How spreadsheets can prepare us for artificial intelligence

How spreadsheets can prepare us for artificial intelligence

Zak Yudhishthu, Staff Writer

In the past few months, it has been hard to ignore recent advances in artificial intelligence (AI). Various content-generating AI technologies have been rapidly improving. Perhaps the most prominent has been ChatGPT and its family of personal chat companions, which write computer code, draft short essays and respond to most any query — although not always correctly. Other recently-improved AIs can use text prompts to generate just about any kind of image, or artificially replicate celebrity voices. Across these programs, people have generated a host of results that are alternatively wondrous, bewildering and dumb.

It’s hard to know what lies ahead for these platforms, which could accelerate at unimaginably fast speeds — or not. Progress in the last year has been impressive and rapid, while also sparking both deep public interest and sizable private investments in AI. Most anyone in the business of making predictions about AI will inevitably be somewhat wrong — leading AI researcher Eliezer Yudkowsky has warned that “by far the greatest danger of Artificial Intelligence is that people conclude too early that they understand it.”

We can, however, attempt to broadly anticipate how AI might reshape the world we inhabit. And for this, we should look to the history of today’s most boring digital technology: the spreadsheet.

I use digital spreadsheets for a lot of things. I’ve used them for school homework assignments, to track volunteer sign-ups and to look at government data. The Mac Weekly uses Google Sheets every week to allocate the pages of the print edition. As a tool, spreadsheets are often helpful and occasionally irreplaceable. 

Of course, things didn’t always function this way. In 1984, journalist Steven Levy wrote a story for Harper’s on the birth of spreadsheets in modern business. Over the early 1980s, people developed flashy programs on fat computers that offered a revolutionary change: performing computations on rows and columns of numbers. With this new technology, two shifts occurred: previously commonplace activities became silly, and people began to interact with the world in new ways. Each of these changes gives us a small lens with which to anticipate AI.

Before the 1980s, accountants and other business employees used to work with a dismal tool called “ledger sheets.” When accountants wanted to perform various business calculations, they would handwrite a spreadsheet — like Excel, but on pen and paper. Instead of using handy spreadsheet formulas, they would use calculators to compute every single sum or balance. If just one or two of the inputs changed — say, a monthly expense became cheaper, or banking conditions shifted — then an accountant would have to recalculate all of the ensuing values by hand. 

Digital spreadsheets made such manual operations completely useless. Instead of taking days to complete a ledger sheet, and then doing it all again with new information the next month, calculations could be made and manipulated in minutes. Thus, tasks that had long been part of everyday business activity were suddenly made useless. 

Spreadsheets didn’t just make existing tasks more efficient and easy, however. As spreadsheets replaced previous work, they also created new ways of interacting with and reading the world.

For example, spreadsheets allow a user to change a single number, while the rest of the calculations autocomplete accordingly. As a result, people did not just use this new technology to measure what was true today; they started using spreadsheets to try and make economic predictions about the future. Digital spreadsheets also allowed greater insights to be pulled from numbers, raising the relative value that people placed on quantitative information over qualitative insights. 

It’s hard to predict exactly how and where, but AI seems sure to do both of these things. Already, it seems like AI has put various everyday tasks on the precipice of frivolity. We might soon be rethinking typical college papers or take-home exams, where AI’s performance is currently decent and sure to improve. As AIs become better writers, coders and designers, more everyday tasks will begin to seem unnecessary.

As with digital spreadsheets, AI’s displacement of some jobs and tasks probably will not leave us without jobs, but instead with new and different things to do. AI will become more powerful, and more institutions will adopt it, creating unfamiliar ways to approach tasks and carving fresh forms of legibility into our world. As the media theorist Marshall McLuhan wrote, “it is the medium that shapes and controls the scale and form of human association and action.” By the very nature of these changes, they are impossible to precisely predict, and they will surely occur in ways both beneficial and harmful.

Analogizing spreadsheets to AI does not give us a full picture of the potential shifts. AI already seems likely to be far more transformational than spreadsheets, and to enact its transformation much more quickly. In the next few years, we might find ourselves living through what journalist Ezra Klein calls “the difficulty of living in exponential time.” At the current pace of technological development, we may soon outrun our existing institutions — both our hard-programmed rules, and our soft forms of social organization — leaving us with a more uncertain and strange world.

More darkly, there’s another technology that we might also find analogous to AI: the nuclear bomb. As neuroscientist Erik Hoel recently wrote, human society has not encountered very many situations in which there was a true risk of mankind’s total elimination — perhaps just the nuclear bomb or climate change. In a 2022 survey, a survey of leading AI researchers showed a median prediction of a 10% chance that AI would cause human extinction. These machines are not only powerful and advancing at a dizzying pace, but being funded by governments and corporations with far more incentive to win the technological race than to carefully weigh societal costs and benefits.

Hopefully, histories of technological advances and societal adjustments can help prepare us for what lies ahead, but advances in AI will nevertheless operate in strange and unexpected ways. Nothing is inevitable — our political and economic conditions have massive power to shape AI and its impacts. But be prepared for it to shape us back.