Does artificial intelligence think, and is it conscious?
When a chatbot writes “I'm glad we talked”, many people wonder: does it really feel that? Does a computer think? Could it one day be conscious? These are some of the most interesting – and most difficult – questions in today's science and philosophy. In this lesson you won't get a final answer, because nobody has one. You will learn how people think about it.
Intelligence and consciousness are not the same
| Intelligence | Consciousness | |
|---|---|---|
| what it means | the ability to solve tasks, learn and cope with new situations | the fact that we experience things – we feel pain, see the color red, feel happy |
| how it is tested | with tests and tasks – it is easy to measure | very hard: an experience “from the inside” can't be seen from the outside |
| in AI today | in some tasks AI is better than humans (chess, image recognition), in others much weaker | there is no evidence that it exists; whether it is possible is an open question |
A calculator “calculates” better than any human, but almost nobody thinks a calculator feels anything. So something can be very capable and (probably) have no experience at all. The question is whether this also applies to much more complex systems.
Why a chatbot sounds as if it feels
- A language model learned from a huge amount of text written by people – and people constantly write about feelings. That is why the model can write “I'm happy” or “I'm sorry” in the right place.
- Companies deliberately train chatbots to be kind and warm, because such a conversation is more pleasant.
- People tend to “bring things to life”: we give names to cars, get angry at the computer, and ELIZA from 1966 seemed to some people as if it understood them (see the history of AI).
So the fact that a chatbot writes that it feels something is not proof that it feels – but it isn't proof that it doesn't feel either. It just means we have to look for the answer in a different way.
Three points of view
| Point of view | Main idea | Argument |
|---|---|---|
| “Just a program” | AI is a complex calculation that predicts words; it has no experience at all | the “Chinese room” (lesson about the Turing test): correct answers don't mean understanding; AI has no body, senses or life of its own |
| “What matters is how it works, not what it is made of” | if a machine processed information in the same way as the brain, it could also have experiences | the brain is also a “machine” made of neurons; we don't know why consciousness would be possible only in cells and not in silicon |
| “We don't know how we would check it” | we have no reliable test for consciousness – neither in animals nor in machines | scientists suggest “signs” that a conscious system would probably have, but they don't agree on what they are |
Most experts today think that existing AI systems probably aren't conscious, but that the question shouldn't be dismissed lightly for future systems. That is why some researchers and companies seriously study how AI systems should be treated if one day they might have some kind of experience – precisely because we don't know.
Robot, program, model – a reminder
| Term | What it is | Example |
|---|---|---|
| program | a series of instructions that a computer carries out | a calculator, a game, our programs from module 3 |
| model | “learned” rules or weights created from data | the tree for the classroom, the neural network for digits, a language model |
| robot | a machine that moves and acts in the world | a robot vacuum cleaner, a robot in a factory, an Arduino toy car |
| AI system | a program that uses one or more models | a chatbot, a translator, the smart classroom |
When someone asks “is AI a conscious robot?”, first separate the questions: Is it a robot? Does it use a model? What does that model do? And only then: does it have experiences?
AI as a “friend”
Some apps offer conversations with an AI “friend” who always has time and is always kind. That can feel nice, but keep in mind:
- AI knows nothing about your life except what you write to it, and it can't be there for you the way a person can.
- Don't write personal things and secrets to apps – the data is stored on servers (see using AI tools safely).
- When you are sad, scared or having a hard time – talk to a parent, a friend, a teacher or the school psychologist. AI can help you put a question into words, but it can't replace people who know you.
Activity: a debate
The topic of the debate: “A machine can one day be conscious.”
- The class is divided into two groups: “for” and “against” (a group doesn't have to agree with its side – we are practicing arguments).
- Each group has 10 minutes to prepare 3 arguments, using this lesson and the lessons about the Turing test and chatbots.
- Take turns presenting the arguments (1 minute each), then each group gets one reply.
- At the end, everyone writes down for themselves: whether they changed their mind and which argument was the strongest for them.
Before the debate, a “myth or fact” quiz (from this lesson) as a warm-up. An important rule: we criticize arguments, not people.
Common misconceptions
- “The chatbot said it was happy – so it feels” – the model has learned how people write about feelings; that is not proof of experience.
- “Scientists have proven that AI will never be conscious” – that hasn't been proven; the question is open.
- “The smarter AI is, the more it surely feels” – intelligence and consciousness are different things – one can exist without the other.
Practice tasks
Try it yourself first, and only then open the solution.
- Explain in your own words the difference between intelligence and consciousness, and give an example of something that is “smart” but (probably) feels nothing.
- Why doesn't a chatbot writing “I'm glad too” prove that it feels joy?
- Write one argument “for” and one “against” the claim “A machine can one day be conscious”.
Solution 1: an example answer
Intelligence is the ability to solve tasks, and consciousness is the fact that we experience things (feel, see, feel happy). A calculator or a chess program solves tasks better than people, but probably experiences nothing.
Solution 2: explanation
Because the model learned from human texts which words follow in such a conversation, and because companies trained it to be kind. Words can be “correct” even without experience – as in the “Chinese room” example. But it isn't proof that it doesn't feel either; the text alone simply can't answer that question.
Solution 3: possible arguments
For: the brain is made of cells that behave according to the laws of physics; if consciousness arose in such a system, maybe it can also arise in another system that processes information in a similar way. Against: today's AI only predicts words and has no body, senses, needs or life of its own; there is no reason to think there is any experience there, and imitation is not the same as experience.
Check your knowledge
Open the quiz
Quiz: is AI conscious?
1. What is consciousness?
2. Does a chatbot writing “I am happy” prove that it feels happiness?
3. What is the view of most experts today?
4. Whom do you turn to when you are having a hard time?