We talk about so many things when we talk about AI. The conversation can roam from self-driving cars to dynamic video generation, from conversational chatbots to satellite imagery object detection, and from better search engines to dreamlike imagery generation. You get the point.
It gets confusing! For laypeople, it’s hard to nail down what AI actually does (and doesn’t) do. For those in the field, we often have to break down and overspecify our terms before we can get to our desired conversations.
After plenty of discussions and tons of exploration, I think we can simplify the world of AI use cases into three simple, distinct buckets:
Let’s break these down, one by one.
Gods are the human-replacement use cases. The much hyped artificial general intelligences (AGI) that are allegedly just around the corner…
(Or maybe they’re just imminent when you’re trying to close one of the largest funding rounds in history? After the round is finalized, you can go back to realism. But I digress…)
Gods require gigantic models. Their autonomous nature – their defining quality – means a low error tolerance and a broad, general model. But a giant LLM surely isn’t enough. We could throw every data center and every database at the problem and still be short. We’re going to need breakthroughs, oodles of researchers to find these breakthroughs, and billions of dollars (at least).
Few entities will work towards the AGI or god use case because the barriers are so high. But, so are the rewards. The massively funded private ventures and sovereign-backed labs will continue their work.
But for most of us the Gods use case will most impact us in the form of hype and hand-wringing. It will affect funding, regulation, and politics. (Though if a would-be god-builder visibly hits a wall, we’ll feel it in other ways.)
Interns are the copilots (and a term I’ve shamelessly borrowed from Simon Willison). Their defining quality is that they are used and supervised by experts. They have a high tolerance for errors because said expert is reviewing their output, and can prevent embarrassing mistakes from going further. Interns are focused on the grunt work: remembering documentation and navigating references, filling in the details after the broad strokes are defined, assisting with ideation by being a dynamic sounding board, and much more.
Github Copilot and Cursor are interns for programmers. Adobe Firefly and Visual Electric are interns for designers. Microsoft Copilot and Grammarly are interns for writers. NotebookLM is an intern for researchers. Monterey is an intern for product managers. There are many more examples.
Adobe’s Project Turntable – a utility for rotating 2D designs as if they were 3D objects – is a perfect example of the intern form. Jess Weatherbed, writing for the Verge, sums it up:
The tool allows users to click a button and then drag a slider along to automatically view and snap a vector image into a different viewing perspective — something that would typically require an artist to redraw the image entirely from scratch. The examples demonstrated at the event retained their original designs when rotated without warping into a new overall shape.
The work being done here – redrawing the vector drawing from many different angles – is the type of grunt work an intern at a design studio might perform. The look and feel of the design and its proportions are already set. The work required is tedious but easily defined. Further, the output of this tool is editable by the expert using it. If Turntable is sloppy with a line or two, the artist can drop in and tweak it.
Because they are tools for specific types of experts, Interns are limited to specific domains. They don’t have to be generalists. Their model sizes are large, but not massive. One could build new interns – starting from scratch – for millions. If you use an open model, the costs are dramatically lower.
Today, Interns are delivering the lion’s share of the realized value from AI. Engineering copilots alone are delivering massive returns, increasing the output and capabilities of expert programmers. I’ve heard of a few large companies who have slowed down hiring due to their effects and have witnessed many tiny teams ship products years beyond what they could have delivered even 3 years ago, thanks to Github Copilot and similar tools.
And while Interns are delivering tremendous value, they are secondary to the experts driving them. How do you improve the output of an AI copilot? Simple: find a better expert.
“But Drew,” you might say, “what about the fun image generator or friendly chatbot I play with occasionally? It doesn’t really fit in the intern bucket.”
You’re right. I think these are Toys, a subcategory of Interns defined by their usage by non-experts. Toys have a high tolerance for errors because they’re not being relied on for much beyond entertainment.
To improve a toy you don’t usually need to improve the user or the model; rather, the opportunity is usually the interface. We, as a community, just tend to focus on the models because we’re nerds.
But moving on…
Cogs is the last use case bucket. Cogs are comparable to functions. They’re designed to do one task, unsupervised, very well. Cogs have a low tolerance for errors because they run with little expert oversight.
Cogs may be built into data pipelines, performing some enrichment or transformation step, alongside bog-standard functions powered by regex or SQL. Cogs may be built into interfaces, transforming human output like speech or scribbles into machine-readable input.
Cogs exist in systems as reliable components. Their focus on one task and a low tolerance for errors mean they are usually built with fine-tuned or heavily prompted small, open models. Cogs are cheap to run and relatively cheap to ship. They are enabled and improved by data, used for fine-tuning and/or developing prompts.
We used a cog here recently! Ollama and an embedding model powered a step in our conflation pipeline, nestled among some SQL queries and string distance functions. This cog was cheap, quick, and reliable after some initial configuration.
Cogs are, by far, the dominant use case amongst enterprise teams building with AI. As we covered after the DataBricks summit, in most discussions among builders, “AI looks like just another pipeline function.”
Cloud platforms – DataBricks, Snowflake, Azure, AWS, and others – have recognized the dominance of the cog use case and have sprinted towards delivering tools and hardware for building, testing, and running them.
Sorting the ocean of AI use cases into the Gods, Interns, and Cogs (and fine, Toys) buckets has helped me tremendously. It’s easier to navigate the noise, ask questions about new products, and identify the bottlenecks holding back each category.
It’s challenging to work in emergent fields as the language has yet to settle. Communication challenges are a tax that must be dealt with before productive conversations can occur. By scaffolding the mess with appropriate structures we can ease exchange.
When I was 7 years old, living in a flat overlooking Hamra street in Ras Beirut, I read The Hobbit. I fell in love with it. I memorized all the songs and made up tunes to them; I memorized all the riddles and asked them of whoever would listen; I made up my own adventures in Mirkwood, my own encounters with Gandalf and Beorn and the Elves. I also read everything I could about Tolkien, and went in search of anything else he'd written. I decided that I too would be a writer, and that I would start with poems and work my way into fiction the way he did. In many ways I can trace much of my life's trajectory to that encounter with a single book at a delicate age — a time when all the world's paths are laid out before you, and you wait for someone or something to beckon you on to one instead of another, into one self rather than another.
I say this because almost 20 years later — sitting on my bed in a cold, damp room in Cornwall, floundering toward the end of a second graduate degree — I read Naomi Mitchison's Travel Light, and suddenly felt as if I were seeing my life thus far from a great height. I felt, very powerfully, that I had been waiting for it, and that it was telling me the story of the person I might have been had I read it when I was a child.
Travel Light is the story of Halla, a girl born to a king but cast out onto the hills to die. She lives among bears; she lives among dragons. But the time of dragons is passing, and Odin All-Father offers Halla a choice: Will she stay dragonish and hoard wealth and possessions, or will she travel light?
Contrary to appearances, this is not a simple book with a didactic moral at the end. It begins as a fable but resists conventional fable structures, moving from mythic to recorded history in the way the sun moves across the sky. In the dawn of the book, Halla is called Bearsbairn and Heroesbane; at noon she is Halla Godsgift; as evening draws on, her earlier names cast shadows over her narrative. The story shifts smoothly from bear-nurses to the heart of Constantinople, from the dragon-imparted economics of sheep and princesses to the intrigues of the Holy Roman Empire. And through it all Halla remains Halla, changing from protagonist of her own story to miracle of someone else's, but always and utterly herself.
I had never encountered Mitchison's work before reading Travel Light. A cursory Googling revealed, to my astonishment, that there were good reasons for me to think of this book and The Hobbit as two sides of my heart's coin: Mitchison and Tolkien were good friends, and Mitchison was among the first readers of The Lord of the Rings before it was published (Travel Light was published in 1952, Lord of the Rings in 1954). Reading further I discovered, to my astonishment, that Mitchison had written more than 90 books, that she died in 1999 at the age of 101, that she had led a spectacular life full of travel and social activism, that she had written science fiction, historical fiction, nonfiction and poetry — and that she was nowhere to be found in the canon of genre fiction. Here was a woman who had, in Travel Light, almost certainly written certain points in conversation with The Hobbit, and more: In her Memoirs of a Spacewoman (1962), she wrote in the voice of a character who pursued space exploration that privileged communication over conquest; looked at free love, birth control and child-rearing with delight; and seriously considered Islam as a viable religious choice for herself. Naomi Mitchison was imaginative, progressive and astonishing, but in the course of three English degrees — almost 10 years of studying literature — I had never even heard of her.
That Mitchison's life and works should have been so unfairly relegated to secret history drove home my feeling of books as points of divergence to alternate timelines; that having read The Hobbit rather than Travel Light at that fragile, formative moment of being a child in Lebanon standing at a crossroads of languages, religions and literary traditions nudged me into a different life. Who might I have been if I had met Halla Bearsbairn before Bilbo Baggins? How different might my attitude toward dragons have been if I'd met Uggi before Smaug? How different would the spiritual landscapes of fantasy and science fiction be if they had accepted as antecedents works that showed a corrupt Byzantine Christianity and sympathy toward Islam?
But, most crucially for me, I wonder: Where might I have gone if, instead of a middle-aged Hobbit enamored of his pantry, I had embraced a girl who lost three homes before choosing the open road?
I don't regret, at all, having The Hobbit at the core of me, and will defend its songs and riddles and elves and spiders to the end of my days. But reading Travel Light unseamed something in me, made me feel that my certainties needed revisiting, and assured me that somewhere within me was, still, a 7-year-old girl waiting to be beckoned onto a path of luggage-less travel, of dragons and Valkyries, languages and air — and that with Travel Light, she'd taken the first step in their direction.
Following Donald Trump’s victory Tuesday night, I’ve seen calls on social media to support independent news organizations like ProPublica and The Guardian rather than traditional outlets. It’s a good idea, though I value the work done by mainstream journalism as well.
But let me suggest a different approach to funding media: using your subscription money or tax-deductible donations to support news at the local level. I’ve been writing about the local news crisis for a decade and a half, and during that time I’ve come to believe that one of the reasons we’re so polarized is that low-quality national news has moved in to fill the vacuum created by the decline of community journalism.
Civic life depends on reliable news and information. Without it, you have people showing up at school committee meetings to complain about phony, Fox News-driven issues like transgender sports and critical race theory rather than test scores and the cost of funding a new teachers’ contract.
Academic studies have shown that a lack of local news leads to fewer people running for local office, lower voter turnout, measurable increases in polarization and what my research partner, Ellen Clegg, and I like to call the “corruption tax” — that is, lenders demanding a higher rate of return when municipalities in news deserts seek to borrow money for such worthy causes as a new middle school or fire station. The lenders, it seems, want a premium if no one is going to keep an eye on how their money is being spent.
Rebuilding civic life is a way of lowering temperatures and encouraging cooperation. When people learn they can work with their neighbors to solve local problems even if they hold different views about national politics, that enables them to see those neighbors as fully rounded human beings rather than as partisan Republicans or Democrats.
The news desert problem is serious and getting worse. According to the latest State of Local News report from Northwestern University’s Medill School, some 3,200 print newspapers have disappeared since 2005. Most of them were weekly papers that provided exactly the sort of coverage needed to build and maintain a sense of community.
At the same time, though, hundreds of independent news projects have launched in recent years. Most but not all are digital-only; many are nonprofit, some are for-profit.
As it happens, this is the time of year when it makes the most sense to support local news, especially nonprofits. Every year, the Institute for Nonprofit News, through its NewsMatch program, provides funds to nonprofits to match some of what they are able to raise within their communities. This year’s campaign began Nov. 1. As INN explains:
Eighteen national and regional funders have pledged $7.5 million to NewsMatch, the largest grassroots fundraising campaign to support nonprofit news in the U.S. Since 2017, participating news organizations in the INN Network have leveraged $31 million in NewsMatch funding to help generate nearly $300 million in support from their communities. All of these newsrooms have met INN’s membership standards for financial transparency, editorial independence, and original public service reporting. Not every nonprofit news outlet meets those standards and is able to become an INN member.
Ellen and I wrote our book, “What Works in Community News,” to profile independent local and regional news organizations that are finding ways to serve the public despite the ongoing financial challenge of paying for journalism. We also talk with news entrepreneurs and thought leaders on our podcast, “What Works: The Future of Local News.” Our hope is that the people and projects we highlight will inspire others to fill the information gap in their own communities.
Philanthropy will remain an important source of funding for some time to come. We should assume that long-stalled federal efforts to provide tax credits for local news aren’t suddenly going to start moving forward during the Age of Trump II. Efforts in states that include New York, New Jersey, Illinois and California are worthwhile but limited.
Ultimately the news desert problem will be solved, or not, without government assistance. If your community has an independent news outlet, please support it. And if it doesn’t, I suggest you look into what it would take to get involved in starting one.
When my Dad was in surgery, I remember sitting in the hospital waiting room coding a work project. I didn’t have to – my boss made it clear that I should take as much time off as I needed and not worry about work. I knew my co-workers had my back. And I wasn’t trying to score points on the job.
But I needed something to take my mind off my father in the operating room. Tapping away on my laptop helped pass some time. As stressful as my job could feel at times, that stress was tiny compared to a loved one’s health emergency.
Dad pulled through that incident, and I mostly put work aside once he was in recovery. But even then sometimes I’d check into work. It was a way to feel tethered to “normal life” when real life felt especially scary and unpredictable.
In other words, somewhat counterintuitively, a stressful job can occasionally be useful when navigating hard times. But is that a good reason to keep working? Or does it make sense to try to craft a life where you find other ways of coping? In my case, I want a new path – although I’m not passing judgment on people who find their work helpful in difficult moments. It can be a constructive coping mechanism.
You can follow My Next Chapter by email newsletter or RSS feed. Blog content © Sharon Machlis.
Data dashboards are now one of the standard storytelling templates at reuters.com, joining the familiar journalistic arsenal of text articles, photo galleries and blogs. Going forward, publishing a live-updating collection of charts, maps and tables will require nothing more than a few points and clicks.
First through the gate is "The Magnificent Seven Monitor," which tracks the ongoing performance of the heavyweight U.S. technology stocks that dominate the market.
Anyone who has toiled in the vineyards of news knows it's hard to integrate custom designs like this into the rigid, cookie-cutter "content-management systems" that power most media sites. As a result, many data projects and long-form investigations are published off-the-reservation using shadow technology.
While there are advantages to working outside your core platform, there are downsides, too. Every new publishing tool must tackle how to integrate with your site design, your homepage, your recirculation system, your ads, your analytics, your paywall, and the workflow of your colleagues. That's a lot.
At Reuters, we aim to bridge the divide once and for all. That requires treating the innovative story forms pioneered over the last twenty years—multimedia stories, interactive visuals, live-updating data—like the other widgets we manufacture in our "CMS."
Today's release is a modest first step in that direction. Now that we have the technology issues sorted out, we'll do more and better in the months ahead.