AI War The Winner You Don't See Coming
If you're anything like me, you love a good rivalry. I'm talking Nike versus Reebok, Pepsi versus Coke, Ferrari versus Lamborghini.
But now there's a new showdown brewing that's got the whole world buzzing. Two trillion dollar companies are going at it full, fast, and furious. I'm talking about Google and Microsoft.
As these two heavyweights duke it out for AI supremacy, there's only one question for people like you and me. Who is going to win?
Let's figure it out together.
Well, this AI drama might seem like it just started, but it's been going on for a while. And I can promise you this, it's going to be going on for much longer from here. And perhaps it might even go definitely into the future.
But if you ask me who's winning this war, I would say neither of them. There's something else that's going to take the crown in this AI battle for supremacy.
Before we talk about that, let's talk about a few other important things.
First, how did this whole AI situation get started? Well, in truth, it really is Google that made this AI revolution begin. They have been at the forefront of AI for years now, for almost a decade, basically. They have invested heavily in cutting edge research firms like DeepMind.
That team put together something called AlphaGo. In 2016, AlphaGo made headlines when it defeated the world's top ranked Go player, Lee Seidol, in a historic match. This was a significant milestone for AI because for decades, people thought AI could never beat a human in Go, unlike in chess.
Bad news for all the Go players out there. It certainly can, and it did.
After that, Google was also responsible for the primary and fundamental innovation that all of the AI world that we know currently is built on, an idea called transformers. In 2017, the Google Brain team released a paper called Attention is All You Need, which described this concept of how to use these models for deep learning.
The Google Brain team actually invented these types of deep learning models way back in 2017. Today, nearly every large language model you've heard of, like ChatGPT, relies on transformer technology to give you the output that it does. In fact, that's what the T in GPT actually stands for, is transformers.
But being first does not mean that you are gonna be best. Despite inventing this technology six years ago, Google was not able to meaningfully parlay that into a large consumer success.
To many people in the outside world, for years, it didn't look like any major progress was happening in AI at all, until a little company called OpenAI came to town.
Initially founded as a nonprofit by a bunch of tech legends like Elon Musk, Peter Thiel, and Sam Altman, OpenAI became a for-profit company in 2019. That was the year that they caught Microsoft's attention, and Microsoft invested more than $1 billion into the company.
Fast forward a few years, and OpenAI releases its GPT model that is gonna light the world on fire. It's called ChatGPT. Surely you've heard of it by now, but here's what you might not know. In its first five days, ChatGPT got a million users on board, which is crazy. For context, it took Instagram 2.5 months to get a million users. It took Netflix 3.5 years to get a million users, and ChatGPT did it in five days.
ChatGPT is the fastest growing product in human history. It has more than 100 million users already now. That is why in January of 2023, Microsoft decided to up its level of investment. It invested a further $10 billion into OpenAI in order to make sure that this technology could power its whole suite of products. I'm talking Office, I'm talking Bing search, and many other things.
But big daddy Google was not gonna go down without a fight. That's why they released Google Bard, their version of a GPT chatbot. It is based on that same transformer technology that they invented so many years ago, and it is freely available in 180 countries right now with no waiting list.
Google has nine products with more than a billion users. They have a significant advantage in distribution versus OpenAI and Microsoft. There's all sorts of products, whether it's YouTube or Gmail, that they can integrate this Google Bard technology into to improve the product and start to win this AI race.
Google has the distribution, they have the money, and they have the need to make this happen. As Google CEO Sundar Pichai has said himself, AI is the most profound technology humanity is working on, more profound than fire or electricity or anything that we've done in the past. That is the voice of a man who knows he cannot afford to lose this fight.
So who is gonna win this battle? Big daddy Google or Microsoft and OpenAI?
Well, if you ask me, it's not Google and it's not Microsoft. In fact, it's not even a company or an individual person. It is open source technology.
If you're not aware of the open source community, let me quickly define it and how it works. Open source is a way of developing software which is fully open and transparent and collaborative in nature. Literally what happens is developers all over the world take their code and rather than keeping it a secret, they release it for free to everyone in the world to see and iterate and evolve.
The best example of open source in the past that you might've heard of is Linux. Linux is the second most used operating system in the world and it was developed for free by a distributed group of global strangers, which is pretty crazy when you think about it.
Another example is Apache Web Server, which maybe you've never heard of, but it powers nearly 32% of all websites that exist. It also was built for free by a bunch of different internet strangers coming together to build something larger than themselves.
When it comes to AI, open source developers can work on models and other applications together asynchronously from across the world, building together towards something greater. Open source AI developers can collaborate from anywhere in the world on any type of project with no red tape or approval process or anything like that.
My bet is that these open source developers are gonna be able to build systems that can out-compete Google and Microsoft.
Now, I know it can sound crazy, right? How are a bunch of random internet weirdos gonna be able to beat trillion dollar companies like Google and Microsoft who can hire the best talent in the world?
But I promise you, I'm not the only one that sees this coming. A leaked document out of Google claims that there is no moat for companies like Google and Microsoft when it comes to AI. If you don't know, a moat is just a competitive advantage of some kind. In order to build a sustainable and long lasting business, you need to have some kind of moat.
Netflix has the moat where they own their IP or Disney has their moat where they own IP. When you don't have a moat, there's nothing unique that you can do that your competitors cannot do, which means you are gonna have a pretty bad time of things.
This leaked document has caused massive concern throughout Google and outside in the tech industry as people have come to grips with the fact that there might not be any long-term advantage. And you can see the lack of this moat in the world right now. As I speak, you can run a large language model on a Pixel 6 smartphone. Anyone in the world can access these things.
You can personalize your AI in a single afternoon on your own writing. You can even train an image model on hundreds of open source data sets right now.
So what makes open source so powerful? Why am I so bullish on open source? Well, there are two things. It's collaboration and it's experimentation.
First, let's talk about collaboration. There are a lot of smart people who work at Google and Microsoft. There's no question about that. Some of the smartest people in the world work at those two companies, but just by how numbers work, there are way more smart people that do not work at Google and Microsoft.
And a lot of those smart people are interested in AI right now. All of those smart people can work on the same projects at the same time without having to get approval or permission because the code is open source. With all of these unique skills and perspectives and just honestly the sheer amount of man hours that you can throw at the problem, you get tons of rapid innovation.
Here's a great example of what I mean. Meta, formerly Facebook, their model was called Llama. That was their version of GPT-4. It was leaked onto the internet such that anyone could all of a sudden use the exact same model that Meta had spent all this time and money building up and making better.
At the time it was leaked, Llama was only about 68% as accurate as ChatGPT. But within three weeks, the open source community was able to do what Meta was not able to do in months. They got the accuracy all the way up to 92%, which is just about as good as Google Bard. They did this by creating and evolving a model called Vicuna 13b based on Meta's initial model.
Because all of these different people around the world were able to collaborate on this shared and open model, really exciting things could happen that never would have been possible with just the small group of close-minded people who work at Meta.
Again, this is not a shot at anyone who works at Meta or at the company. It is very important for people to be able to coordinate at these types of centralized companies. But open source allows for a level and degree of collaboration that is simply not possible when you have middle managers and daily standup meetings and all of the rest.
The craziest part of all of this, by the way, the total cost for training Vicuna, that open source model, was $300. There are engineers at Google that make more than $300 an hour. That is the power of open source.
The second reason that open source is gonna end up winning here is experimentation. In order to innovate, you need to run a lot of small experiments. It's the only way forward. Only by running these experiments can you learn enough that you can start to develop better and better products.
The open source community, again, because there are so many people out there without any limitations, is able to experiment way faster than centralized companies like Google and Microsoft.
Here's another example. There is an open source image generation AI called Stable Diffusion. Maybe you've heard of it. When it first came out, it was substantially worse than Dall E, which was the model created by OpenAI, the centralized company.
However, because it was open source, it very rapidly was edited, remixed, evolved by developers all over the world. Right now, you can go to Hugging Face and find hundreds of different types of experiments on the Stable Diffusion model, each one customized for particular use cases.
And as a result of all this experimentation, making it better at all these individual little things, now Stable Diffusion is actually way better than Dall E, despite being open source. It is so good that it is used by centralized companies like MidJourney, who are closed source, to develop their own models.
That is the power of open source. When you experiment all over the place, you end up learning much more. And as a community, if you can come together and transfer those learnings into new products, you're able to iterate and innovate much faster than your competition.
Honestly, the proof is in the pudding. Just go right now, you can pause this video, go use Stable Diffusion and go use Dall E and tell me which one is better. I can promise you it is the open source one.
Now, what's the final verdict? The paradigm shift towards a decentralized and open model of AI development through open source is a game changer and it is here to stay. Who wouldn't want a world where there are no secrets to artificial intelligence, where it is cheap, customizable, freely available to anyone, anywhere in the world at any time, and does not give more power to these giga companies that it exists.
Well, let me slow you down a little bit here because open source is actually a double-edged sword in this case. Normally, there's no problems with open source because if a bad guy or a bad government gets access to code of, let's say, a video game, they can't really do anything too bad with it, right?
But in my opinion, AI is more like nuclear energy than it is like normal software. By which I mean the power of AI to transform society, to cause harm, and to cause good is much, much larger than any individual piece of software ever has been before. And in the same way that you would not wanna open source the plans to make nuclear reactors, because then any bad guy or bad government can create their own nuclear missiles, you probably don't want tons of open source innovation in AI either, right?
It can lead to bad people developing bad things that regulation cannot stop because they live on the internet perpetually. If hearing that scares you, good. You should be a little bit scared. We are in uncharted waters. Never before has technology so powerful been available to so many people with such little restrictions.
The truth is this, there is no putting the genie back in the bottle. There is no way to suddenly take the llama model out of the internet. Instead, we are now going to live in a world where open source development happens at a pace that is mind-boggling.
Click Here to visit my channel and subscribe for the latest videos and stay current on all new uploads. Join the community - Subscribe Today!
Like This? You'll probably like my newsletter too
Every week, I send out a newsletter with thought provoking articles, videos, and quotes to make your weekend more interesting. It's free and takes 5 minutes to read - sign up below.