Co-Intelligence: Living and Working with AI

Food for thought
I will definitely go and rethink how I integrate LLMs and other ML tools into my work after reading this book.Helpful and encouraging advice + fun to read

A well-written, timely intro to AI
Ethan Mollick’s *Co-Intelligence: Living and Working with AI* presents a compelling exploration of artificial intelligence and its impact on our lives. Mollick’s clear and accessible writing makes complex topics understandable. His fearless approach ensures readers are both informed and engaged.Mollick balances optimism with caution in his assessments of AI. He discusses potential risks and ethical concerns while highlighting AI’s transformative potential. His thought-provoking analysis encourages readers to consider how AI can be harnessed responsibly.Mollick introduces the concept of “co-intelligence,” where humans and AI collaborate symbiotically. He argues that AI should augment human intelligence rather than replace it, presenting a refreshing and optimistic perspective.Mollick excels in making complex ideas accessible without oversimplifying. His clear and engaging prose appeals to both tech-savvy readers and those new to AI.For those interested in AI, this is a good primer.

Primer for understanding AI
Just finished reading Co-Intelligence by Ethan Mollick. In many ways it is an introductory ‘primer’ on AI (artificial intelligence).Written by Ethan Mollick, a professor at Wharton, it explores what AI is, how we can utilize is as a ‘co-intelligence’ and what the future might look like with AI.As someone who has used ChatGPT and other tools, but did not have much understanding of how they work, the book provided that background. On the other hand, I found it lacking in the depth and creativity I hoped for in exploring what a future world with AI might look like.Some insights from the book:- AI is built on Large-Language Models (LLM’s). These models have huge datasets at their core. By using this information, it can generate responses by predicting (and displaying) the word that is most likely to come next as a response to a prompt (and as a response to what it has already generated. The association between words are called ‘weights’ and these are adjusted based on the frequency and association between terms (if I have understood this correctly.)- Because of the way that LLM’s are built, the biases that we have as humans will get translated into AI. AI tools have a stage called Reinforced Learning from Human Feedback (RLHF) which is a process by which humans look to remove the bias. (Of course, this introduces another source of bias, but is important in ensuring that AI does not become a mirror of the uglier parts of ourselves which often manifest online and in other places which may be sources of data for AI.)- This process means that unlike other areas of technological development, AI finds it harder to be consistent and structured and instead seems to be actively creative. An analogy which is not found in a book, but which seems apropos was the shift in computer graphics from Euclidian Geometry to Fractal Geometry as the basis of graphics.- Mollick notes that the nature of AI makes it extremely difficult to predict exactly what it will be great at and what it will not. It also means that AI will include mistakes and fabrications without any self-awareness that it is doing so.- Mollick believes we are better for ‘inviting AI to the table’ while maintaining human involvement (you can get great results by working with AI, and poorer results [at least today] from handing over tasks to AI without oversight. He recommends viewing AI as a person in the sense that AI will be quirky and unpredictable. He also notes that whatever AI you are using today will likely be the worse than any version you ever use in the future as the technology continues to develop.- In the book Mollick shows how to use AI (engage with it, provide it with specific instructions “Give me a Mexican recipe that takes under 20 minutes to make using ingredients X, Y and Z [items in my fridge] and can be done by a 12 year-old without any instruments that would potentially be dangerous [like blender’s etc].” He recommends breaking down problems into steps (a process you can ask AI to assist etc.)- Mollick is generally pretty optimistic about AI. He points to studies that show that it helps people who struggle in areas to elevate themselves and believes it may help resolve income inequality etc.- Mollick mentions, but in my opinion does not give enough attention to potential issues. He notes that people often hand off tasks to AI without engaging mentally and we are in real trouble if we take another step back in critical thinking (in my opinion). He does not focus on the ways that AI could easily exacerbate income inequality as those with money put guardrails around the best tools etc. He also does give almost any focus on the way that AI can spit out endless content in little time and in the hands of bad actors, can take our current issues with fakenews, and make it 1000x worse very quickly.- Mollick notes that practical changes occur within systems and the changes we are talking about need to contend with the structures (economic, social, political etc) which currently exist. This reminds me of Kuhn’s descriptions of changes and the slow pace of the Copernican Revolution, in part because it bumped into theological systems.- As I noted above, I felt like Mollick did not spend nearly enough time exploring the issues that could easily arise from AI. What happens in a world where there is endless content being generated and no clarity on the legitimacy of it? Does it lead to additional conflict? How does it play a role in elections, campaigns etc.? Is additional government regulation generated to crack down? A social media platform exclusive to non-AI where you have to validate everything you are writing (somehow)? Influencers with “Real Human – no AI content” badges? What about Human-Reviewed AI Content? Is there a reason that is a bad thing?All in all, I think that AI is likely to have a dramatic impact on the world of the future. This book is a good intro. In all likelihood large aspects of it will become outdated rather quickly. And I wish it had done some more deep thinking about the future. It feels like just a shallow exploration.

Ótimo livro, linguagem simples e direta, tocando todos os pontos importantes de umPonto de vista estratégico, mas sem muitos achismos. Valioso!

This is the age of LLMs, and as stated in the book. We are using the worst possible LLM.Though the future (as they say in star wars) is clouded. But there are certainly ways to make these tools help us. This books gives us the idea of how to do that.Not only this book, but Ethan Mollick has a must read substack worth following and reading