Nieman Foundation at Harvard
HOME
          
LATEST STORY
Newsweek is making generative AI a fixture in its newsroom
ABOUT                    SUBSCRIBE
Feb. 7, 2023, 11:51 a.m.

Google now wants to answer your questions without links and with AI. Where does that leave publishers?

A dozen years ago, Eric Schmidt forecast the AI pivot that’s playing out this week. And the questions it prompts — around the link economy, fair use, and aggregation — are more real than ever.

If you stick around long enough, even the most exotic future turns into the present.

At least that’s what I thought when I remembered an old piece I wrote here at Nieman Lab all the way back in 2011 — one that seems newly relevant today.

It was about an on-stage interview that then-Google executive chairman Eric Schmidt gave to Walt Mossberg and Kara Swisher at what was then the annual All Things Digital conference. They cover a lot of then-hot topics, from Google losing to Facebook on “the friend thing” to whether it should buy Twitter. But there’s one stretch that stood out, then and now, as having outsized importance to news companies.

Mossberg started out by ragging on the declining quality of Google’s search results:

Speaking as a consumer, I find my Google search results to be more and more polluted with junk that I don’t want to see or that doesn’t seem relevant.

Is your basic algorithmic approach — which was so successful and so different from everybody else, in PageRank — still the right way to go?

Or is there some opportunity for somebody to come in and do to you what you did to Altavista?

(Altavista being the web’s leading search engine circa 1998, before Google ate its lunch.)

Schmidt defends the quality of Google’s results, but then pivots:

But the other thing that we’re doing that’s more strategic is we’re trying to move from answers that are link-based to answers that are algorithmically based, where we can actually compute the right answer. And we now have enough artificial intelligence technology and enough scale and so forth that we can, for example, give you — literally compute the right answer.

Anything sound familiar there? Google has built an empire by “polluting” its search results with ads. Its defenses may seem impenetrable, but it has one potential point of weakness: a competitor who’d be able to give users answers rather than links to websites that might have answers.

At 1 p.m. ET today, Microsoft is holding a press event at which it is expected to unveil a new version of its search engine, Bing, that incorporates ChatGPT, the fairly astonishing conversational, generative AI unveiled late last year. If a leaked version is to be believed, the new Bing is built around this promise: “Ask real questions. Get complete answers.” Along with traditional search results, you’ll also be able to chat with “your AI-powered answer engine.” The new Bing is part of Microsoft’s massive $10 billion investment into ChatGPT’s maker, OpenAI.

Google could see this coming, of course, so it tried to get ahead of the news yesterday by announcing its own AI-powered chatbot, Bard. It’s available to “trusted testers” now, becoming “more widely available to the public in the coming weeks.”

(Amusingly enough, even back in that 2011 interview, it was Bing that Mossberg argued was already ahead of Google in giving direct answers to search queries. Schmidt even agreed, “in some narrow cases.”)

This is all extraordinarily exciting stuff — the most important shift in consumer technology since the iPhone, I’d wager. In my nearly 30 years on the Internet, I’ve never been less confident predicting what it’ll all look like in a year or two. Will these new interfaces simply mean a modest improvement in 5% or 10% of our search requests? Or will AI become the dominant way we interact with information online? At the rate of improvement we’re seeing — “the scale of the largest AI computations is doubling every six months, far outpacing Moore’s Law” — putting any cap on the possibilities feels perilous.

But there’s a more narrow matter, of interest to the news business, that we can foresee. Right now, in lieu of those direct “answers,” Google sends a lot of traffic to news sites — more than any other external source by far. Newsrooms spend time and money figuring out a lot of those answers, and the reward they’ve traditionally reaped for them has been web traffic, monetized through ads.

But the more questions Google answers without a click, the less traffic those “answering” news sites will get. Less traffic means fewer ad impressions, which means lower revenue.1

As I put it back in 2011:

The kind of shift Schmidt is talking about — from “link-based” to “algorithmically based” — could have a big impact on publishers in the business of providing answers to searchers’ questions. And not just the Demand Medias of the world who are attached at the neck to search — traditional publishers too.

In 2011, these were the sort of questions that Google felt confident enough to answer without a link. They were all based on structured data, mostly hard numbers.

But even then, it was clear that their range would grow:

For now, Google’s ability to answer questions directly is bound by the sorts of things its algorithms can know. But they’ll get smarter — and Schmidt’s comments make clear it’s a strategic goal of the company to ensure they get smarter. So imagine a point in the near future where Google can give direct answers to questions like:

What time is Modern Family on?

Who are the Republicans running for president?

What red blouses are on sale at Macy’s?

Who’s the backup left tackle for the New Orleans Saints?

Modern Family is long gone, but Google will tell you directly when Abbott Elementary is on. And it’s happy to let Macy’s buy some ad space to talk about their red blouses. But it’ll still rely on links for G.O.P. candidates and the Saints offensive line.

ChatGPT, in its current public form, isn’t connected to the open internet and is only trained on information up to 2021, so it isn’t great on current events. (A rival generative AI, from Anthropic, tried to tell me Terron Armstead plays tackle for the Saints. Alas, if only.) But both Google and Microsoft have said their bots will be able to draw on the entire web, in real time — so their results will likely be strong.

Let’s go back to 2011 for a minute. Things like TV schedules, team rosters, and announced presidential candidates are still structured data — the sort of thing that didn’t seem too far out of reach even then. But what about the next level of difficulty?

And imagine, further down the line, that Google’s AI improves to the point where it can answer questions like these:

Did Dallas city council approve that zoning change last night?

Was the stimulus package too small to be effective?

What’s going to replace the Space Shuttle program?

Which is Terrence Malick’s best movie?

Did Osama bin Laden really use his wife as a human shield?

Is the new My Morning Jacket album any good?

Some of those are complex enough that Google probably wouldn’t be able to give a single definitive answer, the way it can with a database of census data. But it’s not hard to imagine it could provide a Metacritic-like look at the summary critical opinion of the My Morning Jacket record, or an analysis of customer reviews of Malick’s DVDs at Amazon. It could dip into the growing sea of public data about government activity to tell you what happened at city council (and maybe figure out which parts of the agenda were important, based on news stories, community bloggers, and social media traffic). It could gather up articles from high-trust news and government sources on NASA and algorithmically combine them into just as much info as the searcher wants. It’s a shift in the focus of Google’s judgment; websites shift from competitors to be ranked against each other to data sources to be diced and analyzed to figure out an answer.

I asked each of those questions to ChatGPT this morning. It…did pretty well. It nailed these two:

What’s going to replace the Space Shuttle program?

The Space Shuttle program was officially retired in 2011 and was replaced by the Commercial Crew Program. The Commercial Crew Program is a NASA initiative aimed at outsourcing the design, development, and operation of spacecraft for human space transportation to low Earth orbit. The program was created to provide reliable and safe access to the International Space Station (ISS) for both NASA astronauts and international partners.

The two companies that were selected to participate in the program are Space and Boeing. SpaceX’s spacecraft, the Crew Dragon, has already completed multiple successful missions to the IS, and Boeing’s spacecraft, the CST-100 Starliner, is currently in the testing and certification phase

These new spacecraft are designed to be more cost-effective and efficient compared to the Space Shuttle, and they are expected to provide reliable and safe transportation for astronauts to the ISS for many years to come.

Did Osama bin Laden really use his wife as a human shield?

Yes, reports indicate that Osama bin Laden used his wife as a human shield during the raid on his compound in Abbottabad, Pakistan on May 2, 2011. According to the official account of the raid, bin Laden’s wife, Amal al-Sadah, was shot in the leg as she was used as a shield by bin Laden in an attempt to protect himself. The incident was reported by various news sources, including US government officials, and was later confirmed in the investigation of the raid.

It was useful, if equivocal, on these:

Which is Terrence Malick’s best movie?

Terrence Malick is a highly regarded filmmaker, and his works are often considered some of the best in the industry. It is subjective to name one specific film as his “best”, as it depends on personal taste and interpretation. However, some of his most highly acclaimed works include:

• “The Tree of Life” (2011)
• “Badlands” (1973)
• “Days of Heaven” (1978)

These films are often cited as some of the best in his filmography, and they showcase his unique style, beautiful cinematography, and themes of nature and spirituality.

Was the stimulus package too small to be effective?2

The effectiveness of the stimulus package is a matter of debate and depends on various economic, political, and social factors. Some economists argue that the stimulus package was too small to have a significant impact on the economy, as it did not provide enough support to those who were most affected by the economic downturn.

Others argue that the size of the stimulus package was appropriate given the constraints of the political and economic environment at the time. They argue that the package provided enough support to help stabilize the economy and prevent it from collapsing further, while also providing funding for important infrastructure projects and programs that helped create jobs and stimulate growth.

Ultimately, the effectiveness of the stimulus package is difficult to quantify and measure, and it depends on a variety of factors and perspectives.

ChatGPT didn’t know what the “latest” My Morning Jacket album is, so I asked it specifically about their most recent release as of 2011:

Is the album “Circuital” by My Morning Jacket good?

“Circuital” is the sixth studio album by My Morning Jacket, released in 2011. The album received generally positive reviews from music critics, who praised its mix of eclectic musical styles and the band’s ability to craft compelling and atmospheric songs.

Many fans of the band consider “Circuital” to be one of the strongest albums in their discography, and it is often praised for its cohesive sound and memorable songs. The album was also well-received commercially, reaching the top 30 on the Billboard 200 chart and earning positive reviews from many well-known music publications.

Ultimately, whether or not “Circuital” is a good album is subjective and depends on personal taste. If you are a fan of My Morning Jacket’s music or enjoy a mix of rock, alternative, and psychedelic influences, it is worth checking out.

(“Cohesive sound and memorable songs” is some pretty non-specific praise, it should be noted, and the album actually hit No. 5 on the Billboard 200, not “just” the top 30.)

The last question — “Did Dallas city council approve that zoning change last night?” — is one ChatGPT can’t currently answer, being informationally stuck in 2021. But again, Google and Microsoft won’t have that problem. And for local news organizations, tech companies’ ability to get questions like that right will be critical. Local information, gathered by local reporters, can be exceptionally valuable in the right contexts. Will search engines be able to ingest that knowledge and chat it back out to users without sending even a pageview the publisher’s way?

The debates (and lawsuits) around the ethics of AI-generated images ask similar questions. (Though training an image AI to learn the style of an artist seems distinct from training a chat AI to take specific facts from a publication. “How do you write a city council story?” ≠ “What did city council do last night?”) Does this sort of amalgamation differ, legally or ethically, from the indexing of content that search engines have always done? As publishers around the world have sought hush money, er, I mean compensation from Google, the company’s reply has been simple: But we send you all this traffic! Does the conversation change as that becomes decreasingly true? Or will Google, seeing that risk, decide to focus on the more strategic universe of non-local information, reasoning the legal and PR risks are lower when you’re summarizing Wikipedia than The Hometown Gazette?

Who knows? Not me. 2011 me didn’t know either:

These things aren’t right around the corner — they quickly get to be really complicated AI problems. But they all point to the fact that Google is working hard to reduce the number of times searchers need to leave google.com to get answers to their questions. For all the times that Google has said it’s not in the content business, it’s not hard to imagine a future where its mission to “organize the world’s information” goes way beyond spidering and linking and into algorithmically processing for answers instead of PageRank.

That — much more than news organizations’ previous complaints about Google — could put real pressure on the business models of news websites. It challenges ideas of how to navigate the link economy and what concepts like search engine optimization, fair use, and aggregation mean. And it sure looked like Schmidt pointed the way last night.

A dozen years later, all these questions are becoming more than rhetorical.

Illustration — because we live in an irony-free zone — by Midjourney AI.

  1. Not to mention the role external traffic plays in lead generation for paid products like subscriptions. ↩︎
  2. Note I didn’t specify which stimulus package. ↩︎
Joshua Benton is the senior writer and former director of Nieman Lab. You can reach him via email (joshua_benton@harvard.edu) or Twitter DM (@jbenton).
POSTED     Feb. 7, 2023, 11:51 a.m.
Show tags
 
Join the 60,000 who get the freshest future-of-journalism news in our daily email.
Newsweek is making generative AI a fixture in its newsroom
The legacy publication is leaning on AI for video production, a new breaking news team, and first drafts of some stories.
Rumble Strip creator Erica Heilman on making independent audio and asking people about class
“I only make unimportant things now, but it’s all the unimportant things that really make up our lives.”
PressPad, an attempt to bring some class diversity to posh British journalism, is shutting down
“While there is even more need for this intervention than when we began the project, the initiative needs more resources than the current team can provide.”