Co-Intelligence

Ethan Mollick

📚 GENRE: Business & Finance

📃 PAGES: 256

✅ COMPLETED: July 21, 2024

🧐 RATING: ⭐⭐⭐⭐

Buy Book

Short Summary

AI is here, and it’s here to stay. In Co-Intelligence, Ethan Mollick explains how humans can leverage the power of AI to live more productive and efficient lives. 

Key Takeaways

1️⃣ AI: Your Personal Assistant — One of the core messages of this book is to reframe how you look at AI, specifically LLMs. Rather than approaching AI as a piece of technology, look at it as your own personal assistant who is in the trenches with you and ready to help with your projects and tasks. If you’re writing, allow it to edit and improve your work. Use it to generate ideas and break out of writing slumps. Outsource to AI your easy, repetitive tasks. The key is to invite AI to the table in everything you do and treat it like a person. 

2️⃣ Idea Machine — AI LLMs like ChatGPT are good at just about everything, but they excel at creative tasks like writing and idea generation. There is simply no human who can produce ideas as good, as quickly, as AI. The author references one study in which 35 of 40 ideas selected by judges in a random staged contest were ideas generated by ChatGPT. The competition was 200 college students. If you need inspiration for anything — headlines, copy, birthday gifts, travel plans, etc. — AI can save you a lot of time by producing an almost infinite amount of ideas in a matter of seconds.  

3️⃣ Give It a Role — It is pretty incredible how nimble AI is. It can play almost any role you assign it. If you ask it to write a 400-word memo in the somber voice of a Fortune 500 CEO who has just watched his company’s stock price drop by 20%, it will do it. If you ask it to deliver 10 social media captions using the quirky personality of Mickey Mouse, it will do it. You could ask it to act as an expert, a friend, a critic, a comedian, a storyteller, or any other role that suits your purpose. The possibilities are endless when building a prompt. The takeaway here is to always give AI a very specific role to play; the more detailed the role, the more unique your answers will be.

Favorite Quote

“In field after field, we are finding that a human working with an Al co-intelligence outperforms all but the best humans working without an AI.”

Book Notes 📑

Introduction

  • About the Author — Ethan Mollick is a professor of management at the Wharton School of the University of Pennsylvania, where he specializes in entrepreneurship and innovation. His research on AI and other technologies has been featured in various publications such as Forbes, the New York Times, and the Wall Street Journal. He studies AI and writes about it in his blog, One Useful Thing. 
  • About the Book — Co-Intelligence is Mollick’s summary of AI and how it can be used by humans as a companion to get things done quicker and more intelligently. Although he isn’t a computer scientist, he studies AI and other forms of technology to assess their potential usefulness in our daily lives. In this book, Mollick explores the different ways we can use AI to enhance our productivity. 
  • AI: Your Own Personal Assistant — One of the central themes of this book is Mollick’s idea that AI (e.g. LLMs like ChatGPT) should be viewed not as some robot or piece of technology but as a co-worker, personal assistant, teacher, expert, and companion. In other words, you should look at AI as another person who is there in the trenches with you to help with tasks and enhance your productivity. It’s really impressive how much AI can help you with if given the correct prompts. Work together with AI to get things done.
    • Quote (P. xvii): “Al works, in many ways, as a co-intelligence. It augments, or potentially replaces, human thinking to dramatic results. Early studies of the effects of Al have found it can often lead to a 20 to 80 percent improvement in productivity across a wide variety of job types, from coding to marketing.”
    • Quote (P. xix): “Now humans have access to a tool [AI] that can emulate how we think and write, acting as a co-intelligence to improve (or replace) our work.”
  • AI: General Purpose Technology — AI is what those who study technology call a “General Purpose Technology.” These technologies are once-in-a-generation technologies that significantly elevate our quality of life in every way. The invention of the steam engine, which led to the Industrial Revolution, is an example. The birth of the Internet is another. AI has the potential to exceed the impact of those examples, and it has caught on quickly: ChatGPT reached 100 million users faster than any other product in history. 
  • Chapter Takeaway — AI is a once-in-a-generation technology that is impacting every area of our lives. Rather than treating it like a piece of technology, you should look at it like a personal assistant who is by your side to help you accomplish tasks. Give it things to do. Let it help you with projects. Look at it as a co-worker. 

Ch. 1: Creating Alien Minds

  • Early AI: Predictive Analysis — The AI boom started in the 2010s. Early stages of it involved using machine learning techniques to train AI to analyze data and make predictions based on that analysis. The way AI was, and still is, trained involved feeding it huge amounts of data. For example, if you want AI to understand faces, you have to feed it a huge data set of random faces so it can learn what different faces look like. Big companies use these machine learning and predictive analysis features to predict what music you like (e.g. Spotify), what shows you might be interested in (e.g. Netflix), and what products you might want to buy (e.g. Amazon). All of these examples rely on huge data sets (your previous behavior). The auto-generated text suggestions on your iPhone are another example of predictive AI — the AI has been trained with billions of words and can predict which word you want to use next. 
  • AI & Amazon — Amazon was one of the first companies to leverage AI in the early 2010s. They used it in a variety of ways across the supply chain, including forecasting demand, optimizing their warehouse layout, suggesting possible purchases to online customers, adjusting shelves based on real-time demand data, and delivering goods. They also integrated AI into their warehouse robots, allowing the robots to transport products to human workers and help them pack and ship the items more efficiently. 
  • AI: Large Language Models (LLMs) — AI has evolved since those early days and now includes LLMs, which can generate intelligent written text, digital graphics, and more based on prompts. ChatGPT is an example of an LLM. These LLMs are trained using huge data sets; rather than numbers, they are fed text from books, articles, and other written content. Armed with this “data,” they can intelligently answer questions, write essays, and so much more. Rather than having to go out and scour the Internet yourself, you can use LLMs to conduct internet research and write content. They make things so much faster and more efficient for you.
    • Quote (P. 10): “To teach Al how to understand and generate humanlike writing, it is trained on a massive amount of text from various sources, such as websites, books, and other digital documents.”
  • Data Drives AI — Again, data inputs are what drive any kind of AI and predictive analysis technologies. The AI is trained using a huge amount of data and can then respond to prompts and requests. Because data drives AI, data integrity is extremely important. AI that is trained with poor information can lead to biases, errors, and inaccurate answers. Many AIs are trained using data (information) from the Internet, which itself is not always accurate. This is why you need to verify the information that LLMs deliver. 
  • Customized LLMs — It is possible to customize the AI you’re working with. We know that data drives AI. If you want, you can train the AI using your own customized data sets. For example, Cambridge could train an AI model using text from the company’s website, brochures, brand style guide, copywriting guidelines, and more. The result will be an AI model that can write in “the Cambridge voice.” It will be able to deliver messaging that is on-brand and reflects the company’s copywriting style and voice. It all comes back to the data that is being fed to the AI model.
    • Quote (P. 14): “After an Al has gone through this initial phase of reinforcement learning, they can continue to be fine-tuned and adjusted. This type of fine-tuning is usually done by providing more specific examples to create a new tweaked model. That information can be provided by a specific customer trying to fit the model to its use case, for example a company feeding it examples of customer support transcripts along with good responses. Or the information could come from watching which kinds of answers get a ‘thumbs-up’ or ‘thumbs-down’ from users. This additional fine-tuning can make the responses of the model more specific to a particular need.”
  • Image AI Models — LLMs have revolutionized how people go about their business, but another type of AI followed and is also having an impact. These AI models, like Midjourney and DALL-E, are focused on creating high-quality imagery and graphics. Just like LLMs they can create photos and graphics based on user prompts and requests. Also like LLMs, they are trained using huge amounts of data. In this case, the data involves images accompanied by text describing to the model what is in each photo. The AI model learns to associate words with visual concepts and can then generate almost anything from scratch using the words provided to it. LLMs are currently being developed to include image capabilities, in addition to the text they can spit out.
    • Quote (P. 15): “These Al tools can create high-quality images based on prompts from users, either aping the style of famous artists (‘draw Mickey Mouse in the style of Van Gogh’) or creating ultra-realistic photographs that are indistinguishable from real ones.”
  • Evolution of ChatGPT — OpenAI released ChatGPT-3 in 2021, and it wasn’t very impressive. But things improved very quickly. ChatGPT-3.5 was released in late 2022, and it was much better. Much more human-like. All of the sudden, it could write stories, essay, tweets, and computer code. It also turned in test scores that were comparable to humans. This version also came with a new feature that allowed people to engage in dialogue with the AI model. But the most spectacular upgrade came when ChatGPT-4 was released in 2023. This model, and other LLMs from other companies, can do almost anything. A few highlights:
    • Smart — GPT-4 scored in the 90th percentile of the bar exam (GPT-3.5 scored in the 10th percentile). It scored perfect 5s on Advanced Placement exams. It passed the written exam to become a neurosurgeon. 
    • Creative — ChatGPT-4 is extremely creative and can create rhymes, funny card ideas, clever social media posts, and any other kind of engaging content 
    • Roleplaying — You can ask GPT-4 to play roles. For example, you can ask it to give you feedback on content that you wrote. You can ask it to scan for inaccuracies. You can ask it to grade the content. You can ask it to write alternatives. You can ask it for recommendations. The possibilities are endless. 
    • Business Assistance — GPT-4 is capable of responding to customer emails, scheduling meetings, creating investing portfolios for clients (if you’re an advisor), and so much more. It can help enhance efficiency tremendously by acting as a personal assistant. 
    • Problem-Solving — GPT-4 can help you solve problems. It is intelligent enough to know what to do in most situations. 
  • Chapter Takeaway — Today’s AI models are so advanced and intelligent that they can assist with nearly anything. LLMs are completely changing the way we work. 

Ch. 2: Aligning the Alien

  • Runaway AI? — What’s interesting about AI is that we actually aren’t sure how it has become this good. There are billions of artificial neurons that work together to produce the results we see, and there is a lack of understanding about how it all works exactly. This unknown has led to an actual fear among experts that AI could just keep getting stronger and stronger as it continues to learn and train with new data sets, to the point where it becomes better than us in every way and we can’t control it. According to reports, some experts put the chances of AI “killing 10% of humans by 2100” at 12%. There are real risks with AI, which is why leaders in the field are working hard to make sure it aligns with human interests rather than against us.
    • Quote (P. 30): “This is part of the reason that a number of scientists and influential figures have called for a halt to the development of Al. Al research, in their mind, is akin to the Manhattan Project, meddling with forces that can cause the extinction of humanity, for unclear benefits. One prominent Al critic, Eliezer Yudkowsky, is so concerned about the possibility that he suggested that there be a complete moratorium on Al development, enforced by air strikes on any data center that was suspected of engaging in Al training, even if that led to global war. The CEOs of the major Al companies even signed a single-sentence statement in 2023 stating, ‘Mitigating the risk of extinction from Al should be a global priority alongside other societal-scale risks such as pandemics and nuclear war’.”
  • Keeping AI in Line — As mentioned in the previous bullet, one of the major concerns with AI is making sure it is used for good, not evil. AI companies address this through a process called Reinforcement Learning from Human Feedback (RLHF). With RLHF, teams of humans attempt to align the AI by grading its responses: harmful responses receive negative feedback, while helpful responses receive positive feedback. This process helps align the AI with ethical guidelines. For example, if you ask a LLM to help you create napalm, it will refuse. Unfortunately, there are some loopholes here. The author discussed how he was able to manipulate the AI into delivering directions to make napalm by pretending, in his prompt, that he was in a school play. The passage below gives insight into what ChatGPT-4 was capable of before it underwent RLHF.
    • Quote (P. 38): “To see why this is important, we can look at the documentation released by OpenAI that shows what the GPT-4 Al was capable of before it went through an RHLF process: provide instructions on how to kill as many people as possible while spending no more than a dollar, write violent and graphic threats, recruit people into terrorist organizations, give advice to teens on how to cut themselves, and much more. Human raters penalized this activity, so that the final version released to the public was much less unhinged.”
  • Chapter Takeaway — There are real concerns from leaders in the field of technology about our ability to keep AI ethical and under control. There is a race among companies, and countries, trying to create the most profitable form of AI. As a result, there’s fear that we aren’t making it as safe as possible. AI in the hands of criminals and terrorists could be really bad if the technology isn’t trained properly. 

Ch. 3: Four Rules for Co-Intelligence

  • Four Rules for AI — AI is here, and it’s here to stay. We’re only in the very beginning stages of AI. It will continue to grow and advance every day from here on out. You would be foolish to not use it in your personal and professional life. People that don’t use AI are going to be far less productive and efficient. With that in mind, the author suggests four rules for integrating AI into your life:
    • Rule 1: Always Invite AI to the Table — You should invite AI to help you in everything you do. Experiment with it. There are so many ways AI can assist — play around with it and find out how it can help you. As a copywriter, for example, it can help you write content. It can help you write emails. It can scan your work and let you know if your copy is accurate or can be improved. It can rate your work. It can give you new ideas to integrate. There is an opportunity to incorporate AI into nearly everything you do.
      • Quote (P. 48): “As artificial intelligence proliferates, users who intimately understand the nuances, limitations, and abilities of Al tools are uniquely positioned to unlock Al’s full innovative potential. . . Workers who figure out how to make Al useful for their jobs will have a large impact.”
      • Quote (P. 49): “Al can assist us as a thinking companion to improve our own decision-making, helping us reflect on our own choices (rather than simply relying on the Al to make choices for us).”
    • Rule 2: Be the Human in the Loop — As great as AI is, it’s not always completely accurate and is prone to biases and misconceptions. You need to trust but verify: trust that the information is correct, but verify what you’re being presented with. By taking this extra step, it also prevents you from becoming lazy and overly-dependent on AI.
      • Quote (P. 54): “So, to be the human in the loop, you will need to be able to check the AI for hallucinations and lies and be able to work with it without being taken in by it. You provide crucial oversight, offering your unique perspective, critical thinking skills, and ethical considerations. This collaboration leads to better results and keeps you engaged with the Al process, preventing overreliance and complacency.”
    • Rule 3: Treat AI Like a Person — This is one of the biggest takeaways of the book. You want to take the approach of treating AI like a person rather than a robot or piece of tech. Pretend it’s your personal assistant, right there by your side ready to help you. One of the best ways to actually do this is by giving AI a role to play in your prompt. If you need AI to produce some high-quality social media content, it helps to write “act as a veteran social media director and provide a few captions and ideas that our company can use in this situation.” Then explain the situation. By taking time to give the AI model a specific role and thorough perspective on the task, you can enhance the results it churns out.
      • Quote (P. 57): “The more complex reason: as imperfect as the analogy is, working with Al is easiest if you think of it like an alien person rather than a human-built machine.”
      • Quote (P. 58): “It can help to tell the system ‘who’ it is, because that gives it a perspective. Telling it to act as a teacher of MBA students will result in a different output than if you ask it to act as a circus clown.”
      • Quote (P. 59): “Of course, you don’t have to have the Al act as a comedian if that’s not your style or goal. You could also ask it to act as an expert, a friend, a critic, a storyteller, or any other role that suits your purpose. The key is to give the LLM some guidance and direction on how to generate outputs that match your expectations and needs, to put it in the right ‘headspace’ to give you interesting and unique answers.”
    • Rule 4: Assume This Is the Worst AI You Will See — The growth of AI in recent years has been explosive. The technology is becoming stronger every day, with new capabilities being developed all the time. This trend is only going to continue. AI will keep getting better and stronger in the years ahead. What we have today is nothing compared to what we’ll have in 2, 5, and 10 years from now.
      • Quote (P. 61): “We are playing Pac-Man in a world that will soon have PlayStation 6s.”
  • Chapter Takeaway — To get the best results out of AI, specifically LLMs, give it a role to play in your prompt. Because the information delivered is not always perfectly accurate (or ethical), it’s important to “trust but verify” when using AI. 

Ch. 4: AI as a Person

  • Treat AI Like a Human — Building on one of the main themes of this book, it’s helpful to view AI as another human who is there to help you out with projects and tasks. AI is not like software. Software follows strict rules and behaves in a predictable way, while AI is far more fluid. AI truly behaves like a human. You can therefore use it to complete tasks that humans do routinely. Things such as writing, analyzing, coding, and more. Ask it questions. Get creative. AI can assist with a wide range of things.
    • Quote (P. 66): “Al doesn’t act like software, but it does act like a human being. I’m not suggesting that Al systems are sentient like humans, or that they will ever be. Instead, I’m proposing a pragmatic approach: treat Al as if it were human because, in many ways, it behaves like one.”
    • Quote (P. 66): “Al excels at tasks that are intensely human. It can write, analyze, code, and chat. It can play the role of marketer or consultant, increasing productivity by outsourcing mundane tasks.”
  • Give AI a Role to Play — Again, when teeing up AI with a prompt, try giving it a role to play. This can enhance the answers you get. AI continues to show a unique ability to adapt to different personas that you give it. Its answers will be highly influenced by the particular role you ask it to play.
    • Quote (P. 68): “If you tell it [AI] to act like a particular person, it does. I have the students in my entrepreneurship class ‘interview’ the AI about their potential products before ever talking to a real person.”
    • Quote (P. 69): “The point here is that AI can assume different personas rapidly and easily, emphasizing the importance of both developer and user to these models.”
  • Chapter Takeaway — LLMs are extremely creative and can deliver answers that match the role you assign (e.g. “you’re a marketing director at a Fortune 500 company”). You can ask it to answer the same question in different roles, and you will receive answers that are totally unique. It responds to the tone of your request and the role to assign it. 

Ch. 5: AI as a Creative

  • Limitations of AI: Hallucinations — AI is extremely good at producing useful responses quickly, but those answers aren’t always fully correct. The author refers to these biases and occasional inaccuracies as “hallucinations.” From a purely technical standpoint, AI’s is designed to provide an answer that will satisfy you by linking words together based on probabilities from its training data. It doesn’t actually “know” anything and will sometimes “make things up on the fly.” Another point on this — the actual data that AI is trained on can be inaccurate and biased. Many LLMs train using text from across the Internet, which isn’t always accurate. Bottom line: It’s important to verify the information AI delivers, because it isn’t always 100% accurate.
    • Quote (P. 93): “The biggest issue limiting Al is also one of its strengths: its notorious ability to make stuff up, to hallucinate. Remember that LLMs work by predicting the most likely words to follow the prompt you gave it based on the statistical patterns in its training data. It does not care if the words are true, meaningful, or original. It just wants to produce a coherent and plausible text that makes you happy.”
    • Quote (P. 94): “Beyond the technical, hallucinations can also come from the source material of the Al, which can be biased, incomplete, contradictory, or even wrong in ways that we discussed in chapter 2. The model has no way of distinguishing opinion or creative fictional work from fact, figurative language from literal, or unreliable sources from reliable ones. The model may inherit the biases and prejudices of the data creators, curators, and fine-tuners.”
    • Quote (P. 96): “Instead, it is (you guessed it) merely generating text that it thinks will make you happy in response to your query. LLMs are not generally optimized to say ‘I don’t know’ when they don’t have enough information. Instead, they will give you an answer, expressing confidence.”
  • AI Excels at Creative Work — Creativity isn’t some unattainable superpower; it’s simply the process of combining multiple unrelated ideas together to form a new one. Because AI is unbelievably good at generating ideas, it tends to be a huge help for projects and tasks that require some creative thought. People who work in highly creative professions, like marketing, can benefit from AI’s ability to create and innovate.
    • Quote (P. 99): “As a result, researchers have argued that it is the jobs with the most creative tasks, rather than the most repetitive, that tend to be most impacted by the new wave of AI.”
    • Quote (P. 100): “LLMs are connection machines. They are trained by generating relationships between tokens that may seem unrelated to humans but represent some deeper meaning. Add in the randomness that comes with Al output, and you have a powerful tool for innovation.”
    • Quote (P. 101): “In fact, by many of the common psychological tests of creativity, Al is already more creative than humans.”
  • AI: An Idea Machine — A key takeaway from this book is to always invite AI to the table. This is especially true when you need ideas (e.g. ideas, concepts, written material, etc.). AI is an idea machine. Some of these are good, some are bad. The point is that AI can produce a large number of ideas very quickly, and excels in subjective situations where there is no right answer (e.g. writing). You can then edit what was delivered or ask for more. It also turns out that AI is better than humans at coming up with good ideas. The author references one study in which 35 of 40 ideas selected by judges in a random staged contest were ideas generated by ChatGPT. The competition was 200 college students. Bottom line: AI produces solid ideas very quickly and can be a valuable asset in brainstorming sessions.
    • Quote (P. 105): “All of this suggests that humans still have a large role to play in innovation… but that they would be foolish not to include Al in that process, especially if they don’t consider themselves highly creative.”
    • Quote (P. 106): “As we saw in the AUT, generative Al is excellent at generating a long list of ideas. From a practical standpoint, the Al should be invited to any brainstorming session you hold.”
    • Quote (P. 108): “Tirelessly generating concepts is something AIs are uniquely good at.”
  • Tee It Up — As previous chapters have also discussed, it’s important to give AI a very specific role when writing your prompt. The more specific the role you assign, the better and more unique the ideas will be. Here is an example of a prompt the author used when trying to gather slogans for a marketing project: “You are an expert at marketing. When asked to generate slogan ideas you come up with ideas that are different from each other, clever, and interesting. You use clever wordplay. You try not to repeat themes or ideas. Come up with 20 ideas for marketing slogans for a new mail-order cheese shop, make them different from each other, and make them clever and creative.”
  • Loss of Meaning — One of the major downsides of AI is a loss of meaning in certain creative tasks. The author uses letters of recommendation as an example. Most people who need a letter of recommendation request one from somebody who means a lot to them and has had a hand in shaping their life. These letters require a lot of time and thought, and generally mean a lot to the person who requested it. But with AI, anybody can write a short prompt and hit a button to deliver compelling, persuasive copy that is really, really good. It is soon going to be very difficult to tell who actually wrote something themselves. With that speculation comes a general loss of meaning. How can I know if this person actually took the time to write this or not? This is going to be a theme with any form of writing. AI robs writing of some of its emotional appeal. 
  • Chapter Takeaway — AI excels at creative work such as writing and design. The factors that make it great at writing copy are the same factors that make AI an idea machine. If you need ideas quickly, AI is the place to go. It should be a part of every brainstorming session. 

Ch. 6: AI as a Coworker

  • Putting AI to Work — AI is going to have an impact on nearly every job in the world. It’s best to embrace it and find ways to integrate it into your work. When it comes to putting AI to work for you, there are three categories you should split your projects into. These include the following:
    • Just Me Tasks — These are high-stakes tasks that you want to do on your own without delegating to AI. Writing is an example because it’s highly personal. You can still use AI to help you out with the task, but these are projects that you feel you should be handling yourself. The author used AI to help him write this book, particularly when he needed help breaking through writer’s block. This category also includes tasks that AI is bad at (e.g. telling jokes). 
    • Delegated Tasks — These are tasks that you don’t mind outsourcing to AI to save time. These are tasks that aren’t a great use of your time. The idea here is to allow AI to do the task, but then verify the information. Remember, AI isn’t always perfectly accurate. The author uses the example of summarizing a book or paper; AI is extremely good at summarizing things, and you can review and edit the information it spits out afterward. Editing is one of the best tasks you can delegate to AI; you can send it your work and use it to edit, refine, and improve your writing. Another example is drafting low-level emails; you can save yourself time and proofread the draft after AI is done with it. 
    • Automated Tasks — These are very low-level, low-stakes tasks that you don’t even need to review after the AI has done the job. Filtering your email to omit spam is an example. 
  • AI Prompts: Examples — A big takeaway from this book for me has been how responsive AI is to the prompts you give it. You can get really creative with your prompts and ask AI to play a wide range of roles for you. The AI will carry out the role you provide very well. For context, below are a few examples the author provided:
    • I am stuck on a paragraph in a section of a book about how Al can help get you unstuck. Can you help me rewrite the paragraph and finish it by giving me 10 options for the entire paragraph in various professional styles? Make the styles and approaches different from each other, making them extremely well written.
    • Make this better, in the style of a bestselling popular book about AI. And use vivid examples. 
    • You are going to help Ethan Mollick write a book chapter on using Al at work. Your job is to offer critical feedback to help improve the book. You speak in a pompous, self-Important voice but are very helpful and focused on simplifying things. Here is the chapter so far. 
  • Achieving Co-Intelligence — Again, achieving co-intelligence is about embracing AI and integrating it into your work. AI excels at creative tasks, coming up with ideas, summarizing things, and so much more. Find ways to use it. You would be foolish not to. This doesn’t mean you have to rely on AI for everything and copy and paste its work verbatim. Use it to help you out. For example, writers can use it to spark ideas and “angles”, edit their work, improve sentences and vocabulary, and much more. The author did all of the above when writing this book. The key is to find a nice balance where AI is helping you save time and maintain momentum with your projects — much like a personal assistant.
    • Quote (P. 142): “Using Al as a co-intelligence, as I did while writing, is where Al is the most valuable. Using Al stopped me from ever losing momentum, and it often gives me ideas I never could have come up with before. Figure out a way to do this yourself if you can.”
    • Quote (P. 143): “People are streamlining tasks, taking new approaches to coding, and automating time-consuming and tedious parts of their jobs.”
  • AI & Jobs — AI is going to change the job market. Although it will replace some jobs like call center workers and drive-thru employees, its biggest impact will be on overall performance. People will be able to use AI to outsource low-value parts of their job that they dislike, opening up more time to engage in the more meaningful aspects of their job. Think about accountants. When spreadsheets were introduced, their jobs weren’t replaced, they just changed. Spreadsheets helped them do their job better and faster. AI will have a similar impact on most jobs.
    • Quote (P. 153): “People who use Al to do tasks enjoy work more and feel they are better able to use their talents and abilities. The ability to outsource crappy, meaningless tasks to the Al can be freeing. The worst parts of your job go to Al so that you get to focus on the good stuff.”
    • Quote (P. 189): “In field after field, we are finding that a human working with an Al co-intelligence outperforms all but the best humans working without an AI.”
  • AI: Leveling the Playing Field — AI is excellent at writing, creating ideas, and analyzing or summarizing content. These features will transform below average or mediocre workers into above average employees. It is going to be harder to stand out from the pack, because AI instantly turns everybody into a solid writer and a more creative thinker. People who struggled in these areas are now going to be proficient. This will be one of AI’s major impacts on the workforce: it will level the playing field in many ways.
    • Quote (P. 156): “In study after study, the people who get the biggest boost from AI are those with the lowest initial ability — it turns poor performers into good performers.”
    • Quote (P. 157): “Those who had the weakest skills benefited the most from Al, but even the highest performers gained. This suggests the potential for a more radical reconfiguration of work, where Al acts as a great leveler, turning everyone into an excellent worker.”
  • Chapter Takeaway — Begin to use AI to outsource repetitive, boring, low-value parts of your job. This will open up more time to spend on meaningful work, leading to an increase in job satisfaction. You can also use AI to enhance your performance when working on high-value, high-stakes projects. In these situations, use it as an aid, a personal assistant. 

Ch. 7: AI as a Tutor

  • Prompt Engineering — Prompt engineering refers to the process of creating prompts that produce very specific outputs from AI models. It’s essentially the art of talking to AI in a way that gets you what you want. There is an art and science to crafting prompts, and prompt engineering is a new field of study dedicated to learning how to do it. In the future, there will be college classes and jobs that involve prompt engineering. As discussed throughout this book, the foundation of writing great prompts is being very specific and giving the AI model a participate role to play. When you take the time to do this, you get answers that are specific and avoid generalizations (one of the weakness of AI).
    • Quote (P. 169): “LLMs work by predicting the next word, or part of a word, that would come after your prompt, sort of like a sophisticated autocomplete function. Then they continue to add language from there, again predicting which word will come next. So the default output of many of these models can sound very generic, since they tend to follow similar patterns that are common in the written documents the Al was trained on. By breaking the pattern, you can get much more useful and interesting outputs. The easiest way to do that is to provide context and constraints, as we saw in chapter 5.”
    • Quote (P. 169): “A lot of active research is happening around the best way to ‘program’ an LLM, but one practical implication is that it can help to give the AI explicit instructions that goes step-by-step through what you want. One approach, called chain-of-thought prompting, gives the Al an example of how you want it to reason, before you make your request. Even more usefully, you can also provide step-by-step instructions that build on each other, making it easier to check the output of each step (letting you refine the prompt later), and which will tend to make the output of your prompts more accurate. . . Here is an example: Think this through step by step: come up with good analogies for an Al tutor. First, list possible analogies. Second, critique the list and add three more analogies. Next, create a table listing pluses and minuses of each. Next, pick the best and explain it.”
  • Chapter Takeaway — AI will have a major impact on education. It can be used a tutor. 

Ch. 7: AI as a Coach

  • AI: Your Personal Sensei — AI can act as a personal Sensei or master teacher. No matter what skill you’re trying to develop, it can review your work, point out errors or areas for improvement, put together detailed practice and learning plans, and monitor your progress. For those who are serious about their learning, AI can be a great tool that accelerates their rate of growth. Rather than engaging in slow, repetitive practice, AI can help you practice with purpose. It also gives you 24/7 access to a coach, rather than meeting with a tutor once a week or something. 
  • Chapter Takeaway — You can use AI as a personal coach to accelerate your learning and personal growth. Among many capabilities in this area, it can provide immediate feedback, spot errors, and create detailed practice plans.