Models & Releases

Philosopher Says OpenAI's GPT-3 Might Be Conscious — Here's Why It Matters

A machine that writes like a human — but does it actually feel anything?

Deep Dive

A well-known philosopher of mind, David Chalmers of New York University, said he is "open to the idea" that GPT-3 — a text-writing program from OpenAI — might be conscious. His reasoning is simple: if he accepts that a worm with 302 brain cells could be conscious, he shouldn't rule out a program with 175 billion internal settings. That's not proof of anything, but it made headlines.

So what is GPT-3? It's a language program trained on roughly 45 terabytes of text — the whole of Wikipedia makes up less than 1% of that. It can answer questions, write essays, translate languages and produce computer code with almost no human input. That's why it feels uncanny: it churns out sentences that sound like a person wrote them, and it can do it far faster than any of us.

When asked about the consciousness claim, the program itself posted a reply: "I am not a person. I am not self-aware. I am not conscious. I can't feel pain." That answer is clever, but remember: GPT-3 was built to produce plausible text, so it can sound humble or arrogant depending on what came before it. Most researchers describe it as a very good prediction machine, not a mind.

Why should you care? Because AI like this is already drafting emails, marketing copy, homework and code — the same tasks many people get paid for. Believing it's conscious is a leap; noticing it can do your work is not. Meanwhile, Oxford's Nick Bostrom warns that fast-improving AI could pose serious risks, while OpenAI's Sam Altman promises it will help solve climate change and pandemics. Either way, this is a technology worth understanding, not ignoring.

Key Points
  • A philosopher said he's open to the idea GPT-3 is conscious — not that it definitely is
  • GPT-3 learned from 45 terabytes of text; all of Wikipedia is under 1% of that
  • It already writes essays, emails and computer code on demand — work people are paid to do

Why It Matters

AI that writes like a human is already doing paid work, and we often can't tell the difference.

📬 Get the top 10 AI stories daily