The potential of LeoAI is just getting better.
What I am going to write about in this article stems from a discussion that I saw Jensen Huang, CEO of NVIDIA, having with Mark Zuckerberg. He touched upon something that, when he described it, he was talking about LeoAI.
This is how far advanced some of this stuff is.
How would you like to have your own chatbot? What do you think about having one that is personal to you?
Huang talked about having this based upon all the articles he wrote and interviews he did. The idea is to fill the database with information that could be used to basically create a digital version of him.
This is exactly what LeoAI is going to do.
Image generated by Ideogram
What does all this mean?
To understand, let's dive into a few of the technical details.
The essence of this is called retrieval augmented generation. Basically, we are seeing the database set up in such a way where it is dealing with vectorized data. This is a way of connecting it so the model has the ability to pull in a manner that is concise to the individual.
Hence, whatever is input by the individual will be in the database. If, over time, all the data is input, such as articles and videos, the ability for the model to learn about that individual starts to emerge.
This is where the idea of a personal chatbot trained on you comes about.
Naturally, this is true for anyone else who does this. The model is processing all data placed on the blockchain. What stands out, however, is the fact that this is all on one system. That is not the case when articles are posted on Medium, Wordpress, or Facebook activity.
The specialization of this model is what brings it to the personal level.
Over the next few years, we are going to come to know the phrase "digital twin".
For the moment, this is applicable to the manufacturing world where the idea of holograms or digital versions of a factory or product are created. This can be helpful in the layout of equipment even before the plant is built.
In this scenario, digital twin carries a different meaning. It is basically one's digital life.
We leave a footprint wherever we go. This is not surprising to anyone. The issue is we are on a variety of platforms, all gathering data in a proprietary manner.
With this version of ourselves, since any activity that is placed on the blockchain is fed into the model. This means LeoAI is developing an ongoing digital version of ourselves. It is tied to any posts or comments that we make. Our votes along with reblogs are all available for processing.
LeoAI is always watching.
Using the database vectoring, this enables the model to design a framework from the data. This is where we begin to emerge in digital form.
Of course, this is very select in what it can process. Since it is not tied to anything outside what is posted here, it has no way to implementing that information. For example, Netflix has an idea of your movie tastes based upon your viewing habits. LeoAI will not have any clue unless you write articles about that.
Another factor to consider is when you interact with the chatbot itself.
The data posted on chain is not the only source of information. These chatbots "learn" from the direct interaction that individuals have. In other words, when you have "conversations" with them, they are processing that data also.
Like most of these machine learning systems, the more engagement means better results. Those who spend little time feeding in data are going to find they have little output. This is true if the majority of one's online life is spent on traditional social media. Elon Musk is willing to feed your data to Grok but that isn't likely to do much in terms of developing something useful.
LeoAI provides something specific. Here is where the "small language model" in front of the LLM enters. The training on general Internet data is done by the model builder (Meta, Google, etc...) whereas the SLM comes from the Hive Database.
This is then fed into LeoAI, providing a profile unique to each individual (account).
We will have to see how this all fits together when it goes live. What we do know is all activity is being fed into the model.
