For those who think that things will progress slower in 2024 regarding technology, this is not going to be the case.
It appears that while 2023 was the year of LLM (Large Learning Models), 2024 is going to capture the world with LBMs (Large Behavioral Models). This is what is now being used to train robots, a bridge from the digital to physical world.
As we can imagine, this could have a profound impact upon society. This is going to cause a lot of issues, fear, and discussions about how to solve this problem. From my perspective, I think it emphasizes the needs for Web3 even more.
In this article we will go through what is taking place along with how to offset this.
Rise of the Robots
It appears the world revolves around data. Over the past 13 months, the online world is able to play with the chatbots produced by Google, Twitter, Meta, and OpenAi. We can see some impressive stuff resulting. While this is really glorified search, it does get the imagination of many going.
Whatever the next phase of this evolution, we now have something that could dwarf it in impact. Large Behavioral Models (LBMs) could serve up the next breakthroughs. This is not tied to the digital world per se but, rather, robotics. Here is where we could start to see the "rise of the robots".
Over the last few days it appears many robotic and AI experts were caught off guard. We saw a video posted by FigureAI that stunned many.
Here is the short video:
While that might not seem impressive to the average person, this is a huge breakthrough. The key here is the training took about 10 hours to complete. With these systems, once the task is learned, it is never forgotten.
Another vital point is the fact the system learns from each robot. If it had 50 different ones focusing upon tasks, aa simple update means provides network ability. Unlike humans who have to learn individually, these systems operate at the collective level.
Video Training Changing Everything
At the core is the ability to train these systems from video. It is something that Tesla stumbled upon early last year. Since that time, it spread throughout the entire industry. Many companies are feeding labelled video through the neural networks, training them on a variety of tasks.
One company that is moving in a massive way is Google. According to a post on its blog, they are now using Visual Learning Models (VLM) which can give the robots an understanding of its layout. When this is combines with LLM, we are seeing the system better comprehend what humans are seeking. This allows for the training with a lot less code.
Another point is they are finding the system learns by viewing a short series of images. This means scaling is possible.
In extensive real-world evaluations over seven months, the system safely orchestrated as many as 20 robots simultaneously, and up to 52 unique robots in total, in a variety of office buildings, gathering a diverse dataset comprising 77,000 robotic trials across 6,650 unique tasks.
This is laying the foundation for these types of machines to operate in a particular environment, like a home or factory, and execute commands such as "cook me an omelet" or "organize those parts based upon SIC categorization".
The Usual Suspects
It should come as no surprise that we start to see the same companies in the news with these breakthroughs. While there are some start ups, we find that Meta, Google, Amazon, and X are getting the headlines. Also, even though they are not public about things, we have to presume Apple is somewhere in the equation.
What does this mean for the future of humanity? We can easily see how a handful of technology companies are looking to replicate the Internet. By that we are referring to the siloed system we now have. A handful of corporations basically control the digital realm. What happens if these extends to the physical world?
It is evident that these corporations are looking to control the means of production. This is going to remove it from humanity. Labor is estimated to be around $30 trillion annually. The potential exists for this to be diverted into the pockets of these entities.
Production being done by machine is a major component of the age of abundance. It is what is moving our economy away from scarcity. Nevertheless, is it an improvement if this only further empowers a small handful of companies?
Here is where I believe Web3 enters the picture.
New Ownership Model
Each of the companies mentioned are owned by their shareholders. This is the decentralization mechanisms of the 19th and 20th centuries. The rise of the corporation allowed many to participate in globalization and profit from the growth of these entities.
Of course, while this expanded things from the "Robber Baron" days, we are still dealing with a relatively small group of people as compared to the general population. Many are going to call for governments to step in. Unfortunately, this is a history littered with failure, especially as corporations are not starting to rival governments in power.
Web 3.0 is going to transform humanity. The question is how does it unfold. Here is where we are at a crossroads.
To start, if hundreds of millions of jobs are eliminated over a 3-5 year period, how does people sustain themselves? Again, we are going to see many call for government intervention. Does anyone really think being subservient to a government is the best course of action?
As detailed in past articles, the replacement for labor income could be staking. Here is where people have financial stake in networks that are growing in importance. The next for public data should be evident by now. Overlooking this means continually feeding the largest tech companies in the world, something that might not have a great outcome for the majority of the population.
How do we offset a handful of companies controlling the entire means of production? This is a greater concern to me than the Terminator scenario. The likelihood of that is much lower than a small number of corporations taking the control of the digital realm and extending it into the physical.
Once again, we are back to centralization and the danger it poses. Here we are talking about data. Think of all the video training material Google has simply from YouTube. The sick part of it is each day, It is reported that 500 hours of video are uploaded to that platform every minutes. This works out to 720K hours per day.
We keep feeding the beast.
Of course, we follow this by our activity on the Facebook family of applications, X, and Amazon. We are being tracked everywhere. The same is true on our phones and laptops.
While we cannot eliminate all of this, we can start to consciously choose to develop more Web3 databases. At the basic level, this means activity.
As always, nothing operates in a vacuum. People need to wake up to the backdrop of what everything is operating within. Obviously, we cannot turn away from these entities immediately. However, we can start the process.
Recent breakthroughs should get the attention of everyone and for a multitude of reasons.