Humans would have the ideas. AI's would be the builders.

Regarding the origin of ideas, AI can invent ideas. I am not saying it cannot. I’m saying that an idea a human being can invent may be an idea AI may not be able to invent. A human being’s idea is based on their experience, which is highly complex.

If you think about it, an idea has building blocks. AI has its own training. The human mind has its own training. On those building blocks, an idea originates. The idea of a human being is likely to be different because of its unique building blocks. Even with similar building blocks, some nuances might be different.

I think human intelligence is highly complex. It is shaped by a variety of experiences that a person has had over many years. Every second and every minute, they have an experience. Those experiences shape their thinking. Their mind is constantly being trained.

But the training of the human mind is not happening in a neat model. I think the training of the human mind is very random. A person might right now be in a small city. Then that person travels to a foreign country. They see different cultures. They meet different people. They have different experiences on a daily basis. One experience may have nothing to do with another experience. But all of them are shaping the same human mind.

AI is also trained, but its need for training makes this different. It may learn from what people have written online, from books and research papers, or even forums. AI can absorb that information differently from a human being whose experiences are changing. For example, there is the experience of buying vegetables, going to a café, going to the gym, having a weekend trip, having honest conversations with friends, and having a difficult conversation with work colleagues. Those are all a variety of experiences which AI is not trained on, or it may not be easy to train AI to create a universal assortment of experiences which captures the uniqueness of a human experience.

Ideas are plenty. But I think it all depends on who the idea is for. You have to think about the end consumer. The end consumer is most likely going to be a human being.

For AI to comprehend the psychology of a potential end consumer, given its need for training, may be more complex. If AI was generally building software for AI, business-to-business AI might do the best job. One AI company may communicate very well with another AI company because they understand that language. But even there, the business they are serving is eventually serving a human being.

Take financial data, for example. A company could get its financial data analysed beautifully by an AI company. The AI company may organise all of it extremely well. It may create the dashboard. Two companies may communicate very well with each other about that data.

But the end result is that the dashboard is being read by a human being to understand market conditions. So even when the communication between the businesses is highly technical, eventually it has to reach a human being.

Human beings understand human language. That is why, with AI, my understanding is that a lot of people might be saying, “I don’t understand what you’re saying.” They might say, “Can you say it more plainly?” or, when dealing with a chat agent, “Can I please talk to a human being, please?”

AI has taken human-written content and transformed it into AI-written content. It has perhaps been fed the idea that this is a more efficient way of communication in general. But an efficient way of communication is not necessarily the way human beings naturally communicate.

Conversations move in different directions. People get distracted. People go off on an assortment of tangents. I don’t think AI naturally does that in the same way because it is trying to organise things neatly.

Human beings may start talking about one thing and suddenly move somewhere else. Then they might come back. Something completely unrelated may remind them of something. That becomes part of the conversation as well.

Therefore, I think the content human beings consume also has to have that similar assortment.

Think of a movie. Sometimes there is extreme drama going on. Then suddenly there is a lighter moment. You move between different emotions. It is an assortment of emotions.

I think human beings want to experience that assortment. There has to be an idea which can imagine that assortment of emotions. For a neatly trained AI, that might be difficult. It may naturally want to organise everything into the most efficient and logical sequence, while human beings do not necessarily experience life in that sequence.

AI can invent an assortment of experiences for human beings to have. However, the essence would lie in how natural it feels, and that is the main thing. If it begins to actually, literally start feeling artificially put together, those experiences may not go further. The word artificial then becomes the key. Can AI get rid of itself from the tag of being artificial? That is going to be the greatest challenge.

That is why I think the human being may remain important at the level of the idea. AI can become exceptional at execution. It can become extraordinarily good at building things once you tell it what has to be built.

The human being may increasingly provide the direction. AI can say, “I can build almost anything.” The human being still has to say, “This is the thing worth building.”