Exploration #114

Hate Longscrolling? Jump Around!

• If You Click Only One…

• Things to Think About…

• Things to Cringe About…

• Things to Know About…

• And finally…

Hi all. This week we’ve got a BBC-led survey on audience reactions to generative AI, Anthropic’s newest, most powerful version of Claude, and a look at the bizarre world of farming simulator eSports. But first…

3rd Thursday Recap 

It was great to see so many of you at our 3rd Thursday webinar last week. If you missed any part of Integrating AI with DEI in Public Media: Challenges, Solutions, and Opportunities you can catch it here. And we know that a number of you expressed an interest in the slide deck. You can download that directly here.

We're taking July off, have already started booking fall guests and will be back in August. In the meantime, you can see any of our winter or spring webinars at the Public Media Innovators events page.

This Task Goes to 11 

I picked up a new tool yesterday that I’m already finding pretty handy. 11 Labs, the platform we’ve experimented with for voice synthesis, has released a reader app for iOS that reads text you feed it with a realistic human voice. And frankly, some of the voices would be right at home on our air. The best part for me is the integration with the iOS share function. I can send long-read web pages straight to the reader and have them read to me. This huge because I’m on the move a lot. And while I can listen to podcasts or YouTube videos, long-reads (key to this newsletter) have been elusive. Some apps like the NYT and Atlantic (which also uses 11 Labs) provide their own readers with better than average human voices, but this gives me a one stop shop for Medium pages, Substacks, and all the rest. Thus far, the service is free. We’ll see how long they let me chase that dragon before hitting up my ApplePay.

Back with Another One of Those Bot Blockin’ Feats 

All the major gAI chatbot tools have achieved their levels of prominence by crawling the web for content to be used as training data. And that has become the fundamental IP issue of our day, meaning that every media company has to ultimately decide whether they will block the bots or not. I’ve mentioned here and in various public settings that our stance at Nebraska Public Media is to let AI bots scrape our site for training content. I’m not second-guessing any other company’s choice, but our rationale is that if you deny the systems quality content for training you necessarily get worse gAI tools.

We’ve had to amend stance though, and I wanted to be transparent about that here. It started the evening of April 24 when our lead front end developer, Steve Exon, posted to our Labs Discord server:

“Website is super slow & CPU levels are elevated. Going to try to restart [the] Apache [server]. [30 minutes later] Just looked at the logs. Claudebot is relentless. Wow!!”

Claude, the super friendly, constitutional AI, was being a very bad robot.

For years the internet has run on what is essentially a handshake agreement that web crawlers would receive instruction from individual sites on how to handle crawling and indexing sites. (David Pierce’s “The text file that runs the internet” offers a good explanation.) But like so much else with the AI era, things apparently move too fast to be bothered with highly functional convention. In this instance ClaudeBot was ignoring our crawl-delay request and (in the words of our Director of Labs) “hammering our site.” Taken to an extreme, this type of activity is also how DDOS (distributed denial of service) attacks take down websites every day. So, as a result we were forced to block our first bot.

Since then, we haven’t identified any additional site performance issues as a result of other AI bots, though others have reported OpenAI and Perplexity bots will also go a-Viking. But we keep an eye on it and it’s something you should be aware of as well.

Next Exploration July 10 

Barring any major emerging tech news, I’ll be back with our next exploration on July 10.

Okay, on to the links…

If You Click Only One…

Embedding the Audience: Putting Audiences at the Heart of Generative AI (Peter Archer - BBC) - This is a long read (33 pages), but an important one, so at least give the forward and executive summary a chance. And while the survey data is mostly of British and Australian media consumers, some Americans were surveyed online by Ipsos. So, there are extrapolations that, absent local survey data, we can make about the sophistication of our audiences relative to generative AI. If nothing else, this is at least worth sharing this with your boards, advisory boards and/or commissions to see what degree the results and conclusions resonate with them.

Things to Think About…

Netflix Games’ Second Season: Challenges, Changes, and the Path Forward (Michail Katkoff - Deconstructor of Fun) - These past couple of weeks I've been talking more about general audience video games as a potential future for public media. But my excitement and optimism about that path don't mean that I think it'll be an easy one to walk. But we can learn from others, and for those of us thinking about strategy this is a good read.

New Cultural Codes (Zoe Scaman - Musings of a Wandering Mind) - This is more about innovation than about tech, but if you are in marketing, there is good food for thought here. Definitely worth checking out.
—Jump straight to Scaman’s deck on Canva.

How generative AI could reinvent what it means to play (Niall Firth - MIT Technology Review) - One of the current uses of generative AI is for improved customer service. In the context of video games though, that approach could result in more immersive experiences, which can often make for more successful games.

Algorithmaxxing (Anna Gorham - Kernel Magazine) - This one is for anyone who is fascinated by language and how the precision of technology forces people to be even more fluid in our use of language.

A Generation of AI Guinea Pigs ($) (Caroline Mimbs Nyce - The Atlantic) - This one shifted my thinking a bit on the use of AI in education. I tend to be an techno-optimist (though not a techno-solutionist) where AI is concerned. But framing AI alongside the unknown-unknowns of social media's impact on kids, and I won't lie that I felt a little doubt creeping in.

Forget Deepfakes: Social Listening Might be the Most Consequential Use of Generative AI in Politics (Dean Jackson & Zelly Martin - Tech Policy Press) - I like a good alternate point-of-view, and this one looks past the well-wrung fear of deepfakes to AI's other strengths (or threats).

Things to Cringe About…

AI is Already Wreaking Havoc on Global Power Systems (Josh Saul, Leonardo Nicoletti, Saritha Rai, Dina Bass, Ian King, & Jennifer Duggan - Bloomberg) - This isn't the first time we've touched on the externalities of emerging tech like crypto and (in this case) AI. This cleverly assembled web piece shows just how large the numbers have grown (8% of US energy consumption, up from 3% two years ago). I don't have a suggested solution for this. Some may suggest boycotting these tools, but I think that's cutting off your nose to spite your face. But I also don't think we should turn a blind eye to it either.

Google News Promotes “News” From Dozens of AI-Generated Content Farms (Jack Brewster, Isaiah Glick, & Kate Reynolds - NewsGuard) - As I said relative to a similar story last week (and many times before), garbage in = garbage out.
—In the same vein is this piece from Emanuel Maiberg in 404 Media: AI Images in Google Search Results Have Opened a Portal to Hell

Things to Know About…

Anthropic has a fast new AI model — and a clever new way to interact with chatbots (David Pierce - The Verge) - The “big” news in AI this past week was the announcement of a new model for the Claude chatbot. I use quotes though because new model announcements feel less and less like news to me these days. The arms race between the frontier models is such that it almost creates an ennui that borderlines on entitlement. The inurement of amazing times. In this instance though, the improvements seems mostly performance based. Claude's feature set still pales in comparison to ChatGPT Plus. That said, it's worth remembering that Anthropic claims to have a more strategic B2B focus than B2C.
—Michael Nuñez over at Venture beat also weighed in with Why Anthropic’s Artifacts may be this year’s most important AI feature: Unveiling the interface battle. Here’s a key line: While OpenAI wows us with ChatGPT’s newfound voice capabilities and Google touts the expanded knowledge of its Gemini models, Anthropic is taking aim at a far more fundamental question: How do we turn AI from a fancy chatbot into a true collaborative partner?
—See Anthropic's announcement directly: Claude 3.5 Sonnet
—Plus, just announced prior to publishing this exploration, Claude Projects
—Fans of Claude may also want to delve into Anthropic's explanation of how they fine tuned Claude’s Character

Record companies sue AI music generators Suno, Udio ($) (Gerrit De Vynck - Washington Post) - In a surprise to no one, this was announced Monday. Feelings of Napster nostalgia would be a natural reaction, but the argument being put forth here is that the innovations of Suno and Udio are actually harshing record labels ability to innovate. To me the bigger issue has to do with replicating singer voices. Udio gave me a track once that had a vocal which was a dead ringer for Annie Lennox. And that places this more in the arena of the OpenAI's did they/didn't they use of Scarlett Johannsson's voice.
—Then, shortly after writing the text above, I ran across this from Ethan Hillman in Rolling Stone: The World’s Largest Music Company Is Helping Musicians Make Their Own AI Voice Clones
—Also potentially of interest is this manifesto of sorts from UMG and Roland: Principles for Music Creation with AI

Publishers Are Increasingly Irked at Perplexity Bots Circumventing Blocks (Trishla Ostwal & Mark Stenberg - AdWeek) - Perplexity has had a complicated PR year, thus far. Initially touted as a David to Google-iath on the battlefield of web search, they've slowly shown themselves to be, at best, an anti-hero.
—Wired puts it a bit more bluntly in the headline to Dhruv Mehrotra and Tim Marksman's article, Perplexity Is a Bullshit Machine

Google is bringing Gemini access to teens using their school accounts (Aisha Malik - TechCrunch) - Key line: "Gemini will be off by default for teens until admins choose to turn it on."

Amazon mulls $5 to $10 monthly price tag for unprofitable Alexa service, AI revamp (Greg Bensinger - Reuters) - I mean, who wouldn't want to pay for a service that was previously free and which now barely registers as an also-ran on the emerging media landscape.

What We Learned from a Year of Building with LLMs (Part III) (Eugene Yan, Bryan Bischof, Charles Frye, Hamel Husain, Jason Liu and Shreya Shankar - O'Reilly.com) - This series is more geared toward the product folks in public media. I published part 1 a couple weeks back, and since then the other two parts have dropped. You’ll find the first two below.
—What We Learned from a Year of Building with LLMs (Part II)
—What We Learned from a Year of Building with LLMs (Part I)

BBC partners with Condense for immersive live events (Jeremy Walker - BBC) - Here's one to share with the engagement and outreach folks in your world. Some potential ‘steal this idea’ material.

Roblox’s Road to 4D Generative AI (Morgan McGuire - Roblox) - Gentle reminder that the major gaming platforms like Roblox and Fortnite are incubators of innovation that will help determine the future of the internet. Here the company's chief scientist walks us through how the gaming platform is continuing to incorporate generative AI and machine learning. And to spare you the suspense, the 4th "D" is interactivity.

Mixed Reality and Mobile AR Dominate AWE (Charlie Fink - AR Insider) - Last week the American version of the Augmented World Expo (AWE) came and went. This one has been on my list to attend for a few years, but South by Southwest always wins out. When VR/AR, spatial media and/or the metaverse if truly back, I'll probably hit it up. Meanwhile, Charlie Fink sums up the three-day expo for us.

And finally…

Inside the Peculiar World of Farming Simulator eSports (People Make Games via YouTube) - And finally, you kind of just have to see it to believe it. But once you see it, it does kind of make sense.
—And if that wasn’t peculiar enough for you (or video isn’t an option right now), try Rich Pelly's piece from The Guardian: ‘It’s impossible to play for more than 30 minutes without feeling I’m about to die’: lawn-mowing games uncut

Have a creative, productive couple of weeks, and safe 4th of July!