Unless you’re on a media cleanse, you’ve probably come across someone excited about AI recently. For me, it seems like every day there’s a new article to read, podcast to listen to, or Twitter thread to unfold about the implications of the latest breakthroughs. Just a few weeks ago, The New York Times dedicated an entire episode of The Daily podcast to wondering “Did Artificial Intelligence Just Get Too Smart?” Meanwhile, on a different daily, Daily Show host Trevor Noah mused about AI-generated pirates of the future that teach kids quantum physics.
Progress in the field over the last decade has been remarkable to witness. But it’s really just in the last few months that our collective imaginations have been set ablaze, particularly by generative AI models like DALL-E 2 and ChatGPT, two systems recently released by OpenAI. These models and others like them respond to simple textual prompts with novel computer-generated content seemingly pulled from thin air, but actually computationally synthesized from massive amounts of data from the web. Different models respond to your query with different types of media—ChatGPT responds with text, DALLE-2 with images, and Copilot from Microsoft responds with code. The results are often wildly creative and spookily accurate, giving these models a human-like feel.
People are comparing this moment to the dawn of the modern internet in the early 1990s. Just as the internet revolutionized the way we communicate and access information, generative AI is poised to revolutionize the ways we create and process information. Although the technology is developing quickly, it will take time for the full potential of generative AI to be realized, and we will likely see many new applications and unexpected developments in the coming years. The massive opportunity is clear, but it is still early days.
There are disruptive first-order effects of generative AI that we’re already seeing play out, particularly relevant to the future of work. New, AI-powered word processors like Jasper and Lex are promising to help you write faster and even cure writer’s block by using generative AI to suggest the next paragraph as you write. The old guard is taking notice: Microsoft plans to integrate OpenAI’s models into their Office products. Similarly, Shutterstock recently announced it will also integrate AI-generated stock “photos” from DALL-E 2 in its search results, angering the content creators who power their platform (and as it turned out, also helped train DALL-E 2). The promise of these new technologies at helping people be more productive is exciting, but they also raise thorny questions about trust, intellectual property, and plagiarism that we’re just starting to grapple with.
However, lately, I’ve been thinking about some second-order effects that may be less obvious. Here are a few ideas I came up with. Since I’m an optimist by nature, these all lean toward the brighter side of things. But, I’d love to hear counterarguments and more pessimistic takes in the comments here or on Twitter.
In 1996, Bill Gates famously wrote an essay called Content is King. In it, he predicted that by reducing the marginal costs of distributing information basically to zero, the internet creates enormous opportunities for content creators of all kinds. In the last 25 years, his predictions have largely come true, with the internet leading to massive disruptions in various industries such as media, entertainment, and advertising.
Now enter generative AI. Since these new models literally generate content—text, images, music, code—we are entering a world where the costs for creating and synthesizing content will decrease dramatically, changing the nature of the jobs and industries where content creation is a central focus, such as writing, journalism, design, research, marketing, and programming—many of the same jobs that were also transformed during the internet revolution.
In the midst of all this, AI is more likely to transform content-creation jobs rather than replace them. Instead of completely replacing human workers, AI will augment their abilities and automate certain tasks, allowing them to focus on higher-level, more creative, and more strategic work. Working with generative AI will be like having an all-knowing collaborator who can instantly help you explore and evaluate different directions on creative projects. With the help of generative AI, people will produce more creative content faster.
While it’s not clear how all this will play out, there are two market forces that seem almost certain with respect to content: (1) the amount and quality of per-capita content produced by organizations will increase, and (2) how people value this content will qualitatively change.
The second point deserves unpacking. With much more content that is easier to produce, at least marginally speaking, individual pieces of content will likely be devalued. However, there will probably be a long-tail effect here, and, on the whole, the collective value of content will increase. Just as Gates predicted about the internet, there’s a ton of collective value to be unlocked by generative AI through content creation. However, with AI able to help generate tailor-made content nearly on-demand, people are sure to change their internal metrics for how they evaluate content. How this will change is uncertain. Will people start to place higher values on things that are much more niche and specialized to their individual interests? Or maybe more people will start investing their time as content creators/consumers, bypassing the publishers and distributors to generate content right from the AI source for an audience of one? Or maybe a hipster market will emerge that places high value on AI-free “vegan” content? It’s all too soon to tell.
Content may still be king, but for the first time, humans are not the only kingmakers.
When I was at Microsoft Research, my team and I built an AI-based chatbot called Calendar.help that could schedule your meetings for you over email. The initial prototype did not have any AI. Instead, we built a multi-layered task-execution workflow, that split up the complex but tedious task of scheduling a meeting into manageable bite-sized chunks that a human executive assistant could easily execute.
Our research vision was the “assembly-line-ification” of knowledge work. We could make people much more productive by streamlining the routine work they do into simple, easy-to-execute micro-tasks, all the while training machine-learning models that could automate these micro-tasks using the data the platform gathered. Our plan worked. After running the beta internally for a while, we saved people a ton of time they would have spent manually scheduling and were able to fully automate the system.
When we first started the project in 2015, we had a grand vision of extending the system beyond meeting scheduling to other routine workplace tasks—things like documentation, task tracking and delegation, triage, summarization, information synthesis, and other common chores of information work that waste people’s time and energy but are not directly critical to their work. Although this vision was exciting, it took us much longer to automate scheduling tasks than we had hoped, and our bigger vision proved much too ambitious given the AI technology of the time.
However, today, with the latest advancements in generative AI, I believe we are on the verge of an “industrial revolution” for knowledge work. In the near future, we will start to see large language models in the workplace as the first station in a knowledge work assembly line—as a sort of conversational routing layer that can hand off your requests to the right internal human + AI systems. These new systems will free us from the many tedious tasks of knowledge work, so we can focus on higher-value strategic thinking.
Any organization is a collection of interdependent systems—systems of people, systems of software, systems of operations, and systems of processes. When different systems interact, there is always some degree of friction that can cause delays, errors, and inefficiencies, ultimately impacting the overall performance and success of the organization. Whether it’s collaborating across different divisions, moving information from one app to another, connecting different API endpoints, or composing interdependent workflows, it is often at these frictionful seams between systems where things fall through the cracks and work and doesn’t get done.
I’m sure you’ve been there before. Maybe you’re always forgetting to capture follow-ups in your task tracking app when your Zoom meeting ends, or maybe it’s every time you have to get your copy approved from legal before a public release. Each of these tasks is simple enough on its own, but they require you to change contexts and interact with entirely different systems, each with its own schemas, interfaces, rules, and cultures. This is where things break.
Widespread adoption of generative AI will act as a lubricant between systems, reducing friction and improving the ease with which work moves across systems. How would this happen? By synthesizing the activity that happens within a system, and transmitting it in a format that other systems can understand. For example, if your team is talking about a task in Slack, AI will be able to synthesize that discussion and automatically update your task-tracking system. Or when the marketing team notices a bug with similar symptoms to the one you’re trying to solve in a different part of the product, you should get notified about it since it’s relevant to your current task. It is just a matter of time before these types of frictionless cross-system workplace automations are commonplace.
The common thread of all of these transformations is the ability to listen at scale to the deluge of information produced by the modern workplace and make sense of it all. We’re building a product called Maestro AI that does just that. We’re starting with workplace chat, because, as we noted in another post, chat has become the de facto knowledge base for most teams and companies. Imagine turning your Slack into a super-smart AI-powered chief of staff that can answer any questions you had about what’s going on. This is our vision for Maestro.
We’re opening up our invite-only private beta to a handful of users this week. If you’d like a sneak peek at our approach, ping me on Twitter and I’ll get you signed up.