The Hated One
patreon
Mark Zuckerberg Is Unironically Based Now
🕑 Added 2024-05-24 21:38:25 +0000 UTCZUCK IS BASED
I am not a fan of Mark Zuckerberg. The evil illuminati triangle of his social media apps is a poison for your mental health, it’s riddled with manipulation and it’s a privacy nightmare. But for once, Zuck did something right. Here, let me show you.
I am running Large Language Model Application, or Llama, directly on my laptop that you’ve seen me editing my videos on. This is an AI developed and owned by Meta, Facebook’s parent company. [1]
But here is the catch. I don’t have to sign up or pay a subscription fee. I don’t need an account or be connected to the cloud. In fact, I can completely disconnect my Internet access and Llama will still run.
You can’t do this with ChatGPT, Copilot or Google Gemini. But Llama, is very much like those three. So what’s the difference?
The difference is in this document. This is the official license you agree to when you use Llama 3, the latest version of the model. [2]
And this is the most important line in the whole document. “You are granted a non-exclusive, worldwide, non-transferable and royalty-free limited license under Meta’s intellectual property or other rights owned by Meta embodied in the Llama Materials to use, reproduce, distribute, copy, create derivative works of, and make modifications to the Llama Materials.”
That’s a long way of saying Llama is open source. And that’s a game changer. [3] [4]
I am the Hated One, and this is why you should care about open source AI. Please subscribe and watch another video of mine. If you don’t, fine no problem.
Corporate takeover
The big tech is dumping billions into AI. So much so that even their green pledges go out the window once they start evaporating city-scales of water just to train their AI. The hype is inescapable. Every new product is now trying to shove AI down your throat no matter how useful or useless it is. [5] [6]
But despite what it might look like on the surface, you don’t really have that many choices. The biggest players in the AI industry are the big tech incumbents. Microsoft, Google, Amazon, Apple… these are dinosaurs overpowered by the infinite money glitch granted to them via their trillion-dollar market evaluation. And they are not looking to share their top spot. [7] [8]
So all of their AI products are proprietary. Which means whatever they develop, only they get to distribute and modify. They set the rules for their AI but the public doesn’t get to see them and you will never have a say in them. If you want to use ChatGPT or Gemini, you have to sign up with an account and pay or be greatly limited in your interactions. It’s an enshitified version of a technology that hasn’t yet had a chance to get good first. It’s like pre-shitification. [9] [10]
Meta’s proposition
But then there is Meta. The shining beacon of hope and freedom. Yeah, not really. But Zuck decided to go the opposite direction. As of 2023, Meta’s Llama models are open source, which means anyone can access them for free and what’s most important is that they can modify and redistribute their own versions of Llama without Meta’s explicit permission. [11]
This allows for tremendous freedoms. It enables anyone in the world to take Llama and customize it to their own needs, fine-tune it, train it on their own data sets. You don’t need a data center to do this. It’s doable on consumer-grade hardware. [12]
The barrier to entry is as low as a laptop-grade processor. Or if you wanna go really crazy, you can get two graphics cards and run the larger, more powerful versions, still without connecting to the cloud. [13]
Now I am priced out of the latter market, patreon.com/thehatedone.
But I can still enjoy the lighter version of Llama 3 on my 3-year-old editing laptop, which means you can probably run Llama on your own computer too. [Unless… “what’s a computer”]
The conflict
But now, Zuck’s open source AI is in direct conflict with the big tech in two big ways.
The first is, open source AI is just cheaper. It’s free, essentially. Why would anyone want to pay to interact with AI when they can use an open source model with a very similar level of quality? [14]
The second battle is return on investment. Even if proprietary models take over the market, their only return on investment is money and user data. They don’t get anything else and they have to bear all of the research and development costs. [15]
For open source models, you are still gonna have customers who will pay for a cloud-based AI, but you will also benefit from all the contributions and improvements to the source code from the community. Open source code is everyone’s code. There is already a global community of researchers and hobbists fine-tuning, tweaking and enhancing open source AI models for everyone. And just like Meta gives away their AI, they can also take all of those contributions for themselves and build more improvements. Everybody wins.
In contrast, the rest of the big tech is gambling on the believe there will always be enough paying customers. And freely available open source AI is an obstacle to that strategy. So where they can’t win in free market, they turn to the government.
Regulatory capture
In my previous video, I talked extensively about how the big tech is fear mongering about AI as an extinction threat. They do this because they want to convince regulators to restrict who can and cannot develop AI. [16 – 19]
They want to do this by imposing a strict licensing regime, where only government-authorized models would be legal and the source code will be confidential. They want to achieve this by holding AI developers liable for abuse from their users. This would criminalize open source development, because no small team could afford litigation when some jerk decides to generate questionable looking horse pictures in their tax folder. [20] [7]
But nonetheless, bills and orders are being proposed locally and nationally in the US, UK and EU. Billionaire lobby with philanthropic networks of organization are paying staffers and fellows that help legislators draft their laws. [21]
This strategy is working. The media always rushes to pick up the latest sensation about how “industry leaders” are warning of AI extinction. There are comparisons of AI to nukes, which only brings us to one conclusion – you cannot allow anyone to build their own AI.
But, let’s settle this question once and for all. Hey, Llama! Are you an extinction threat? See, nothing to worry about!
This extinction threat notion has no basis in scientific evidence. It’s one of those “non-zero-chance” scenarios that could make a great story line in a movie. But it only serves to cement the position of the big tech who was the first on the AI scene and wants to be the only one on the scene. [17] [10]
Here is the thing. If the AI is such a severe threat, should we trust a handful of emotionally dubious Silicon Valley tech bros to run it at their discretion? Proprietary AI isn’t any less vulnerable to prompt-injection attacks, which is when a user attempts to bypass safety restrictions of the model. If they truly believed AI is so dangerous, they wouldn’t spend all of their resources developing it. [17], [22]
[23]
Open source is the right answer. If AI is a threat, only transparency and public access can increase our chances to notice something’s wrong. Clandestine source code is never the answer. Security through obscurity doesn’t work. The more eyes on the code, the higher the chance good actors will notice malicious activity. [9]
Zuck is right
Zuck is right on this. And it’s not just a publicity stunt for him. He is passionately arguing that AI should be open source.
And it’s not like his reasons are messianic all of a sudden. Open source allows for non-exclusivity on the market, which allows Meta’s AI products to disseminate faster and potentially capture a greater market share. This is how Android has taken over the global mobile market. Android is also open source, it’s developed and owned by Google. But anyone can take it and modify it however they please. It’s only when they want to use Google Play products on top of Android, that’s when things get proprietary again. This is how Android got cheap and surpassed Apple’s iOS even though it used to be a lot less mature in the beginning.
Zuck also has plenty of shitty experience with dealing with proprietary APIs, especially from Apple. You’ve heard about these feuds between Facebook and Apple. In the media, they are always framed through Apple’s marketing lens, as a privacy issue. But in reality, it’s about Apple exercising its monopolistic control over the iPhone’s walled garden, and not allowing developers or users to act outside of Apple’s strict conditions.
That’s the same kind of control they want to exercise with proprietary AI. Your only chance of interacting with it is through an API that is going to greatly limit your experience and everything you do will be monitored and monetized.
Open source is da wei
Don’t let them take this away from you. The big tech is lying to you and Mark Zuckerberg out of all people got it right. Not perfect, but still in the right direction. It’d be best if all of Meta’s product were open source, or completely free software under a copy-left license. But I don’t think that’s ever happening. [24]
The big tech is absolutely trying to steal AI from you and they are using fears and hypes to get their way. They don’t deserve to be the sole arbiters of what technology we can and can’t have.
Llama is far from the only open source AI. There is plenty more and if you are interested, I can make more content about open source AI in the future.
I believe that technology should be as free and openly accessible as possible. It’s a belief that I extent to all software as well as medicine, science and education.
We should encourage and support good decisions like Mark Zuckerberg made with Llama. Don’t support proprietary AI with your money or usage data. Use Llama, or some other open source AI, like Mistral, instead. You will go to heaven if you do. You’ll get a first class ticket to heaven if you also support me on Patreon. You can trust me on this. It’s prophecy. Thank you.
Sources
[1] https://ai.meta.com/blog/meta-llama-3/fbnoscript=1
[2] https://llama.meta.com/llama3/license/
[3] https://en.wikipedia.org/wiki/Free_and_open-source_software
[4] https://www.semianalysis.com/p/google-we-have-no-moat-and-neither
[5] https://www.datacenterdynamics.com/en/analysis/data-center-water-usage-remains-hidden/
[6] https://arxiv.org/pdf/2304.03271.pdf
[7] https://www.ft.com/content/2dc07f9e-d2a9-4d98-b746-b051f9352be3
[8] https://www.politico.com/news/2023/12/03/congress-ai-fellows-tech-companies-00129701
[9] https://open.mozilla.org/letter/
[11] https://www.vox.com/future-perfect/23817060/meta-open-source-ai-mark-zuckerberg-facebook-llama2
[13] https://anakin.ai/blog/how-to-run-llama-3-locally/
[14] https://www.semianalysis.com/p/google-we-have-no-moat-and-neither
[15] https://fortune.com/2023/05/05/google-engineer-says-no-moat-artificial-intelligence-warren-buffett/
[16] https://www.politico.com/news/2023/10/13/open-philanthropy-funding-ai-policy-00121362
[18] https://www.politico.com/news/2023/12/03/congress-ai-fellows-tech-companies-00129701
[19] https://www.technologyreview.com/2023/09/25/1080104/inside-congresss-first-ai-insight-forum/
[22] https://www.wired.com/story/chatgpt-prompt-injection-attack-security/
[23] https://www.wired.com/story/generative-ai-prompt-injection-hacking/
Comments
The Hated One
I think you can follow this one: https://anakin.ai/blog/ollama-llama3/ It's basically very simple: 1) Install Ollama 2) Head to Ollama website for library of models 3) Select a model you want and run the commend (usually "ollama run /chosen model/" And this should download the model for you and you can interact with it in teh command line.
The Hated One
Have you tried a different server? I've been using Patreon over VPN exclusively with no issues. Maybe it was a card fraud prevention issue.
Big Machiavellia
I just tried joining this patreon with a monthly contribution and my account was immediately banned by patreon after entering my payment details. The only thing I can think of was that I was on vpn. I went off the vpn and tried again and it worked. Just thought you should know. I'm not even going to bother to get the original account unbanned.
Luis
Is there a guide you used to install Llama AI?
