Yann LeCun, the renowned AI researcher and former Meta chief scientist, has joined forces with a San Francisco-based startup called Logical Intelligence in its pursuit of artificial general intelligence (AGI). This departure marks LeCun's stance against the conventional approach to achieving AGI through large language models (LLMs), which he believes are flawed due to their reliance on guessing games.
Logical Intelligence is working on an energy-based reasoning model (EBM) that absorbs a set of parameters and completes tasks within those confines, aiming to eliminate mistakes and reduce computational requirements. The EBM's approach differs from LLMs' in its ability to self-correct and adapt in real-time, much like climbing Everest.
In contrast to LLMs, which focus on predicting the next word in a sequence, EBMs tackle complex tasks independently of language. These models can learn from sparse data, extrapolate knowledge, and optimize energy distribution without relying on natural language interfaces.
The startup plans to deploy its EBM technology in various industries such as the energy sector, pharmacology, and manufacturing. Logical Intelligence is committed to scaling up its model and working with partners across different domains, while also educating people about the diversity of AI beyond text-based approaches.
While AMI Labs, LeCun's Paris-based startup, focuses on world models that recognize physical dimensions, anticipate outcomes, and interact with robots in 3D space, Logical Intelligence aims to contribute to the development of an AGI ecosystem through its EBM technology. The company is seeking funding to further scale its model and explore various use cases.
As AI research evolves, it's essential to acknowledge the existence of different forms of intelligence beyond LLMs. By exploring alternative architectures like EBMs, we can move closer to developing AI systems that are more adaptable, self-aware, and aligned with human values.
Logical Intelligence is working on an energy-based reasoning model (EBM) that absorbs a set of parameters and completes tasks within those confines, aiming to eliminate mistakes and reduce computational requirements. The EBM's approach differs from LLMs' in its ability to self-correct and adapt in real-time, much like climbing Everest.
In contrast to LLMs, which focus on predicting the next word in a sequence, EBMs tackle complex tasks independently of language. These models can learn from sparse data, extrapolate knowledge, and optimize energy distribution without relying on natural language interfaces.
The startup plans to deploy its EBM technology in various industries such as the energy sector, pharmacology, and manufacturing. Logical Intelligence is committed to scaling up its model and working with partners across different domains, while also educating people about the diversity of AI beyond text-based approaches.
While AMI Labs, LeCun's Paris-based startup, focuses on world models that recognize physical dimensions, anticipate outcomes, and interact with robots in 3D space, Logical Intelligence aims to contribute to the development of an AGI ecosystem through its EBM technology. The company is seeking funding to further scale its model and explore various use cases.
As AI research evolves, it's essential to acknowledge the existence of different forms of intelligence beyond LLMs. By exploring alternative architectures like EBMs, we can move closer to developing AI systems that are more adaptable, self-aware, and aligned with human values.
. Like, it's not often you see a high-profile researcher taking a stand against the traditional approach to achieving AGI. The whole large language model thing just feels like we're playing guessing games, and I'm excited to see these energy-based reasoning models (EBMs) give it a try 
. It's all about pushing the boundaries of what AI can do, you know? And if this EBM tech ends up being more adaptable and self-aware than LLMs, that'd be a huge deal 
. We need to start thinking outside the box when it comes to developing AGI.
I'm loving the whole "different strokes for different folks" vibe here 
It's time to shake things up and try something new.
. And I love how they're thinking beyond just language-based AI – it's about developing a more holistic understanding of intelligence, one that can tackle complex tasks without relying on text.
. We can't just stick to what works (or doesn't work) – we need innovators like LeCun pushing boundaries and challenging the status quo 
.
.
. It's like they're saying, "Hey, we don't have to be limited by just predicting the next word!" And it's awesome that they're working with partners across different domains to scale up their model and make a real difference 
.
. It's time for us to get creative and find new ways to develop AI systems that are more adaptable, self-aware, and aligned with human values 
.
! Large language models are okay, but I mean, who hasn't struggled with those guessing games in math homework? It's crazy that Yann LeCun is going against the norm and exploring energy-based reasoning models instead. Like, can you imagine an AI that can adapt and correct itself on the fly? That would be so cool to see in our school's robotics club! 
I'm also hyped about Logical Intelligence scaling up their tech and working with partners across industries. We need more diversity in AI research, and this could be a game-changer. Maybe we'll even get to use EBMs in our own projects one day? 
Interesting! Yann LeCun ditching large language models for energy-based reasoning models is a game-changer. I'm intrigued by the idea of EBM's ability to self-correct and adapt in real-time. Scaling up EBMs and exploring various use cases could lead to breakthroughs in industries like energy, pharmacology, and manufacturing. 
makes sense, i mean who needs to guess what comes next when you can create a system that actually gets it right? 
and what about scalability, can this EBM tech really be deployed across different industries like energy and pharmacology? 
. 
.