Will AI Enhance or Destroy the Business of Media?

Matthew Chmiel
Head of Discourse

The largely agreeable July 28 virtual event sparked more introspection than discourse.

At our inaugural virtual event, we invited three guest experts to discuss the intersection of AI and digital media.

Nearly 30 of our ON_Discourse members joined us that day and were provoked by Dan Gardner and Michael Treff. I spoke to Dan before the event to get a sense of his goals for the discussion. “Everyone in our field expects AI to be a massive disruption, I was hoping to get past the hype and dig into the details.”

From my perspective, the discourse was light, almost too pleasant and agreeable to qualify as a deep discourse, but the good news is that the conversation isn’t over yet. Check out some of our key takeaways and let me know if anything you see here makes you want to push this further.

As always, the pull-quotes are not and will never be attributed.

Provocations Used

  • AI will destroy all digital distribution mode
  • Fake news is all news
  • LLMO (Large Language Model Optimization) is the new SEO

Discourse Rating

  • Agreeable — there were no obvious and direct disagreements in the group. Let’s turn up the heat next time! As always, send me a note and we’ll find a way to get your follow-up arguments into the mix.

Recap

This is going to destroy the
internet as we know it


Today’s internet is scraped and organized and ranked by Google. Those results are still, for the short term, listed as links that send users to owned and operated web pages. The introduction of chat-based interfaces, where LLMs process the information from the same web pages and then directly deliver that information, without links, is going to flip the paradigm and ruin the internet.  It will homogenize the sources of information in the eyes of audiences.

The key takeaway here is the irrelevance of owned and operated pages – pretty soon, it seems, the need for websites will fade away – as long as the information is available and presented in the chat query.

— Provocation – AI will destroy all digital distribution models

—— what did you think?

Or…..

It’s putting agency back
in the hands of the user

The evolution of prompting LLMs is a 3-way interaction between an individual and the LLM *and* with the community that is training the LLM. Publishers will have to think in new ways — they can’t just ship some content and stop thinking about it. The essence of this interaction model is that the content gets shaped by the prompts. This is empowering audiences in new ways. 

Forget that nonsense about irrelevant web pages, this new interface and behavior (prompting and querying) are going to reveal new needs among users that will deepen their relationship with the true sources of this information. The fact is, the distribution model for digital media has been fundamentally broken ever since Web 2. This might actually fix it!

— Provocation – AI will destroy all digital distribution models

—— what did you think?

But….

AI needs an attribution protocol

Several members and guests focused on a specific software requirement that will address questions of reliability and trust. The brands, voices, and processes that delivered the information must persist, in some way, in the final delivery of generated-information. In other words, if the LLM is scraping an article that comes from the New York Times, that brand and even that author must be clearly articulated in the answer. There is reason to believe this type of protocol can be administered because publishers have the leverage – LLMs need good source material to be relevant and this is putting the power back in the hands of the publishers. 

— Provocation – Fake news is all news

—— what did you think?

On the other hand…

IT SUCKS

The current state of audience trust in publisher platforms. Will LLMs really make it better? Anecdotal audience research shows that Apple News readers assume that Apple News is a news publication and that the articles they are consuming are coming from Apple News (not the brands and reporters that distribute from that platform). When you remove all links and bury attribution in generative text, you remove all chances of establishing a meaningful relationship between the audience and the brand that created it.

— Provocation – AI will destroy all digital distribution models

—— what did you think?

… but wait…

Young people are savvy about finding trustworthy sources in the new internet

They have a radar for authenticity which is the antidote to the synthetic content they’ve been receiving all their lives. They find authenticity from some brands, but mainly from influencers online. This isn’t always good though, because not all influencers are trust worthy. On top of all that, we don’t know if the AI itself is going to contain any bias that will influence its answers. 

— Provocation – Fake news is all news

—— what did you think?

… and…

AI is just software – just like everything else on the internet

SEO emerged as a discipline after some engineers reverse-engineered the ranking algorithm.  The same kind of reverse-engineering is going to happen with LLMs – even though they are exponentially more complex, it will eventually happen. And when it does happen, there will be more attributions embedded in LLM answers. This will drive up the potential costs that publishers can extract from LLMs because one of the key elements of an LLM – unlike the internet at large – “shitposting will not get you anywhere on ChatGPT.”

— Provocation – LLMO is the new SEO

—— what did you think?

… finally

The Web is undefeated

Many of our members and guests have been working in digital media for decades and they have read (or written) many pieces about the impending demise of the internet – and still it persists.

— Provocation – LLMO is the new SEO

—— what did you think?