Signal 25.05: A double bang for NBA TV rights, is A.I a kleptomanic? And Meta's Llama's complicated approach to Australian news

An unusual dispatch hour for Signal this week. I am sitting in Suvarnabhumi Airport on a 4 hour transfer waiting for my flight to the WAN-IFRA World News Media Congress in Copenhagen. In my view this meeting of the global leaders in the news and journalism industries couldn't be happening at a more important time, as AI's emergence, a tough advertising market, and sustained affordability challenges for consumers create headwinds on consumer growth. At the same time, these challenges create great opportunity

ONE: The NBA heading for annual broadcast rights payments of $7B USD, up from the current $2.6B. ESPN, Amazon and Comcast set to win

What’s new: The NBA, arguably one of the premiere sporting leagues in the world, is set to announce its next series of TV broadcast rights will score them around $7b a year in revenue. A whopping increase on the $2.6b currently being paid by ESPN and Warner Bros Discovery. This will likely see WBD replaced by Comcast (who can offer games on FTA network NBC), as well as the addition of Amazon

Why it matters: The NBA value is likely going to be recouped by these platforms via subscription payments. The NBA generated approx. $1.5B in ad revenue this season, which means subscriber volume needs to cover the remaining $5.5B in rights costs. It also shows the appetite of media companies to keep driving sports prices higher and higher even in the face of massive disruption and a hugely changing ad market. For players this could mean a doubling of salaries, which could mean a max level NBA player is paid $100m plus a season in the not too distant future.

Marketer implication?: Sports is finite and unique, and as a result is demonstrating incredible pricing power. The advertiser cost to be involved in sport is just going to keep going up.

TWO: A.I might be coming to steal your job, but it also has IP to steal first.

What’s new: What's the deal with Open AI and taking things without permission? In the last week it's not only been alleged to have copied Scarlet Johannsen's voice for it's new GPT 4.0, but its COO also failed to answer the question on whether Open AI was using YouTube videos as training data. Sam Altman also at the World Economic Forum declined to say that he would only use licenced and truly public domain data to train his value destruction machine.

Why it matters: Generally taking things you didn't create, without the permission of the creator, is not something particularly endorsed. Let alone creating an entire business and technology based around it. Slack came under fire last week for having a convoluted opt-out to stop it using chat text within Slack to train AI models. Linkedin is using its collaborative articles as AI training data too. A big part of the AI race is to find ways to get training data for free (or next to free) or tap into existing sources of content creation (such as Reddit or Twitter) to have a constant hose of new data. The future of content creation could weirdly look like humans in low paying jobs churning out content to train AI models for their robot overlords

Marketer implication?: My view is the hunger for IP from these models doesn't end with just content online such as news and videos. Soon these models will be doing the same internet hoovering on businesses IP, design, pricing, locations, assets, go-to market. Marketers haven't shown much consideration for AI lifting the content of its media partners, you'd hope they'll show a bit more care if the same robots start going after their jobs and employers.

THREE: META's Llama begins taking news from Google search returns, before turning it off and offering users "no up to date news"

What’s new: I ran a prompt into the Llama AI jammed into every Meta product this week and got the below return.

I posted it here on Linkedin and it got some traction on here and also via the Unmade newsletter. My main point was Meta having Llama scrape news when the company has been so strident in not wanting to compensate news organisations for content. It was also worth noting one of the news stories was incorrect. Using Google search on an enterprise level for content scraping is also generally against Google TOS.

On Friday I tried the same prompt again and got the below.

Now Meta recommends checking a search engine (which is what the same AI did just a day before)

Why it matters: "Move fast and break things" is still alive and well at Menlo Park.

Marketer implication?: I wonder what happens when these AI platforms start to get things wrong and say something defamatory against a brand or individual. Or recommends someone do something that may harm them. AI is being presented as something high value, real and trustworthy, but also seems to have the power to push out information that is not only stolen, but incorrect.