In this episode, Pascal interviews Devi, an AI research director at Meta, about the history of AI at Meta and their unique approach to developing and using AI. They discuss topics such as the difference between AI and ML, deep learning and LLMs, Meta's open approach to AI, and exciting future AI developments. They also mention recent product launches, including Meta AI agents and personalized AI assistants.
Meta invests in long-term explanatory research and aims to create an ecosystem of diverse AI agents that can interact with users.
Meta strongly believes in open science and open-source initiatives, aligning with their goal of giving more people access to technology and creating a beneficial ecosystem.
Deep dives
The Distinction Between ML and AI
There is not much distinction between machine learning (ML) and artificial intelligence (AI) in practice. ML focuses on learning from past experiences and making predictions on future data, while AI traditionally involved hand-coding logical rules for machines to make intelligent decisions. However, modern AI tools heavily rely on ML techniques, and the terms ML and AI are often used interchangeably.
The History of AI at Meta
AI has a long history that predates the recent hype around generative AI. Meta has been investing in AI research for over a decade, with a focus on long-term explanatory research. The recent excitement around generative AI, like ChatGPT, is reminiscent of similar cycles of excitement in the past, such as the rise of deep learning in 2013. Meta has a tradition of pushing research innovations and bringing them into various products.
The Vision of Agents and Foundation Models
Meta aims to create an ecosystem of diverse AI agents with different personalities and interests that can interact with users. This is different from other companies that focus on business-to-business use cases. Additionally, Meta envisions foundation models that are multimodal and can generate content across various modalities like language, images, and video. The goal is to create a vibrant ecosystem of agents and allow users to create their own agents for different use cases.
Meta's Approach to AI and Open Science
Meta strongly believes in open science and open-source initiatives as a way to advance AI research. The fundamental AI research lab at Meta has a philosophy of open science, sharing results, and building upon the work of the scientific community. Meta's approach aligns with its broader product philosophy of giving more people access to technology and creating an ecosystem that benefits everyone.
For this last episode of 2024, Pascal talks with Devi, an AI research director at Meta. They talk about the history of AI at Meta, some of the basic terms, how Meta's approach to developing and using AI differs notably from other companies and what the future has in store.