Did you know New Paradigm of AI Architectures Will Arrive in the Next Three to Five Year, Meta’s AI Scientist Claims

 

The world of AI is evolving and the next three to five years will see the uprising of a new paradigm of AI architectures, Meta’s leading AI scientist shared.

Yann LeCun claims the new systems will feature capabilities that go above and beyond those seen in AI models today. Similarly, he predicted how the future would be the decade of robotics. This is where advances in the world of AI and robotics could unlock a new class of the most intelligent apps.

In regards to this session called Debating Technology, LeCun shared how the flavor of AI currently is not up to the level that it should be or can be. Well, he did agree that it was useful, but it does tend to fall short in so many aspects.

The current shelf life for the LLM paradigm is short and could last three to five more years. But within that time frame, we will see more advanced models that leave the older versions behind. This means they would no longer have the same limitations of LLM models that we see today.

The drawbacks inside AI can prevent intelligent behavior from uprising, LeCun shared. He similarly provided reasons as to why that might be happening including diminished comprehension of the physical world, deficient memory, limited reasoning, and less capabilities for complex planning.

While LLMs are not capable of this today, they will be equipped with the right technology in the next few years that will rise as a revolutionary wave. They might even need to alter the name as it might not be generative in how we comprehend it today.

This is a similar echo of what was seen and published in the recent past by Meta’s AI scientist. Right now, many of us are experiencing world models that promise to assist AI systems in comprehending the real world. From memory and intuition to common sense and reasoning - there are so many traits that go beyond the current system. It’s all linked to pattern recognition.

Before, LeCun stated that we are a decade away from all of this. Suddenly, the estimate is bringing things closer than ever. But again, no exact time frame can be predicted. As he claims, today’s models can manipulate language but not a person’s thinking. This is why we’re working more on designing models with mental capabilities. They would have common sense instilled into them. 

 Image: WEF