Sign up to access an inspiration library of 30,000+ posts
Ruopeng An's Recent LinkedIn Posts
Ruopeng An
@ruopeng
Endowed Professor & Director, Data Science Center | AI & Social Impact Innovator | Epidemiologist, Policy Analyst, Author & Speaker | Social Entrepreneur
Posts
Ruopeng An
Tech & AI
3mo
Our C+M Ceter AI Conference united scholars from New York University, Yale University, Stanford University, University of Southern California, Princeton University, IBM, Anthropic, and AARP to explore AI’s role in education, social services, mental health and work. During my fireside chat with Yann LeCun, we tackled a common misconception: that eloquence equals intelligence. I asked him whether real intelligence needs language at all.
His answer was unequivocal. “No—intelligence does not need language,” he said. He pointed out that orangutans are highly intelligent yet not social and have no language. Octopuses are even more striking: they never meet their parents, live only a couple of years, and grow incredibly smart within months. They solve puzzles, open jars, camouflage themselves and adapt—without any chatter. In Yann’s view, language is a powerful tool for social species like ours, but it isn’t the source of intelligence.
That resonated with his broader critique of large language models. Elsewhere he’s argued that thought doesn’t require language; language is an expression of thought. An algorithm can churn out fluent paragraphs without ever grasping the world it describes. When we mistake chatty output for comprehension, we risk overestimating what today’s systems can actually do.
What does that mean for us? It means building intelligence requires grounding AI in perception, action, and memory. It means designing systems that learn through interacting with the world—like babies and animals do—rather than just predicting the next word. And it means being cautious about conflating linguistic polish with cognitive depth. In short: talking isn’t thinking. True AI will need to think before it speaks.
52
Ruopeng An
Tech & AI
3mo
Our C+M Ceter AI Conference united scholars from New York University, Yale University, Stanford University, University of Southern California, Princeton University, IBM, Anthropic, and AARP to explore AI’s role in education, social services, mental health and work. During my fireside chat with Yann LeCun, we tackled a common misconception: that eloquence equals intelligence. I asked him whether real intelligence needs language at all.
His answer was unequivocal. “No—intelligence does not need language,” he said. He pointed out that orangutans are highly intelligent yet not social and have no language. Octopuses are even more striking: they never meet their parents, live only a couple of years, and grow incredibly smart within months. They solve puzzles, open jars, camouflage themselves and adapt—without any chatter. In Yann’s view, language is a powerful tool for social species like ours, but it isn’t the source of intelligence.
That resonated with his broader critique of large language models. Elsewhere he’s argued that thought doesn’t require language; language is an expression of thought. An algorithm can churn out fluent paragraphs without ever grasping the world it describes. When we mistake chatty output for comprehension, we risk overestimating what today’s systems can actually do.
What does that mean for us? It means building intelligence requires grounding AI in perception, action, and memory. It means designing systems that learn through interacting with the world—like babies and animals do—rather than just predicting the next word. And it means being cautious about conflating linguistic polish with cognitive depth. In short: talking isn’t thinking. True AI will need to think before it speaks.
52
Ruopeng An
Tech & AI
3mo
Our C+M Ceter AI Conference united scholars from New York University, Yale University, Stanford University, University of Southern California, Princeton University, IBM, Anthropic, and AARP to explore AI’s role in education, social services, mental health and work. During my fireside chat with Yann LeCun, we tackled a common misconception: that eloquence equals intelligence. I asked him whether real intelligence needs language at all.
His answer was unequivocal. “No—intelligence does not need language,” he said. He pointed out that orangutans are highly intelligent yet not social and have no language. Octopuses are even more striking: they never meet their parents, live only a couple of years, and grow incredibly smart within months. They solve puzzles, open jars, camouflage themselves and adapt—without any chatter. In Yann’s view, language is a powerful tool for social species like ours, but it isn’t the source of intelligence.
That resonated with his broader critique of large language models. Elsewhere he’s argued that thought doesn’t require language; language is an expression of thought. An algorithm can churn out fluent paragraphs without ever grasping the world it describes. When we mistake chatty output for comprehension, we risk overestimating what today’s systems can actually do.
What does that mean for us? It means building intelligence requires grounding AI in perception, action, and memory. It means designing systems that learn through interacting with the world—like babies and animals do—rather than just predicting the next word. And it means being cautious about conflating linguistic polish with cognitive depth. In short: talking isn’t thinking. True AI will need to think before it speaks.
52