Hi there,
One burning question in generative AI is where market power will accrue – to the companies making foundational AI models, or the applications (and publishers) that use those models.
In the first scenario, where just a small number of AI companies have proprietary foundational models, those companies would have a competitive advantage that helps them dominate distribution; collect fees from developers that use their APIs; and negotiate firmly with content providers. But if foundational models are broadly available, then publishers with unique content & data would be better positioned to build their own proprietary AI models – or demand high licensing fees.
Last week Meta released its new AI model, Llama 2, under an open source license, where companies with fewer than 700 million users can use it for commercial purposes, for free. Early reports find that it’s quite good – perhaps comparable to OpenAI’s GPT-3.5 – and the open source community will make it better quickly.
For publishers, the release provides a glimpse of a hopeful future, one where they’re able to build valuable products on top of open source AI models. If, for example, conversational AI turns out to be a great interface for recipes, a strong open source ecosystem makes it more likely that a publisher could leverage its archives to build the dominant product in that space.
Meta’s core business benefits from this approach, making it likely that the company will maintain an open source policy in future AI releases as well. As an aggregator, Meta benefits when there is a lot of user-generated content, which open source models will make cheaper and easier to create (a point explained in greater detail by Ben Thompson in Stratechery). Not to mention that Meta will benefit from influencing the technical standards of a prominent AI model.
It’s hard to predict what AI-powered products will resonate, but it’s now cheaper to experiment and learn – and many publishers are well-positioned to do so.
And here’s the latest in digital media:
Publishers:
Large publishers, including the NYT, IAC, and News Corp, are close to forming a coalition that could lead to suing AI companies.
Court documents offer new details about Vice’s financials: It generated $258m in gross revenue in 2022, and it has raised over $1.3b in debt & equity financing since 2017.
Altice is reportedly exploring a sale of Cheddar. (Read our rundown of Cheddar’s situation here.)
NPR appointed Edith Chapin as editor-in-chief.
Endeavor Business Media is partnering with an ad sales agency in Italy.
A Chinese marketing agency used wire services to place pro-China stories in U.S. outlets.
Semafor issued a correction explaining that there was not a walkout planned at the NYT.
Platforms:
TikTok is enabling text-based posts.
Twitter released its new “X” logo.
Sam Altman’s cryptocurrency initiative Worldcoin launched.
Interesting Reads:
Indy newsletter writer Edwin Dorsey is monetizing accountability journalism by selling high-priced subscriptions to short sellers.
Thanks for being a part of Business Side’s public beta. Have a great day!
Was this forwarded to you? Subscribe here.
Not sure where market power will accrue, but I believe revenue will be lead by those with proprietary data and content. It seems like every week, or day, there is a new LLM open AI model from big tech. Although billion$ are being spent there is not much proprietary in these models; easy to copy. Amazon may be the best and most experienced AI company that is benefiting from AI, and also doesn’t allow its data to be scraped. Other examples might include cos such as LinkedIn, Spotify, Yelp, Airbnb, or GitHub. Many more like these. Anyway it will be fun to watch!