Anthropic vs Apple: Can LLMs Reason?

About this lesson

Can Large Language Models, like ChatGPT and Claude, actually reason? A recent heated debate between Apple and Anthropic highlights this exact question. Apple researchers argued that current AI models can't truly reason. They tested LLMs with structured logic puzzles and found the models repeatedly fail when complexity rises—claiming they just mimic patterns instead of genuinely understanding. But Anthropic countered, suggesting Apple’s tests were unfair. Anthropic demonstrated that when given more appropriate conditions—like sufficient token limits and solvable problems—LLMs can perform structured reasoning tasks effectively. So, who's right? Well, both, actually. If you define reasoning strictly as human-like logical thinking, current AI struggles. But if reasoning includes structured problem-solving within defined contexts, today's LLMs clearly demonstrate this ability. What do you think? Let me know in the comments! #apple #llm #anthropic #genai #reasoning

Original Description