<html>
<center><h2>Robots learn with videos.</h2></center>
<p><br></p>
<center>
https://cdn.steemitimages.com/DQmZBjrNGZwJat4y7e4m9TtcMjsEzPeX3mcKkLCWGf75o94/imagen.png <br />
</center>
<a href="https://www.youtube.com/watch?v=G5YH869B-X0">Image source: </a>
<p><br></p>
<div class="text-justify">
Meta's new AI that teaches by observing.
<p><br></p>
Imagine artificial intelligence that learns just by observing, the world without the need for explanations, without labels, simply by seeing how things happen, I know, it seems like something from science fiction, but it is already happening in META, the same company that owns Facebook and Instagram has just announced V-JEPA 2, an AI model that is completely changing the way robots interpret and interact with the environment around them.
<p><br></p>
I'm talking about machines that begin to understand physics, the logic of cause and effect, even to anticipate what is going to happen next just by watching raw videos, at least that's what Mark Zuckerberg's company claims.
</div>
<p><br></p>
<center>
https://www.youtube.com/watch?v=G5YH869B-X0 <br />
</center>
<p><br></p>
<div class="text-justify">
Robots with predictive artificial intelligence
<p><br /></p>
V-JEPA 2 is a direct evolution of the first model of mere in this line, but now with a leap in its capacity, the big difference does not require a human being to label each part of the video saying "This is a ball", "that is a table" or "that object is going to fall" instead V-JEPA 2 analyzes hours and hours of video identifying on its own the patterns that govern the physical world, that includes things like gravity, collisions, movement of objects and even the concept of permanence, that basic idea that one object hidden behind another still exists.
<p><br /></p>
All this happens in a space called “Latent”, a kind of internal representation that the AI creates to predict the future based on the information it itself has learned and the theory has already become practical, laboratory robbers trained with the V-JEPA 2 are managing to perform tasks that were previously almost impossible for machines that did not have specialized sensors, nor detailed programming, they are taking on strangers by positioning random items and places and even navigating chaotic environments, all only with visual learning.
<p><br /></p>
This means that instead of reprogramming the robots every time the environment changes, it simply adapts its behavior based on what it has already seen and what it expects to happen next, it is like in quotes giving the machine an instinct, something it uses to act more safely and efficiently and the impact of this technology could be giants, delivery thefts, autonomous vehicles drones, all of them could benefit from this new ability to predict the future based on visual experience.
<p><br /></p>
Imagine an autonomous car that, upon seeing a child running towards the street, even before the sensors detect an imminent collision, manages to understand in fractions of a second that this will end in something bad. This predictive intelligence can redefine safety in various sectors from logistics to domestic assistance and not only Meta is in this race, Google with Deep Mind is already working on the Genie model and Wordlabs of the renowned scientist Fefi Lee is also investing heavily in this new generation of physical AI, but of course along with all this evolution there are also questions.
<p><br /></p>
To what extent is it safe to let machines make decisions based only on their own interpretations of the physical world, will be that at some point these robberies could begin to make unforeseen decisions based on correlations that they themselves created, but that humans cannot track and the most provocative thing will be that we are seeing the first real steps of an AI that begins to develop rudimentary forms of awareness of the environment, the ethical debate is just beginning.
</div>
<p><br></p>
<p><br></p>
<a href="https://about.fb.com/news/2025/06/our-new-model-helps-ai-think-before-it-acts/">References 1</a>
<p><br /></p>
<center>
<h2> <div class="phishy">
Follow my publications with the latest in artificial intelligence, robotics and technology. <br />
If you like to read about science, health and how to improve your life with science, I invite you to go to the previous publications. <br />
You want to win, play <a href="https://hari-raid.h4lab.com/?ref=necho41"> HARRY-RAID </a> <br />
</div> </h2>
</center>
<p><br /></p>
</html>
hiveblocks