Co-Intelligence

I had mixed feelings while reading Co-Intelligence. At times it felt a bit dry and technical, yet it still fascinated me enough to keep me engaged. Reading the book was an interesting experience, although the beginning was quite challenging for me. The first chapters include many technical explanations and because I am not very familiar with much AI terminology, I sometimes struggled to follow. Mollick talks about attention mechanisms, training data, "unknown unknowns" and different versions of language model and also about the evolution of Chat GPT from 3.5-4. For someone without a tech background, it can feel a bit overwhelming reading these first chapters. However, from chapter 3 onward the book becomes much easier to read. Mollick begins to use real-life examples of how AI appears in our every day lifes, for example at school, in businesses, and even in simple everyday tasks. These examples helped me to understand what he meant and made the read more accessible and practical. I was able to connect the theory to situations I recognised. One of the ideas I found especially interesting was AI as a coworker. Mollick suggests that humans and AI each have different strenghts. For example AI can be fast, creative and very efficient, while humans bring judgment, emotional understanding and responsibility. He explains that certain tasks should be done by AI, while it is crucial that others must remain in human hands. I thought this was a very realistic way of looking at collaborations. At the same time, Mollick warns that people might become lazy if they rely too much on AI. I agree with him, because if AI does everything for us, we risk losing our own skills. But if used correctly, AI can help us work better and faster. The real challenge I think is finding a healthy balance. Mollick also writes that nobody fully understands AI, not even the people who built it, this statement gave me a lot to think. Mollick states that systems behave like "alien minds", powerful but unpredictable. AI can do impressive things, like the example he mentions in the book; helping Amazon optimise packaging, but it also has serious limitations. For example, it can "hallucinate", meaning it invents facts or confidently states wrong information. I have experienced this myself. When I used AI for a calculation, it gave me a completely wrong answer. When I pointed it out, it didn't admit the mistake at first. Instead, it tried to talk around the problem, almost as if it were trying to protect itself. What surprised me even more was that when I asked the same question again, it gave a different wrong answer. It could not reproduce its own reasoning consistently. This personal experience made Mollick's point very real; AI can sound convincing while being completely incorrect, which can be very dangerous. Mollick also discusses the alignment problem and ethical issues such as bias, deepfakes, political misinformation and non-consensual image creation. Because AI is trained on data from the internet, often biased toward certain groups, it does not always reflect society fairly. This is why he argues for regulation and a broader societal response, not just government action. A chapter I found especially eye-opening was the one about treating AI like a person. Even though it speaks like a human, it is still a software and Mollick really points out the emotions-aspect. And forgetting this can create false trust or unrealistic expectations. In the end, Co-Intelligence helped me understand both the possibilities and risks of AI. Although the beginning was difficult, the later chapters were clear and practical. The book made me reflect on how AI might change the way we live, learn and work and how important it is to use AI thoughtfully.