Creating at the speed of thought
How large language models will mostly just reduce our tolerance for typing and swiping
Science fiction author William Gibson is said to have used the phrase
The future is already here — it’s just not very evenly distributed
and when it comes to AI, both the topic of our best possible future and the topic of egalitarianism are hotly debated. There is the question as to the value of AI and to what degree it us under or overhyped. There is the question of who AI is for. There is the question of how it will impact the future of work. Often there is a suspicion that big tech is shoving AI down our throats.
But how can it help? How can it help in a way that doesn’t just matter to nerds?
AI or machine learning more generally have, you might say, been “quietly” influencing industry for decades (or more) in areas like healthcare, agriculture, decision making, finance, energy etc. These more general forms of machine learning often focus on statistical models for prediction and data analysis. They often work behind the scenes or surface in dashboards or spreadsheets.
This new flavour of generative AI has, in some sense, sucked the energy out of the room in terms of the diversity of research and attention but it has certainly got people talking. And mostly its due to how easy it is to anthropomorphize an AI that is so good at language and so good at generating grotesquely pretty pictures.
Simply put, it is the ability to interface AI with our natural language and our day to day that makes AI so compelling right now. I think its actually more about UX than intelligence. Its more about how it touches us and our lives.
The value has been noticeable in some areas more than others. Many argue that generative AI is not that interesting and only certain folks like coders care about it. Being in that group, I have to wonder about that. Much of the research and benchmarks focus specifically on math and coding and this is one of the strengths and one of the killer apps of generative AI.
But I think we are merely the early adopters. If (as a coder) you have experienced the before and after of having an idea and needing to build it either in a language you know well or not at all, versus having the language model generate code that actually runs from just a high level description… its very hard not to be impressed and very hard not to be excited about the future. This really changes everything. The feeling you get is that you can write code at the speed of thought. Its not that the language model is replacing you, its that its typing faster than you and remembering to add some things like what we call defensive programming, the extra checks we would have added on our second or third pass.
The feeling you get is that you can write code at the speed of thought.
So the question now is, will this feeling of writing at the speed of thought belong to coders alone or will it be something experienced broadly? Maybe there is another question here - if instead we should be asking if AI should be used in this way at all?1 Maybe if we go too far from left brain math and coding, the nuance required in document creation makes this particular flavor of auto-complete on steroids uninteresting?
I googled “creating at the speed of thought” which has been top of mind in my own work and I found an article by Steven Johnson on the NotebookML product, which is one of the best examples of generative AI applied to assisted research and writing -
He touches on interesting points about workflow. Take for example the following -
If you do it the old-fashioned way, using Spotlight on the Mac or Drive search, it goes something like this: type in a query like "Jacobs Gordon ant colonies" and that will bring up a list of suggested documents that might be a match. Then you open those one by one, but at that point you have to command-F with exact key strings: "ant colonies" maybe, or "Gordon." You go through a bunch of documents and eventually you find one of the quotes. You copy it, switch tabs to Bard or ChatGPT, and paste it in, and then go back to searching by hand through your documents, looking for the second quote. Once you find that one, you paste it into the chatbot prompt window with the first quote, and add a line in the prompt requesting a short essay connecting the ideas in the two quotes.
That's what it takes to execute that particular task the old way. I ran the test myself and it took 5 minutes and 8 seconds of genuinely annoying work.
I believe this feeling will be experienced broadly as we bring AI closer to people’s information and their workflows. This is a question of data integration and UX and not a question of the capability of a language model to model other domains or a question of less technical folks than coders being able to benefit.
ChatGPT, the generative AI interface and modality, is more like an evolution of the thing on the left than a step towards the thing on the right. Its a better UX.
I think its pretty simple. Firstly, I have in mind knowledge workers that are already using touch screens and keyboards for their specific jobs to be done. The thing is, there is a genuine tension that exists between having an idea to create something and the limitations of our computing tools in bringing that thing into the world. There is no reason why we want to swipe or type more that we need to. Such human-computer interactions are not making our lives better. They are not making us better humans.
Generative AI is in many ways just a better interface in those contexts in which we already work. It may take awhile for people, generally, to realize how much better it can be. It will also take awhile for UX designers to learn how to rethink the interfaces that we work with today within this new paradigm.
Beneath or beyond UX, we can already see what these models are capable of today. That is why we say that; the future is here but its just not evenly distributed. But, its just a matter of time. (Generative) AI is better search. It is better contextualized explanation. It is certainly better auto-complete. Its sometimes a better critic of your work than you are. Generative AI is about changing the interface for knowledge work. Its about removing an unnecessary tension in the creative process and removing time spent typing or swiping.
Its about creating at the speed of thought.
We should always be asking in what way tools support our humanity