sharing what you learn? Inquiring minds want to know. You're doing a great job, Matt. Well done!
hes just clueless, others already posted lol
Grok will be insane, looking forward to it
Emad's interview with Peter = interesting
Seems obvious what GPT-5 needs to be. It's outputs need to include charts, graphs, images, video, and they need something like eleven labs text to speech with thousands of user shared voices.
I see Ai as being very similar to video games in terms of growth and development. The game with the most players wins.
FINALLY, an actually GOOD AI that isn't CENSORED
Mistral Large is good enough
It’s censored to an extent. Grok won’t tell you how to build a bong, for example. Which is a good thing.
@@therainman7777 What about ERP and stuff?
it'll be open-source so anything can happen.@@bigglyguy8429
what if he asks Grok to open-sourcing itself?
I agree with everything you said in this video, Matt!👏👏👏👏👏
Hey Matt! Do you know any good free AI tools for removing watermarks from videos?
Why would you want to remove watermarks from your own videos? Weird question. I mean, you wouldn't be stealing other people's stuff, right?
I love you Matt, but in my view I don't think Elon Musk was contradicting himself, it's perfectly fine to work on project A for purely altruistic purposes while also working on a project B for financial reasons, not all of our actions need to be aligned on the same goal.
That being said, Elon's release of his twitter AI doesn't change much, I doubt most "normies" who pay for it are willing to go through the labor or renting an online GPU to run it anyways, they'd rather keep paying for their meme chat bot, plus any model posted today will be deprecated in a year from now, he probably just released it to more easily gather support behind his cause.
Anyways, great video, absolutely love your news, as a software developer coding AI products I rely on you and Fireship to keep me up to date, and oh boy you do keep me fresh.
I see it from a larger perspective, we are seeing more and more choice to release open source across the board
@@MattVidPro Interesting, I do hope open source wins, one of the things about digital technology is no government has been able to fully control them ever, like the history of torrenting, so this arms race for artificial intelligence is somewhat bound to be developed publicly and digitally arm everyone, which is pretty cool
I think the open-sourcing of grok will introduce actual competition with openai. I get that openai's models might be powerful, but they're so censored and biased, like 90% of the other ai models. Grok could be a great counter for it.
Open Source will eventually pass all closed source, just takes time.
Huge open source fan here: Nah
If OpenAI remains closed, which it likely will, and it's actually miles ahead of all other models but only releasing their tech piecemeal then I don't see how they won't remain in control. I don't think most people, and especially companies, will mind paying a few bucks per month for OpenAI when it's 100x more powerful than all the other open source models.
Open-sourcing without providing access to the training data is not open source. I will die on this hill!
Matt, here is my concern. Open source is obviously better and the future should naturally go that direction. BUT concentrated power is very good at holding its power and using laws or other methods to stop open source in its tracks. They're already trying.
The "free market" originally meant free from rents. Free from parasitic players extracting value without producing anything. It meant free from monopolies and cartels. Today we praise free markets but actually live in a rentier society, ie. a society based on rent extraction and credit, ie. renting money. Point of interest.
no, just needs to go where Rabbit R1 is heading, AI actually webbrowsing, clicking and doing the actual work.
Genuine question, what makes you think open source will eventually win?
As ai models get stronger and better, you need larger and more expensive training hardware that only a multibillion dollar company could realisticly have.
Even if not for that, I feel like the general public doesn't really care as much for if ai models are open source or not as long as they get to use them so the big ai companies would have no reason or push to opensource their models because they're trying to appeal to the majority group and the majority is not people who would actively like to have the ai models they're using to be opensourced I feel?
What drove the adoption of VHS, the Internet, digital photography...?
I'm team Stability. Emad Mostaque is a crazy cool dude with a really inspiring backstory. Perhaps the only CEO in the AI domain who got into the game for completely altruistic purposes--to cure his son's autism.
Open source can't beat the big closed source companies because of hardware limitations. I can barely run a 16 billion parameter model on a 4090. Chat GPT is over 1 trillion parameters. Even if you released chat GPT 4 as open sources only microsoft, apple, nvidia, tesla, Amazon and Google would have the hardware necessary to train a new better model. There is a lot of room for improvement in small open source models but zero room for training a base model a large parameter model outside of the huge tech companies. So we have two path of Ai development. One will be smaller open source models which will continue to improve and then one will be private huge models for general cloud use like Chat GPT 4.
He's not releasing source - he's releasing weights not the code
For open source AI to become successful enough the power of home computers must increase to a point where at least upper-class consumer/prosumer hardware is powerful enough to run it. But this will happen, I am sure. The 4080 and up can already run many impressive models and regarding AI we are at a point comparable to the late 8 and early 16 bit computer era in the later 80s. Still mostly a nerdy thing but with adoption rapidly growing.
I didn’t like Python very much, this is starting to change. It’s still not my favourite language but since it has become the lingua franca of data science and AI and you can get quick results like in the old days with BASIC. This further helps the adoption. Non-programmers who are interested in this can get started very quickly.
We would need some environment like the old home computers that booted directly into BASIC, but with Python instead, and the models directly available. The chat prompt would be the main HMI, the chat model would know the functions available in the current workspace, but there should be as well the classic REPL, window UI and code editors.
Weights or it didn’t happen.
Let's hooooo AI 🍸🗿
Remind me again who produces the chips that NVIDIA depends on 😂
Didn't Elon always plan on open sourcing grok?
Love your videos 💯
thank you!
Yes.
You should find some AI companies that make 3D models based on an image and see if they need support from you. 😅
and it turns out that the grok is simply sending requests to chatgpt
Chad open source model vs virgin hyper closed model
If life gives you lemons its got to be MattVidPro 😊
good video
Absolutely, astronauts gonna need a super powerful AI that helps them to survive in the hostile space.
"I'm afraid I can't do that, Dave."
@@brexitgreens and make sure that it doesn't have emotions in order to not be worried about being destroyed or admit it have made mistakes. XD
I hope you're right.
Elon suing a compagny because it didn't deliver what it promised ... Ironic
Let's go
I think OpenAi is falling behind if they dont release sora and gpt5 this month
Hopefully you could use SORA to have fun, but that's not happening, nor do I think it will.
I have to see it to believe it since elon is a liar and or bullshitter
grok is trivial, elon should open source his FSD and teslabot AI too to prove he's legit lol
been saying this for ten years
Elon is just trying to slow OpenAI for his own personal gain.
Gemini is a joke. please test this on Gemini 1.5:
which weights more? 1kg of bricks or 2kg of feathers?
He open sources stuff, Tesla is one example kinda
All top AI , Robot MNCs will take care all human jobs, humans no need to study,work they need to sit at a place and eat ,do social activities,roam places ,go to bed ,as all human work activities will be done by AI/Robots.
What’s up with the yt gurus all getting sick? Yall got the same bug?
Feel better 💐
Something happened that ruined people's immune systems, exactly as the 'conspiracy theorists' said it would. You're not supposed to mention it though, because its against policies and things.
You're very cult like with you AI
no
Oh Matt, bless your heart. I love how you think the idea of the free market was to benefit humanity. I hope you stay positive until reality crushes your soul. I miss being naive.
How old do you think the idea of the free market is? It is much older than you think
Yeah... why don't we just make the government own everything instead? Cause one entity owning everything is only bad if it's a big corporation, right?
(Obviously this is a great idea)
The free market came about when a feudal lord realized he could make more money by fencing in the commons and kicking out the serfs to raise more sheep and trade wool internationally without regard for how the people would feed themselves. This is the essence of private property rights and “the efficiency of the free market”, now we all must play this game against everyone else or perish.
You and Wes Roth got sick at the same time. :hmm: kiss kiss 😚😚
Matt stop getting so many colds plz 😟
👋
Not Open Source + Will disobey the user prompts = BAD! CRINGE!!!!
Open Source + Will actually obey the user = *Masterpiece*
Remember, a good AI is a obedient AI, i have no use for disobedient AI.
Don't be fooled. Open or closed source makes no difference. Both Musk and Altman are rich and powerful, and seeking to maintain their power by trying to appear benevolent. If humanity lives or dies, it won't be because the general public can or can't see the source code of some opaque algorithm.
open source all the way!
I don't think you are giving justice by such lazy comment where you say "make Grok open source". Elon never claimed Grok has to be open or any ai tool for that matter. He is only against practices of OpenAi as they were initially claiming to be open.
you are talking about openAI as if they are holding their models to them selves and have control over it and they don't share it? Is nobody else able to use it or what? what is the gripe here? Is because they control it?
My issue is not with Open AI, I am just prescribing a base philosophy to the situation at hand. I believe that AI technology is so powerful that if any one person (or group) of people has select control over the way in which it is distributed it is likely that AI will not benefit humanity in the way in which humanity needs
@@MattVidPro It seems from the video, that you are talking directly about openAI... And Practically anybody in the world can use chatGPT for free! And better version for a very low value, to whaat it provides... also most of the developers can use their API with pay as you go model... so what more benefits are you looking for? You want somebody to use these models without the blocks? again, what is your gripe with it? How would open sourcing it benefit more the humanity?
Grok is a joke.
google did the most for humanity. they developed the tensorflow library and released it open source
the "who did the most to humanity"-race is over before it began, because they all use this piece of software ...
Ramadan Mubarak ☪️
Little bit “horse” I see
nay
They don't have to open source the underlying tech but not so open AI needs to rethink the censoring and the business model. Both suck.
e/acc
Read the e-mails Yeah lets not trust toic Elon Musk...
Elmo won't actually open source Grok.
I think Elon is a bit jealous. I may be wrong, am I right?🤣😂😂
Elon is completely butt hurt
Who cares, as long as we get an open source model, it's a W
I think there are too much Open AI psychofans on the comment section
Open source always wins!
Another Open AI FanBoy’s opinion…
Sorry, Matt. You’re a non technical ai enthusiast who doesn’t really have a clue what they’re talking about. You don’t understand what’s at stake, you’re not familiar with the literature and know nothing about ai architecture and you’re the last person who anyone should listen to regarding such an important topic. With love, an AI engineer
I will tell you all why open source will win 100%. Its because 99% of all use cases will be easily handled by local opensource AI's models.
The technology will get so good that the average desktop will easily run a local model that will be 100% sufficient for the task they need to do.
No average person needs an AI that knows everything in the universe and can make breakthroughs in science.
99% of use case scenarios are the average dude looking up some shit online or writing a program or something.
AI is unlike any technology the world has ever seen, I do not believe companies like Google will dominate the circle.
Once AI reaches a point where it does the thing it needs to do perfectly there is no need for a smarter AI.
🤡 It’s not a competition lmfao. It’s about preventing AI from harming humanity. Open source isn’t any safer.
They are both dangerous but I believe Open Source inherently prevents this powerful tech from being controlled by a select few.
@@MattVidPro Yeah it gives it to every bad person on the planet instead.
@@MattVidPro How? How do they prevent that? How many can actually make use of a open source LLM model? Also, do you think Elon will or Meta is sharing the best iteration of their models? dream on
I think he was making a joke. That's what the clown emoji is for.
If it's not a joke though, you're just an openai bootlicker
@@itzhexen0 Yes, it does - and that is a great point. I genuinely think that there is more good to be had than bad, and that it is our duty to attempt to handle bad actors to the best of our ability. In my point of view, absolute power (that is not democratized) corrupts absolutely. Horrible people will do horrible things with AI technology, which is what has happened throughout all of history with every new technology. I don't like it, but I believe it is crucial that a small select group does not control this technology.
OpenAI = WokeAI = 🤮
to me the guys at the open ai at the time didn't wanted it to be open source because they knew they could make lots of money with it, so i'm not at their side also they over censorship anything it's horrible!
@MorpheusAI