My obsession with “remembering everything” began in 2012, when I came across an article on The Verge about backing up your brain with cloud note-taking software. I think that the article caught my eye because everybody always tell me that I have a terrible memory (sadly – it’s true). If I was late for work every morning because I was forgetting where I put my keys, what chance did I have to remember to cite all the sources for that article due on Friday? This system promised to supercharge my brain for the digital age.
The idea of this ambitious project, in its most basic terms, was to map your brain into a computer database. More specifically, this brain-map (let’s just call it that) was to consist of a series of Evernote notebooks in which you recorded everything that you read, wrote, or wanted to remember. If I ever needed to remember an idea for a blog post or an article, I would be able to trace this idea to its point of origin – a movie I watched, an article I liked, or the introduction of the book that I just skimmed.
In practice, creating this brain-map was hard. The process required, after all, that I record anything and everything, no matter how tedious or seemingly unimportant. Up until around 2016, I diligently kept plugging information into my notebooks. I hoped that I could “teach” my computer my thoughts and memories, and that eventually, with enough data and usage, it would “learn” my habits and help me retrieve the information when I looked for it in the future. In hindsight, I suppose I anticipated a personalized search experience before machine learning became mainstream.
Looking back at this four-year collection, I see that it consists of over 4,000 discrete notes. The range of material is mighty. Some notes are summaries of book chapters, while others are lists of my favorite movies or movies I plan to watch. Mixed into this noise, I have photos of artworks that I particularly liked, shopping lists that I’ve completed in the past, PDFs and other materials for posts that I’ve written, audio and video recordings of meetings or events – the list goes on and on.
If it was the information that I needed to finish the Wednesday brief or if it was simply a recipe that I cooked last Tuesday, I could, in theory, find whatever I needed using my computer’s search functionality instead of relying on my own fallible human brain. At least that’s what I thought.
Did it work?
Yes and no. Since about 2015, the hype around productivity and note-taking software died down. Evernote fell on hard times. It was only after a new CEO, a corporate restructuring, and several application updates that the company became cash-flow positive. At the time, the app was bloated, slow, and often dysfunctional. Things were looking bleak for my digital brain.
Apart from the downturn for Evernote, I realized, too, that my mind-map was utterly useless if it didn’t retrieve the information that I wanted it to retrieve. After all, the biggest problem that I had with my own memory was its inability to perform information recall.
When you’ve learned something but are unable to retrieve it, you’re experiencing something that psychologists call “retrieval failure.” At its very core, retrieval failure means that the information is in there, somewhere; you just can’t access it anymore.
Perhaps somewhat naively, I thought that I could use software to treat my bad memory. In our information-centric world, I hoped that I could offload memory-based decisions – what to remember and what to forget – to a computer system. What I quickly realized, though, is that my system struggled with determining what was relevant in the first place. It’s one thing to have a lot of data – it’s a whole other problem retrieving the right data at the right time and in the right context.
Here’s an example of the system not working: when I conduct a search for “pasta,” it’s probably obvious that I’m looking for a recipe. But what about when I search for a word like “bank”? Am I looking for my bank statements, or am I looking for the “10 Best River Banks in America?” My brain-map couldn’t tell the difference.
The technology just wasn’t there in the early 2010s. I now realize that I gave up on my brain-map due to a poor search experience: one that locked away my data in a vault rather than a filing cabinet. The software, instead of helping me eliminate “retrieval failure”, caused more problems than it solved. I was getting frustrated when I would enter terms into the search box and getting completely irrelevant results, so I quit entirely.
Looking to the future
I don’t want to give up on my dreams of a better brain. AI-powered search could, in the near-future, be the answer to the original demands that I had for my own note-taking system. Evernote just announced at SXSW 2018 that they are experimenting with machine learning solutions that will, in the future, “deliver better search results and suggest relevant tasks.”
If I had unlimited resources, I would set up AI-powered search on my old brain-map and see how an added relevance engine can improve document retrieval and help boost my productivity at work – all the while helping me sift through the ocean that is today’s flow of digital information. Maybe this is why I was drawn to working at Coveo: I want all of my devices to intuitively know what I’m looking for.
Thinking even further into the future, I can imagine a system that would replicate memory retrieval even more closely. Text-powered search is still limited by language and by my own ability to plug in the right “cues” (read: words) to find what I want. As weird as it sounds, I don’t want to be limited by my own brain or by my language. Computers remember and store everything perfectly, so why not build an algorithm that reconstructs – or attempts to reconstruct – my original intent for a particular note or file? Why not automate the process of “searching” itself?
Ideally, computers are here to help us: they should be able to look past human limitations and, in turn, help us work through our own cognitive barriers. Machine learning should be able to adequately assess the context of my needs – let’s say that I’m working on an article – and deliver documents and information in a way that circumvents search altogether. In fact, I no longer want to “categorize” and “tag” anything at all. Classification is a computer’s job: I look forward to the day that AI-powered algorithms automatically inject relevance into all of my software, saving the time I spend organizing and giving me accurate results, every time. Until that happens, though, I think I’ll still be waking up every morning to find that my keys have mysteriously gone missing.
If you’re interested in hearing more about the ways in which Coveo has changed search using machine learning, check out the video below.