With the release of tools like DALL-E and ChatGPT, Google — once the presumptive industry leader in artificial-intelligence research — suddenly finds itself playing catch-up. The rise of OpenAI was reportedly a “code red” emergency at the company, which had quietly been working on, but not really releasing, similar tools for years.
Now, quickly, Google is retrofitting its product line with AI. Last month, it demonstrated its take on a chatty version of its search engine. Yesterday, it shared more details about AI-assisted Gmail and Google Docs. In Gmail, there are tools that will attempt to compose entire emails or edit them for tone as well as tools for ingesting and summarizing long threads. In Docs, there are tools for generating text from simple prompts or other content. A lengthy email discussion is turned into a “brief” and then a slide deck, which is then illustrated with generated imagery.
Google’s promotional video is worth watching for a few reasons. In contrast to tools like ChatGPT, which are quite capable but function mostly as general-purpose tech demos and marketing tools, it’s an example of how a major firm thinks software powered by a large language model, or LLM, will change its existing business and products, some of which you might already use — these features are where the new class of AI tools will face a test of their real utility.
Google, in its eagerness to show off a scattered range of new features, has also unintentionally done something sort of funny. After automating the creation of a marketing campaign with no participation from anyone else, our host ends the video by sending her team a machine-generated thank-you note. Thanks for what? Google makes the decks now and writes all those client emails. I’m sure it’ll be able to write a solid layoff notice, too.
Most interesting are the ways in which these features seem to be in conflict with one another. On the one hand, you’ve got a bunch of ways to generate text easier and quicker: Tools for composing full-length emails from prompts, which is to say turning a small amount of text into a lot of text. Tools for suggesting full responses to emails rather than the already popular automatic replies that Gmail users have been using for years. Tools for generating customized emails for a spreadsheet full of recipients. Tools for generating lots and lots of text. For people whose job involves generating lots of content, or perhaps for whom the creation of lots of content is a convincing performance of work, this has obvious possible uses — if, of course, it’s any good at what it does.
On the other hand, for those who might interact with people who have these jobs — that is, those who can expect to be on the receiving end of this plentiful new content — these features read a bit differently. Are you excited for your co-workers to become way more verbose, turning every tapped-out “Sounds good” into a three-paragraph letter? Are you glad that the sort of semi-customized mass emails you’re used to getting from major brands with marketing departments (or from spammers and phishers) are now within reach for every entity with a Google account? Are you looking forward to wondering if that lovely condolence letter from a long-lost friend was entirely generated by software or if he just smashed the “More Heartfelt” button before sending it?
These features, in other words, solve one set of user problems while running the risk of creating others. For example, in offices already burdened by inefficient communication and processes, it’s easy to see how reducing the cost of creating content might produce weird consequences and externalities. Tim can now send four times as many emails as he used to. Does he have four times as much to say?
Don’t worry, though — Google has AI tools to handle that, too: Tools for summarizing long emails and threads. Tools for shortening content. Tools for ingesting and filtering out the massive quantities of material generated in the world, by people or maybe machines, and synthesizing it into something useful, actionable, or just legible.
In the context of Gmail and collaborative documents, we see suggestions of automation processes at war with one another, feeding problems that must be solved with more automation as Google manufactures demand for its own mitigating products. It’s an arms race in every inbox! It’s textual hyperinflation in every office! It’s a hundred meetings a day scheduled and attended and summarized by bots! Before Google productized this vision, OpenAI’s Sam Altman joked about how ChatGPT users had discovered it themselves.
As these features roll out, Google — and Google’s customers — will start to figure out how people actually use them and which ones catch on. The most likely outcome is that such tools will reveal and exaggerate existing office dynamics, reducing the institutional value of certain tasks (cranking out a deck nobody is going to pay attention to anyway) while emphasizing the value of others (coming up with subjects worth making decks about). There’s no need — or, rather, no time — to game it all out. Google, following the example of OpenAI, is rolling this stuff out right away. We’ll see what happens.
This isn’t entirely new territory. Social-media communication is already substantially mediated by AI, and its example (platforms that address their own content glut with AI-powered algorithmic feeds) contains some suggestions about how this stuff might play out elsewhere or at least how it might feel. Lots of email services already attempt to prioritize messages judged to be urgent or relevant; it’s not hard to imagine that a sorted inbox could be replaced by a written digest or to foresee how such a change could affect the ways people write their emails — for consumption first by the machine that stands between users and their audiences.
Similar tensions have defined Google’s core products for years. For the past two decades, websites have been accommodating the needs of Google’s automated indexing and ranking algorithms, some by following Google guidelines in the process of producing the best and most relevant content possible and others by simply flooding the web with material that’s close enough to snag some search traffic. A great number of them — including basically every major publisher and e-commerce site — work somewhere between the letter and the spirit of Google law. Gmail users have watched a related dynamic play out in their inboxes for years as substantially automated and increasingly convincing spam and marketing have been thwarted, or at least managed, by corresponding automation on Google’s side in the form of filters and inbox-sorting tools. (As for what this long technological battle or dance has meant for the utility and role of email and for the experience of communicating with one another using computers, I leave that up to the reader. It’s not a simple story! And apparently it’s not over.)
We’re going through a spectacular period of hype around AI as users encounter genuinely bracing technologies en masse, each embedded with fresh possibilities and promises: to make work easier or more productive; to give you an edge, all else being equal. Our encounters with anthropomorphized machines emphasize private experience and a feeling of ownership as they engage with us in conversation and answer direct commands. These, mostly, are tools in demo mode, pre-deployment and pre-monetization, often offered for free, at a loss, in guises intended to inspire wonder and enchantment. It’s great but ultimately misleading marketing. All else is not equal. Everyone else is going to be using these things too.
Google, here, has done us a favor by offering a glimpse of what might come after the hype dies down: A concerted upgrade, or at least a change, to the productive machinery with which millions of people interact every day. A top-down change in how people work, carried out one software suite at a time. A thank-you email from a manager who doesn’t need you anymore, written by a machine.