If you visit any news or social media website, you’ll likely stumble upon articles about ChatGPT and other artificial intelligence (AI) tools. Opinion pieces, interviews with experts, latest and greatest prompting tips… Among all the buzz and noise, it can be hard to find quality resources to understand this technological tsunami and how it might impact our future.

Meanwhile, ChatGPT is estimated to have reached 100 million monthly active users just two months after launch, becoming the fastest-growing consumer application ever. A recent survey among US workers found out that 85% have already used AI tools for work-related tasks. The number is pretty consistent across generations. Wanna bet most of this has been unauthorized bottom-up adoption at an unprecedented scale?

Why learning about AI matters

As much as I wish we could have a broader and more inclusive debate on the ethics of such a fast adoption, the golems have emerged from the ovens of Silicon Valley and walk among us, and we can no longer pretend this isn’t happening. Unless you go off the grid, you’ll soon have no choice but to use generative AI. Your AI Copilot is coming to every Windows 11 PC, Microsoft 365 app, Google Workspace app, Slack, and most of the other apps you use daily. You’ll be expected to figure out how to integrate AI into your workflow. If you’re a developer, you’ll soon be asked to integrate generative AI into your product. And if we want to have a say in how this technology is being developed and deployed, we should all learn more about it.

It’s easy to dismiss ChatGPT as a fancy word prediction machine and make fun of the silly outputs it sometimes generates. But I could make the same argument for people. What I’m doing right now as I try to choose which word to type next is not so different from what LLMs – Large Language Models – such as ChatGPT are doing. Language is the operating system of humanity, and now we have software that can easily generate it convincingly, and effortlessly translate between different modalities. This is huge.

This is why I’ve decided to share some of the resources that have helped me learn about different aspects of generative AI beyond the hype. I’ve been following the field of machine learning from the sideways for a while now. Four years ago, I completed the excellent Deep Learning specialization on Coursera, so I have some affinity for the nerdy side of it, but I’m probably even more interested in how this technology will change our social fabric. As always, I believe there’s value in exploring both the forest and the trees, so I’m highlighting resources that can help you gain a systemic view of the field as well as practical resources that might inspire you to start tinkering with GPTs through APIs.

Resources to help you see the AI forest

  • I suggest you start by watching The AI Dilemma, a presentation given by Tristan Harris and Aza Raskin, the co-founders of the Center for Humane Technology, at a private gathering in San Francisco in March this year (before the launch of GPT-4). Harris and Raskin use memorable metaphors to explain what the current wave of generative AI capabilities pose such a great risk to the functioning of our society. The Center for Humane Technology was also involved in the making of The Social Dilemma documentary you can watch on Netflix, and offers an excellent self-paced free course on Foundations of Humane Technology.
  • Next, I highly recommend the New Yorker essay Will A.I. Become the New McKinsey? by Ted Chiang. It looks at how AI is currently designed to concentrate wealth and disempower workers, and reminds us that it’s up to us to imagine different application of this technology. I particularly love this quote: “The tendency to think of A.I. as a magical problem solver is indicative of a desire to avoid the hard work that building a better world requires. That hard work will involve things like addressing wealth inequality and taming capitalism.”
  • If the essay above gives you a taste for a more systemic view of AI, you should also pick up the book Atlas of AI by Kate Crawford. This book is an excellent introduction into how the AI sausage gets made – spoiler alert, it’s not pretty or clean! – and does a great job at surfacing different aspects of political power, social justice, and the environmental impact of AI. For a short sample, here’s a good review and interview with the author of the book.
  • … and a bit of bonus reading that’s close to my heart as a non-native English speaker. The question of language representation isn’t brought up often enough, so I enjoyed the Wired piece The Dire Defect of ‘Multilingual’ AI Content Moderation. And I’m still looking for good resources on how AI might widen the digital divide, a big topic that I rarely see mentioned.

Resources to help you generate & nourish AI trees

Hopefully, you’re now convinced of how important AI is and are ready to take some of its power into your own hands. Here are my recommendations:

  • The State of GPT talk by Andrej Karpathy from this year’s Microsoft Build conference is full of little gems. He starts by explaining how GPT assistants like ChatGPT are trained, and then goes on to provide practical advice on the development and prompting of these models. A very accessible talk, even if you’re not yet familiar with the technical details.
  • Speaking of technical details, as you dive deeper, you’ll start hearing people talk about transformers. Not the action movie ones, but an equally fascinating deep learning architecture that is heavily responsible for the success of generative AI tools. Understanding the transformer architecture isn’t mandatory if you want to focus on building practical applications, but learning about the basic concepts from beginner-friendly explainers on YouTube such as this one from AssemblyAI will help you follow discussions about experts and might even help you think about prompting differently.
  • Another great way to learn about the terminology and basic concepts – inspired by my ResponsibleTech.Work co-conspirator Daniel – is to go straight to the horse’s mouth and have a chat with ChatGPT! Even outside the field of AI, I’m finding it helpful to learn about new concepts through conversations with ChatGPT. The process of formulating questions and the ability to check my understanding with a super patient and super knowledgeable mentor is priceless. Just keep in mind its limitations and hallucinations; you’ll definitely want to check any references the model gives you on your own.
  • If you’re itching to get your hands dirty, the short – and free for now – course ChatGPT Prompt Engineering for Developers with Isa Fulford and Andrew Ng is another great resource to learn about prompting using the OpenAI API. You don’t even need technical expertise to follow along. You’ll learn useful tips for better prompting, and the course includes an interactive notebook where you can experiment based on the examples from the course.
  • If you feel overwhelmed by the proliferation of LLMs and applications – and who doesn’t! – Stanford’s ecosystem graphs website can help you keep track of the emerging ecosystem of datasets, models, applications, and their relationships.
  • If you’re looking for one resource to rule them all, the Prompt Engineering Guide by DAIR.AI is the one that’s worth bookmarking. The guide is full of technical explanations, practical prompting examples, and links to papers, tools, and more.

It’s pretty obvious the field is developing at breakneck speeds, so at some point you’ll have to accept you’re not a LLM and find smarter ways to keep track of news and trends. Eventually, we’ll probably have helpful assistants provide personalized daily or weekly summaries, but until then, many brilliant humans do a great job at distilling the noise. My personal strategy is to subscribe to a couple of newsletters that help me get a sense of what’s going on without overwhelming me with detail. Here are my current favorites:

  • Import AI by Jack Clark: a newsletter focused on AI research, but it’s easy to follow for non-researchers because the author writes highly informative explanations on why each piece of research or news matters in a wider context.
  • Guide to AI by Nathan Benaich: a monthly analysis of AI research, geopolitics, and startups.
  • The AI Exchange: focused on practical applications and current news.

I don’t read these newsletters from header to footer, but I find it helpful to skim them and dig deeper into topics that catch my eye.

And that’s it, for now. I hope these links and tips help, and best of luck in your own exploration journey! If you’ve found this helpful, please pay it forward by sharing your recommendations on resources that helped you learn and reflect more deeply about our future AI overlords.

P.S.: In case you were wondering, this blog post was still 100% human-written, as you can probably tell by the clumsy, non-SEO optimized title. Bonus points if you catch the Easter egg in it.