Ben Young
Ben Young
December 9, 2022

Edition #386

On why content rights will plague ChatGPT, ad growth slows in 23 and universities as the new media companies.


ChatGPT debuted last week, to much fanfare, the NY Times called it “social media’s newest star”. How it works is, you can give it a prompt, and chat with the AI bot. Finding answers, writing scripts, or even interactive games. Leading to StackOverflow to ban AI-generated answers. Many calling that it may replace Google Search results, with WaPo saying Google faces a serious threat from it.

I am probably the person least impressed by ChatGPT. I don’t think it’ll replace search. But do think it is a novel piece of technology which will lead to a flood of new businesses and experiences. It kind of does what I would expect it to do.

The first thing that makes me pause, where did they get the data to inform the model? Elon Musk shares Twitter used to provide data. Wait, did I realize my tweets were being sold to power machine learning? Reasonably, maybe machine learning to improve the Twitter product. But to power other peoples AI – no. As people grow more aware of this issue around content rights, like it did with search engines, it is going to create new problems.

Let’s give an example, let’s say you asked ChatGPT to give your best Ben Young intro, and it had used my tweets and this newsletter. I didn’t give them permission to do that. Wait till politicians dig into this, they’ll have a field day. And should I own part of any output? Or what would incentivize users to give up their rights?

GPT doesn’t give any attribution. I’m not quite clear how it could, but it could go, inspired by X, Y & Z. Attribution to the source is what has given Google leeway. We will send you traffic, if you want to opt out of Google, you can. But ChatGPT isn’t doing that.

At least, data can be removed as a source and the model can be rebuilt. But at some level, the cats already out of the bag,

You may recall that Facebook had some snafus with machine learning, and ownership of content on Facebook, when it realized, it may end up ‘owning’ a lot of illegal material. And in turn be liable for it. That is all about content rights. A16z recently released some proposals for content rights, around NFTs, you can envisage new rights coming for use of original content in machine learning.

Why don’t I think it’ll replace search? I think Google is already search. Search is a consumer product, not a technical one. No other search engine has made inroads into their share. Voice search was touted as the dream, but Amazon just made major cut backs to Alexa. It turns out the dollars aren’t there. And much of Google Search is already ML powered, with snippets and suggested searches. Google has a reasonable idea of what questions you might also ask, and gives you the choice. In all of the change that’s gone on in the web, Google Search has done remarkably well, to continue to maintain share and innovate on the product.

Finally if ChatGPT got major traction, Google would just scoop up them up. Or any others that start to build momentum.

A counter riff to my point is that it is easy to be cynical of new technology. Especially in the AI space. It’s harder to be optimistic. And I would say that’s a fair point.

Don’t get me wrong, I think this is a wonderful pillar of technological progress, I just don’t think it’s gonna change the world, yet. It is great that it has opened many folks eyes to what possibilities there could be if they build it.

Notable stories this week

Deals/M&A

  • WashingtonPost is ready to sells its unprofitable tech division, Arc XP.
  • Yahoo paid zero cash for its 25% stake in Yahoo.

Campaign of the week

https://www.youtube.com/watch?v=c2UF26YpPFg

View all 2022 best campaigns.

Smartest commentary

  • “If you can take rockets to space and back you can figure out advertising,”… “”It’s not about whether advertisers are back. The right advertisers were never there to begin with because Twitter never focused on building a great advertising product.” –Adam Singolda, Taboola, CEO

Datapoints of note

  • Downloads of Twitter in the U.S. averaged about 125,000 a day during a recent 31-day period — a figure that’s 23% higher than during the previous month and 42% higher than in a similar period last year. The growth was slower outside the U.S., however, with 14% growth from a month earlier and 4% growth from a year earlier, Apptopia data showed.
  • Growth in global ad spend is expected to slow to about 5.9% next year, according to media buying firm GroupM, down from 6.4% estimated in June.
  • DisneyPlus and Netflix to add $1.2b to CTV ad spending next year.
  • Peacock paid subscribers are now 18m.

That’s it for this week.


“Join