How to live and work with Artificial Intelligence
Four Rules of Co-Intelligence
Always invite AI to the table
It blows my mind when I hear people don't want to use AI because they don't understand how it works or they are afraid of it. Whether you like it or not, AI is and will be part of everything we do. The best way to not fear AI is to learn more about it, and the best way to learn more about it is to use it.
Start with simple tasks like brainstorming ideas, fixing your spelling, grammar or tone in your email, or analyzing large documents you may have a hard time reading. One of the best ways to understand the limitations of AI (or rather, whether your prompts are good or not) is to test AI on something you know well - whether it's project management, financial investments, or art history.
Be the human in the loop
To that end, maintain oversight to verify accuracy, address biases, and ensure ethical outcomes. When you start using AI on things you know, you will see what exactly AI can and can't do.
As you start using AI on more and more tasks (and more and more use cases are being expanded every day), ensure that you are reviewing. AI, as you may know, is prone to hallucinations, i.e., making things up that may not be true. It does this because AI is rewarded for satisfying its customer (you) and it does this by generating responses, even when the responses aren't based on facts.
Treat AI like a person (but specify its role)
One way to improve AI's responses: tell AI to act like a certain individual or expert. If you're trying to get better marketing strategies and tips, tell AI to act like a marketing expert or Seth Godin.
Or if you're not sure what kind of individual or expert you want the AI to act like, tell it to act like different individuals to get different ideas (act like a scientist, act like a doctor, act like an economist) and then use the different responses to get vastly different ideas.
Assume this is the worst AI you’ll ever use
It seems like every week or month, a new AI model comes out. O4, Gemini Flash, Claude 3.7. Just because you used AI to help you with a task, and it didn't perform well then does not mean it will not perform well today.
It pays to constantly experiment with the AI model of your choice or with different AI models to get the most out of AI.
Collaboration Models
There are two ways Mollick suggests for how to work with AI.
In the centaur model, you divide tasks between humans and AI. For example, if you were writing a book, AI might draft the content and then humans may act as editors, providing additional prompts to direct AI to do certain things.
Recommended by LinkedIn
In the cyborg model, humans and AI are blended seamlessly. Think as you are writing a post how Grammarly highlights specific words, spelling mistakes or grammatical errors. These real-time AI suggestions
With AI improving every day, and use cases expanding, there may come a time when we all work through the cyborg model. Most people (including myself) are more comfortable with the centaur model for now because the centaur model allows humans and AI to leverage their strengths.
AI’s Impact on Work and Learning
One of the most interesting things for me is that AI democratizes expertise. What that means is that AI helps newcomers and the lowest-performing employees the most. AI certainly helps experts, but not as much. Yes, you still need to be an expert to recognize when AI might be hallucinating, but AI helps to remove a lot of the barriers in switching careers, understanding different industries or lines of work, and helping people become close to an expert very quickly.
Given my time in education, Mollick talks about the concept of a 'flipped classroom' where students learn through AI tutors at home, and class time is reserved for discussion and critical thinking. When AI can provide personalized content at a scale much more effectively and efficiently than teachers can, why wouldn't teachers use the class time for teaching skills that (as of now) AI can't teach to students?
AI is extremely powerful in helping generate ideas. However, if you go to different AI models now to generate ideas, you will get a lot of very similar ideas UNLESS your prompt is also unique. When people say AI means you don't need to be creative anymore, Mollick argues that there is still a need for creativity in your inputs and prompts to get non-generic outputs.
Key takeaways
Experiment broadly
Delegate strategically
Stay informed
Hey you got to the end! Want more from me? Did you know I also have a Substack where I share what I learned from the 50+ books I read every year? It's a lot like this newsletter - a weekly article - except I also cover health and other non-work-related topics. Join hundreds of readers and subscribe so you don't miss the next newsletter when it comes out.
