There are a lot of things mentioned when it comes to Web 3.0. The potential is enormous and I think it is a piece of the foundation of this technological revolution we are going to see.
Often, before something is built, it is hard to envision exactly how it will look. We know that technology often takes paths we did not anticipate. After all, considering the potential of the Internet, who could have guessed that Twitter was going to be one of the most popular use cases.
There is no telling what people will embrace or how they will use things.
Here is the quandary with Web 3.0. We all can forecast what we think will be important, yet that is just guesswork.
For this reason, let's dive even deeper towards the core to find out what purpose Web 3.0 will serve.
Image generated by Ideogram
Web 3.0 has only one goal: democratize data.
This is something that is becoming very clear. Over the last 18 months, the discussion surrounding data has completely changed. The focus in the past was how Web 2.0 companies are selling it. We know the main purpose was advertising.
However, there was something else taking place, unknown to most of us. The training of neural networks to create large language models (LLMs) was in full swing. For this to happen, large amounts of data was required.
Companies went around scraping the Internet. This caused companies such as X and Reddit to throttle their APIs. While they got a lot of pushback in the moment, it became clear very quickly.
OpenAi is enjoying lawsuit after lawsuit about how it broke copyright laws by taking the data without permission. We will see how those cases turn out. It will be an interesting study in case law going forward.
It is mind blowing that the conversation turned to the fact we might not have enough data. This was an idea that was insane just a few years ago.
Here is only one problem. The total amount of data gets a lot of attention. What is equally as important is the fact of accessibility to data.
This is where Web 3.0 comes in.
What is the problem with how things are presently structured?
Only those entities with means (i.e. money) are able to get the data required. Companies like X, Google, and Meta all have their own. Others are able to enter the market, such as OpenAI, and purchase it. We see deals being cut with the likes of Reddit.
The result is a financially exclusionary system.
Success is going to depend even more on the ability to access data. For the moment, compute is still an issue. The difference with this is the laws of information technology are always operating. In a couple years, the cost to train a Llama3 is going to drop by orders of magnitude.
Thus, the processing will be available to many more companies. Data, however, is still an issue.
Humanity is better served when more people have access. A world where only a few major technology companies are the ones behind our artificial intelligence is not something that most would opt for.
The solution starts with having vast amounts of data available. We know the corporate entities are not going to simply give their data away. Why would they? They understand how valuable a commodity it is.
It is easy to fall into the notion that text is key. Certainly, it is important. However, we are seeing a rapid advancement in the type of data that these models are using.
One of the central shifts is the use of video. Here is where Web 3.0 can really make a difference.
Video offers a couple benefits.
To start, the amount being generated is only increasing. This is a trend that started years ago and is not slowing down. Web 3.0 needs more platforms to make an impact but the potential is there.
Another is the fact that video can help these models understand the spatial concept. Text provides a 2D world. With video, we start to more into physical as the systems begin to grasp the idea of space. Distance is an important factor introducing things such as depth. All of this is aided by video.
Of course, with most video, there is also audio. This helps the LLMs.
There is another element here. Generative AI is much bigger than just LLMs. This is simply one component to a larger technology.
Generative AI can be used for disease detection, research, and personalized learning experiences. It is going to enter most fields at some point, expanding the capabilities of all involved.
All of this comes back to data. Who is going to have access to a lot of this? Obviously, there is much out there that is not available publicly because it is personal. That will benefits the organizations who have it (and most likely sell it to an OpenAI).
That said, there is a lot that can be produced publicly. Here is the roll that Web 3.0 offers.
By democratizing data, we can provide access to anyone who wants to build something that trains on it. This will not be closed like we are seeing now.
Without the data, even if the model is open source, we are simply running into another wall.
The goal of Web 3.0 is to democratize data.
