Media
6 mins read

Semafor’s Signals: Is the AI-powered curation tool an innovation or a money pact with the devil?

Microsoft and its partner, OpenAI, want to make deals with publishing companies. As Ashley Norris writes, the product the duo launched with Semafor is new ground for all three companies. But is AI-curated content a “magic bullet” for publishers? Not everyone thinks so.

The scene

There has been much talk about how Artificial Intelligence will revolutionise content creation for media companies.

Yet, a year and a half from the momentous launch of ChatGPT in 2022, there are very few examples of publishers harnessing AI to generate content.

This could, of course, be down to the fact that a lot of experimentation with the technology is happening behind the scenes. Several companies, including Sports Illustrated and CNET, have been “rumbled” by other media entities for using AI to create content without acknowledging it. A significant number are probably trialling AI tools, but haven’t yet made their work public.

Another factor could be that media companies’ legal disputes with OpenAI, the company that created ChatGPT, have made innovating with the technology somewhat controversial. 

Yet a growing number of organisations, among them Axel Springer and The Associated Press, have decided to work with OpenAI. They will be paid to licence content for use by the AI technology.

Among them is Semafor, the nascent US news brand launched in 2022. In early February, it announced that it had not only signed an agreement with OpenAI and its partner, Microsoft, but was also using their technology to create content.

It has created Signals, a tool that offers global analysis of the day’s big stories and complements the content created by the Semafor team.

Signals is built around an AI-powered search tool called MISO (multilingual insight search optimiser), which is designed to help reporters curate content quickly by finding a broad range of stories in various languages. In its official release, Semafor describes the tool as “a custom bot built on OpenAI’s platform and using Microsoft’s Bing search engine”.

The Financial Times, which broke the story, reported:  “For a breaking news event, Semafor journalists will use AI tools to quickly search for reporting and commentary from other news sources across the globe in multiple languages. A Signals post might include perspectives from Chinese, Indian or Russian media, for example, with Semafor’s reporters summarising and contextualising the different points of view, while citing its sources.”

No one outside of the two organisations has any visibility on the details of the deal, but OpenAI is rumoured to have invested a fairly significant amount to secure the arrangement.

Semafor co-founder Ben Smith told the Financial Times: “What we’re trying to go after is this really weird space of breaking news on the internet now, in which you have these really splintered, fragmented, rushed efforts to get the first sentence of a story out for search engines  …  and then never really make any effort to provide context.

“We’re trying to go the other way. Here are the confirmed facts. Here are three or four pieces of really sophisticated, meaningful analysis.”

What’s in it for Microsoft and OpenAI?

The deals have been largely about licensing content for use in ChatGPT. Signals is, however, an AI-based tool that is a significant part of how Semafor creates content. 

OpenAI/Microsoft must convince the media that it is a partner rather than a potential rival. A deal with a high-profile, emerging news organisation is a “come on” to other media companies, encouraging them not to follow the New York Times down the legal route.

Crucially, stories that have been created sport a Signals tag (including the Microsoft logo) to differentiate them from other Semafor content. 

What’s in it for Semafor?

One of Semafor’s key innovations is how it structures its articles. The first section is called “The Scoop” or “The News” or “The Scene”, a few paragraphs laying out key facts. The next section is often “Know More”, providing detail. Then follows the writer’s view, with Semafor’s journalists giving their opinion on the story. Some stories also feature a section called “Room for Disagreement”, which outlines views contrary to the writer’s. There is often a link to an established news source at the end for readers who want to find out more.

Signals is slightly different. Its first section tends to be shorter than other Semafor stories, sometimes only one paragraph.

Each story then usually features two or three perspectives mainly from mainstream media. For example, a recent story, US draws widespread criticism after it vetoes UN call for Gaza ceasefire fire, featured three perspectives: “Washington under fire from China, allies”, “US prepping its own draft resolution”, and “US often wields veto power in favor of Israel”.

Each Signals perspective features content (and crucially includes links) curated from diverse media sources like the BBC, Xinhua, The New Arab, CNN and Al Jazeera. 

This writer’s view

So, does it work? Signals certainly has the potential to enable Semafor to publish more stories, many with a global perspective. I guess the rationalisation behind the approach is that the speed and volume of content creation might free up Semafor’s journalists to work on their own stories and deliver more in-depth articles.

That speed does put pressure on the humans editing the content, though. The Signals content is not badly written, but it does have a whiff of automation. Even without the Signals signage, I would wager that regular Semafor readers would be able to spot the difference.

It will likely improve as AI bots become more sophisticated and humans become more skilled with automated content.

From Semafor’s perspective, it feels like a significant win. It gets additional content cheaply, ups its global coverage, and has developed revenue sources. 

But not everyone is impressed

In an article for Defector, Tom Ley makes two key points. Firstly, he asks what the AI is actually doing. Secondly, he wonders why Signals is billed as “innovative” when its function is similar to what news wires have done for decades.

“While reading Signal posts, it’s also hard to discern how Microsoft’s chatbot is actually contributing,” Tom argues. “Smith told the FT that Signals would use AI as a ‘research tool to inform posts’ written by Semafor staffers, and in his post on the site mentioned that editors, after ‘tapping into these AI research tools’, will ‘evaluate and verify sources, compose summaries, and clearly cite and link readers to the original information.’ Again, this is a lot of words to describe what is already a well-established journalistic practice: researching and writing posts. Search engines and press wires have already made that process relatively simple.”

When a media company rolls out a shiny new product, Tom writes, the question is whether the product or the money came first.

“In other words, did Semafor launch Signals—which will now demand time and energy from its staff necessary to pump out a dozen short aggregations every day—because of its inherent journalistic value or because Microsoft came to them and said, Here’s a bag of money. Find some way to make our AI tools look valuable?”

Tom is making some fair points. But he doesn’t acknowledge that Semafor operates in a challenging environment. It is apparently close to turning a profit, with most of its revenue coming from advertising and events. I am sure the chunk of cash from OpenAI/Microsoft has proved very useful. It seems a more palatable way of generating income than plastering the site with ads. Only the reader can decide how valuable the extra content is.

It will be interesting to see if Microsoft/OpenAI offers similar deals to other publishers.

Of course, media companies have found out to their detriment that deals with technology companies may be useful in the short term but could cause problems in the long run. With concepts like Signals, though, the risk factor looks pretty small.

Ultimately

Signals is useful for Semafor for a number of reasons. Likewise, it offers Microsoft/OpenAI an excellent profile. 

Is Signals the killer AI application that will excite publishers across the globe and be widely copied? Probably not. We are still waiting for those innovations.


Join us

  • Join us in person at our Mx3 Barcelona innovation summit on 12-13 March, where we will discuss several themes dominating the current and newly emerging media environment. It’s immersive, off-the-record, and down-to-earth, with places strictly limited in line with Dunbar’s 150 ;). See more at Mx3barcelona.com.
  • On 4-6 June, we are in Cascais, Portugal for the 46th FIPP World Media Congress, which we organise and host on behalf of the international media body FIPP. Learn more at fippcongress.com.
  • There is more on their way soon: check back in at mediamakersmeet.com or sign up for our newsletter to get updated.