AI’nt That Easy #31: LCM vs LLM

In late 2022, the release of ChatGPT brought AI into the global spotlight, and with it, the rise of Large Language Models (LLMs). These models became synonymous with AI’s transformative potential, redefining how we communicate, work, and create. For many, LLMs felt like the pinnacle of innovation.

But now, a question lingers: Have LLMs reached their peak? Enter Meta’s Large Concept Models (LCMs), a groundbreaking evolution in AI that shifts the focus from word-level predictions to reasoning at the level of ideas.

Could LCMs be the key to the next chapter of AI’s story? Let’s dive in.

What Are Large Concept Models (LCMs)?

To grasp the significance of LCMs, we need to compare them to the reigning champions of AI — LLMs.

LLMs, like OpenAI’s GPT or Google’s Bard, operate by predicting the next token (a word or subword) based on a given context. It’s like completing a crossword puzzle one clue at a time. LCMs, however, shift this paradigm.

Instead of working token by token, they focus on concepts — entire ideas, sentences, or thoughts.