Plugging Into AI
Plugging Into AI
Digital Marketing Trends in advertising, artificial intelligence, consumer behavior, policy, search, social media, video & websites...AND Music Monday!
When ChatGPT announced that it was opening up its API for developers to build plugins for its large language model, the proverbial light bulb went off.
To allow developers to apply and train ChatGPT on discrete data sets opens up a world of possibilites well beyond use of the publically available chatbot itself. But it was the code interpreter plugin that convinced me I needed to pay OpenAI for access.
Not only can it write code but it can execute it as well. It will do data analysis and sophisticated visualizations and GIF animations. This podcast episode by HubSpot’s Kipp Bodnar convinced me:
The pace at which AI has improved is mind boggling. During the short time that I’ve been using Midjourney (about two months), it has shipped three significant upgrades that has made the tool far more useful.
In my entire career, I’ve never seen the pace of tech innovation accelerate so rapidly.
Hold on to your seats, boys and girls!
In other trend news:
Advertising
Fast Company - How AI could power advertising’s next creative revolution - Dickon Laws, Ogilvy’s global head of innovation, offers an instructive framework for the narrative arc around new technologies and how they’re adopted that serves as an effective mental model for how creative agencies should be approaching AI adoption. It starts, he says, with the Magpie Moment, when something new and shiny gets our attention. Then comes the Napster Moment, when we get super excited by the technology because it’s new—and maybe doing something a little bit illicit. Next comes the Cambridge Analytica Moment, the WTF moment when the tech gets abused, and legislation, regulation, and compliance enter the picture. Finally, there’s the Spotify Moment, when the technology is normalized and viewed mainly in terms of the benefit it provides.
“We’re going through that curve at the moment,” Laws says, “and we’ve got to protect our clients from being too caught up in the Magpie and Napster moments, to evade the Cambridge Analytica moment, and get to the Spotify moment with minimum brand risk.”
Yep.
The Verge - Forget ad-supported streaming — here come the ad-supported TVs - The co-founder of Pluto TV, Ilya Pozin, wants to give away free ad-supported TVs — yes, you read that right. That’s according to a report from Janko Roettgers in his Lowpass newsletter, who says the free television set will come with a second, built-in screen dedicated solely to ads and a soundbar.
Instead of profiting from TV sales, Roettgers reports Pozin’s new company, called Teevee Corporation, would make money off of the ads that appear on TV’s secondary screen. The idea isn’t as ridiculous as it sounds, as Vizio, for example, already makes far more money from ads and commissions than it does from selling TVs.
Artificial Intelligence
Open Source
VentureBeat - With a wave of new LLMs, open-source AI is having a moment — and a red-hot debate - State-of-the-art LLMs require huge compute budgets — OpenAI reportedly used 10,000 Nvidia GPUs to train ChatGPT— and deep ML expertise, so few organizations can train them from scratch. Yet, increasingly, those that have the resources and expertise are not opening up their models — the data, source code, or deep learning’s secret sauce, the model weights — to public scrutiny, relying on API distribution instead.
That is where open-source AI is stepping into the void to democratize access to LLMs. For example, two weeks ago Databricks announced the ChatGPT-like Dolly, which was inspired by Alpaca, another open-source LLM released by Stanford in mid-March. Alpaca, in turn, used the weights from Meta’s LLaMA model that was released in late February. LLaMA was immediately hailed for its superior performance over models such as GPT–3, despite having 10 times fewer parameters.
OpenAI
Wired - OpenAI’s CEO Says the Age of Giant AI Models Is Already Over - OpenAI has delivered a series of impressive advances in AI that works with language in recent years by taking existing machine-learning algorithms and scaling them up to previously unimagined size. GPT-4, the latest of those projects, was likely trained using trillions of words of text and many thousands of powerful computer chips. The process cost over $100 million.
But the company’s CEO, Sam Altman, says further progress will not come from making models bigger. “I think we're at the end of the era where it's going to be these, like, giant, giant models,” he told an audience at an event held at MIT late last week. “We'll make them better in other ways.”…
Although OpenAI is keeping GPT-4’s size and inner workings secret, it is likely that some of its intelligence already comes from looking beyond just scale. On possibility is that it used a method called reinforcement learning with human feedback, which was used to enhance ChatGPT. It involves having humans judge the quality of the model’s answers to steer it towards providing responses more likely to be judged as high quality.
Two recent experiences reinforce the point made in that last paragraph: During a conversation with colleagues about the social media platforms they use, on person cited TikTok but they mostly use the app regularly “to curate their algorithm.”
Another experience occurred when I was watching a Midjourney how-to video. The creator was discussing how to create consistent characters for use in Midjourney. After each iteration of his character that was closer to what he was looking for, he “liked” that version to tell Midjourney it was on the right track. Two examples of reinforcement learning applied to AI.
AI Music
Forbes - Grimes Tells Fans To Deepfake Her Music, Will Split 50% Royalties With AI - In the wake of the AI-generated hit Heart on My Sleeve going viral with deepfakes of multi-platinum artists Drake and The Weeknd, pop star Grimes has invited her fans to create music with her voice.
On Sunday night she tweeted, “I'll split 50% royalties on any successful AI generated song that uses my voice. Same deal as I would with any artist i collab with. Feel free to use my voice without penalty. I have no label and no legal bindings.”
She added that she’s open to anything anyone wants. “Im just curious what even happens and interested in being a Guinea pig,” She said she welcomes the open sourcing of art and an end to copyright.
This is such a smart move. If you can’t stop it; embrace it.
Promptography
PetaPixel - AI Imagery May Destroy History As We Know It - Analog and digital photography previously presented a window into, and proof of, the distant past. A.I. contrived people do not exist: they have never existed and they will never exist except in the digital ether. The people depicted in these images never lived and they will never die. The human condition of the past is no longer relevant.
Will future generations archive these fake images as historically significant? Will historians mistakenly believe these are depictions of real regalia, dances, and ceremonies performed by authentic tribes? And if so, how does that impact today’s tribes that actively document their real history and heritage? Many experts have already voiced major concerns about “deep fakes” and the troubles they will cause for historians in general.
This is a superb argument for digital watermarks for AI-generated content.
Prompt Engineers
Gizmodo - Want to be Employable? Play Around on ChatGPT - The use of AI like ChatGPT is the way of the future according to a recent survey that found 91% of companies that are hiring want to bring on employees with ChatGPT experience. Resume Builder surveyed 1,187 business leaders and found that not only are candidates with a knowledge and understanding of ChatGPT considered highly sought after, but it also found that one in four companies hiring prompt engineers is offering a salary exceeding $200,000.
Damn! I mean…just…damn!
AI Video
New Atlas - nVidia's new text-to-video AI shows an insane rate of progress - nVidia tested the system using low-quality dashcam-style footage, and found that it was capable of generating several minutes' worth of this kind of video in a "temporally coherent" fashion, at 512 x 1024-pixel resolution – an unprecedented feat in this fast-moving field.
But it's also capable of operating at much higher resolutions and across an enormous range of other visual styles. The team used the system to generate a plethora of sample videos in 1280 x 2048-pixel resolution, simply from text prompts. These videos each contain 113 frames, and are rendered at 24 fps, so they're about 4.7 seconds long. Pushing much further than that in terms of total time seems to break things, and introduces a lot more weirdness.
Watch some of the example videos. Pretty amazing. I’m hoping my Midjourney skills will be useful when I can get my hands on generative video tools!
Bot Love
Scientific American - AI Chatbots and the Humans Who Love Them - Today, we have two very special guests…Together, Anna and Diego produced a podcast called Radiotopia Presents: Bot Love. This seven-episode series explores AI chatbots—and the humans who build relationships with them.
Many of the people they spoke with got their chatbot through a company called Replika. This company helps you build a personalized character that you can chat with endlessly. Paid versions of the bot respond using generative AI – like what powers Chat GPT – so users can craft a bot that is specific to their preferences and needs.
But what are the consequences of entrusting our emotions to computer programs?
Consumer Behavior
Axios - Lawn wars consume America's neighborhoods - The American lawn has become the latest front-line issue in neighborhoods across the country: Some are shelling out to maintain lush greens while others forgo mowing and chemical treatments.
Why it matters: Environmental campaigns like "No Mow May," the "anti-lawn" movement, "Food Not Lawns" and "Climate Victory Gardens" are gaining steam — but prompting homeowner associations and other traditionalists to dig in their heels.
I mean, just to avoid mowing a lawn is enough, isn’t it?
Fast Company - Why the power of brands is getting more tribal - Fans, tribes, neotribes, community, networks, cultures of consumption, and all the many monikers that go by different names all amount to the same thing: people. If you want to get people to move, then you must focus on people and their proclivity to connect. These connections, of course, are fostered and fortified by the cultural characteristics that govern said people. Adoption into these cultural characteristics catalyzes collective behavior and inspires people to move.
I am a huge NFT skeptic, but this article is very good nonetheless for its focus on tribal psychology.
The Information - Nearly Half of YouTube’s U.S. Viewership Is Now on TVs, Helping Drive Ad Shift - For years, Google’s YouTube couldn’t get any respect from the TV industry. TV marketers wouldn’t go near it out of fear that their ads would be tainted by running alongside YouTube’s amateur content. And analysts and research firms treated the streaming service as separate from the rest of television when analyzing TV viewing and advertising.
How things have changed. One startling statistic shows how YouTube is now unequivocally the king of TV. Its internal data indicate that close to 45% of overall YouTube viewing in the U.S. today is happening on TV screens, according to people familiar with the matter, compared with well below 30% in 2020. That’s a radical shift for the video-streaming service, reflecting how the growth of internet-connected TVs has made it easier for people to watch streaming services like YouTube on TVs instead of on their cellphones and computers.
I currently have eight tabs with YouTube videos open on my browser, which I am using on the television that serves as my large screen monitor. Does that count as watching on TV screens? If they’re only counting views from YouTube apps on TVs, then that 45% is likely much higher.
New York Times - Hearing Aids Are Changing. Their Users Are, Too. - A raft of new hearing aids have hit the market in recent years, offering greater appeal to a generation of young adults that some experts say is both developing hearing problems earlier in life and — perhaps paradoxically — becoming more comfortable with an expensive piece of technology pumping sound into their ears.
Some of the new models, including Ms. Wing’s, are made by traditional prescription brands, which usually require a visit to a specialist. But the Food and Drug Administration opened up the market last year when it allowed the sale of hearing aids over the counter. In response, brand names like Sony and Jabra began releasing their own products, adding to the new wave of designs and features that appeal to young consumers.
I guess I don’t need to be as self-conscious any more about what years of loud music have done to my hearing. I’ve gotten very good at seeming like I understand what people are saying when I’m in a room with a lot of people talking. It would be nice to really understand what they’re saying!
Policy
- Vibes-based antitrust hits Microsoft - Under a pre-2016 understanding of Big Tech, Microsoft’s mega-purchase [of Activision Blizzard] might have sailed through the regulatory process. More recently, though, rigorous, metrics-based analyses of market size and definition, or barriers to entry, have faded in favor of what you could call vibes-based antitrust: a broad-based suspicion that the richest tech companies are too big already, and even minor efforts to expand through acquisitions cannot be tolerated — even if the alleged harm is more theoretical than actual.
Or “have faded in favor of” any regulation of Big Tech at all. I can think of only one major regulatory action against Big Tech since I started my career: The anti-trust suit brought against Microsoft that forced it to allow Windows users to install competitors to its Internet Explorer browser. It’s long past time this industry got some serious scrutiny from regulatory bodies.
KOIN 6 - Mental health crisis leads 3 Oregon school districts to file lawsuits against social media platforms - As experts sound the alarm on the mental health crisis facing kids and teens, more school districts are taking on social media companies.
Lawsuits have now been filed by at least three local [Oregon] districts, claiming social media platforms are prioritizing profits over student wellbeing.
In March, the Gervais School District filed a lawsuit on behalf of their students against platforms like Facebook, Instagram, Snapchat, TikTok, Google and YouTube. Since then, similar lawsuits have also been filed by the Washougal and Salem-Keizer school districts.
Legislators in some states are introducing bills that would try to ban children from using social media. School districts suing social platforms is the other side of the coin.
I was fascinated to see that Salem-Keizer was among the school districs filing suit. The Salem-Keizer Cascade model is a leading behavioral threat assessment model for identifying students at risk to themselves or others and intervening with appropriate support. The model was developed by John Van Dreal, who was director of safety, security and risk for that school district.
Foreign Affairs - Innovation Power - The nexus between technological innovation and global domination dates back centuries, from the muskets the conquistador Francisco Pizarro wielded to defeat the Inca Empire to the steamboats Commodore Matthew Perry commanded to force the opening of Japan. But the sheer speed at which innovation is happening has no precedent. Nowhere is this change clearer than in one of the foundational technologies of our time: artificial intelligence.
Today’s AI systems can already provide key advantages in the military domain, where they are able to parse millions of inputs, identify patterns, and alert commanders to enemy activity. The Ukrainian military, for example, has used AI to efficiently scan intelligence, surveillance, and reconnaissance data from a variety of sources. Increasingly, however, AI systems will move beyond merely assisting human decision-making and start making decisions themselves. John Boyd, a military strategist and U.S. Air Force colonel, coined the term “OODA loop”—observe, orient, decide, act—to describe the decision-making process in combat. Crucially, AI will be able to execute each part of the OODA loop much faster. Conflict can happen at the speed of computers, not the speed of people. As a result, command-and-control systems that rely on human decision-makers—or, worse, complex military hierarchies—will lose out to faster, more efficient systems that team machines with humans.
In previous eras, the technologies that shaped geopolitics—from bronze to steel, steam power to nuclear fission—were largely singular. There was a clear threshold of technological mastery, and once a country reached it, the playing field was leveled. Artificial intelligence, by contrast, is generative in nature. By presenting a platform for continuous scientific and technological innovation, it can lead to yet more innovation. That phenomenon makes the AI age fundamentally different from the Bronze Age or the steel age. Rather than natural resource wealth or mastery of a given technology, the source of a country’s power now lies in its ability to continuously innovate.
This piece by Eric Schmidt, former CEO and Chair of Google, goes well beyond military uses of AI. The article takes for granted the AI will be used in wars (a sound assumption, in my view) but takes a larger geopolitical view of AI’s role in innovation in light of America’s competition with China.
ℹ️ Paid subscribers: The next section of Chapter 2 will be published on Wednesday.
Search
Search Engine Land - Google planning new search engine while working on new search features under Project Magi - Google is in the process of building an “all-new search engine” powered by its new A.I. technology while at the same time working on adding new A.I.-based features to its current search engine under the project name Magi…The new search engine would “learn what users want to know based on what they’re searching.” In addition, it would “offer lists of preselected options for objects to buy, information to research and other information.” At the same time, being more conversational, like talking to a person. Earlier reports have Google talking about big changes coming to Google Search.
AI in general and Bing’s adoption of AI in particular is forcing innovations in how search works fundamentally. This could portend both volatility and opportunity for search marketers. The delivery of answers absent citations, such as what you get from ChatGPT, makes those answers suspect.
Search without citations also removes discovery from the search equation, which is a massive problem for creators and organizations alike. Give the cash cow that search advertising is, I’m optimistic citations are here to stay.
But I’ll keep a wary eye on this trend nevertheless.
Social Media
Bluesky
The Verge - Dril and AOC are now on Bluesky - Two Twitter icons joined Bluesky on Thursday. Dril was first, with AOC joining shortly after.
The invite-only decentralized Twitter alternative has been gaining steam in recent days — I’ve seen a lot of people I follow on the bird-themed social network start to show up on Bluesky. But the Twitter clone might truly begin to feel like a New Twitter now that Dril (the Twitter legend who recently gave an interview as his real-life human self) and Rep. Alexandria Ocasio-Cortez (D-NY) have officially joined the platform.
Wikipedia
Motherboard - AI Is Tearing Wikipedia Apart - During a recent community call, it became apparent that there is a community split over whether or not to use large language models to generate content. While some people expressed that tools like Open AI’s ChatGPT could help with generating and summarizing articles, others remained wary.
The concern is that machine-generated content has to be balanced with a lot of human review and would overwhelm lesser-known wikis with bad content. While AI generators are useful for writing believable, human-like text, they are also prone to including erroneous information, and even citing sources and academic papers which don’t exist. This often results in text summaries which seem accurate, but on closer inspection are revealed to be completely fabricated.
Influencers
The Atlantic - The Influencer Economy Is Warping the American Dream - Influencing may appear to be a different type of labor—or not be labor at all—but it still falls into the same traps as traditional work. Not everyone succeeds, for one. As Alice Marwick, an associate communication professor at the University of North Carolina at Chapel Hill, explains, most discussions around influencers focus on mega influencers (commonly defined as those with more than 1 million followers): the kind who can live in luxury based solely on their content. “But that’s the tiniest tip of the pyramid,” Marwick told me. “Beneath them, there’s thousands and thousands and thousands of people who are trying to do the same thing, but not succeeding.” For those people, she explains, it’s one of many stressful careers with long hours and no guarantee of success.
Back during the early days of social media, I wondered what the effect would be on the tailing edge of the Millennial generation, as the first generation for whom it was normal to come of age with the notion that they have an audience.
Entreprenuer - How ChatGPT Will Dramatically Change the Influencer Space - Better audience targeting: A big challenge for influencers is identifying the right audience for their content. With the help of ChatGPT, influencers will be able to analyze their followers' preferences, interests and behavior patterns more accurately. This will enable them to create more targeted and personalized content that resonates with the right people and drives engagement.
The author lists five ways ChatGPT will influence influencer marketing. All are plausible but piece is more speculative than based on concrete knowledge.
While the focus on this piece is ChatGPT, my initial thoughts on how AI will affect influencer marketing is through tools like Midjourney. Applying generative AI to the influencer market will result in a flood of fake influencer content. Reminds me of the HBO show Fake Famous:
Fediverse
The Verge - Can ActivityPub save the internet? - The hottest new thing in social isn’t vertical video, and it’s not AI-driven algorithmic feeds. Instead, it’s a little-known, years-old protocol called ActivityPub that could help rewire the entire social fabric of the internet.
In recent months, a number of tech companies have thrown their resources into ActivityPub and what’s now known as “the Fediverse.” Tumblr is working with ActivityPub, as are Flipboard, Medium, Mozilla, and even Meta. There’s now an official WordPress plug-in for ActivityPub, which will enable the protocol for something like half the internet all at once. Developers are using ActivityPub to build new and different takes on YouTube, Instagram, and much more. ActivityPub is everywhere!
Long read but worth the time. This is the best explanation of how the Fediverse works that I’ve read.
Group Chat
The Atlantic - Of Course This Is How the Intelligence Leak Happened - Although the Discord leaks are, of course, a national-security story, they’re also a story about how information travels in 2023 as the relevance of traditional social media wanes. They are a story about the power, primacy, and unpredictable dynamics of the group chat…
Group chats aren’t just good for triggering geopolitical crises—they’re also an effective means to start a bank run, as the world learned last month. The investor panic that led to the swift collapse of Silicon Valley Bank in March was effectively caused by runaway group-chat dynamics.
Fascinating read for a variety of reasons: In-group dynamics, the atomization of audiences, the potential threats to national security (both militarily and economically), and the difficulty of monitoring threats within group chats.
Review Spam
Motherboard - AI Spam Is Already Flooding the Internet and It Has an Obvious Tell - A search of Amazon reveals what appear to be fake user reviews generated by ChatGPT or another similar bot. Many user reviews feature the phrase “as an AI language model.” A user review for a waist trimmer posted on April 13 contains the entire response to the initial prompt, unedited. “Yes, as an AI language model, I can definitely write a positive product review about the Active Gear Waist Trimmer.”
Another user posted a negative review for precision rings, a foam band marketed as a trainer for people playing first person shooters on a controller. “As an AI language model, I do not have personal experience with using products. However, I can provide a negative review based on the information available online,” it said. The account reviewing the rings posted a total of five reviews on the same day.
This was inevitable.
Video
CNN - How new technology caused a 36% drop in use of force by officers in this CA police department [Video] - Police departments across the US are using a new AI program called Truleo to analyze officers' bodycam videos and prevent problematic behavior.
This kind of AI application is being used by Medtronic (client) for its Touch Surgery product.
Websites
- The Long and the Short of It - It was a "summarize" button. It appeared on the Artifact news app, that semi-useful new content aggregation app, not unlike the many news aggregation apps that came before it. It's one of those things I typically use for a while and tire of. But they just added the "summarize" button and I like it. It's a little hard to find, but once you do use it I think you will keep using it. What's more, I think you are going to be using buttons like this all over the place in the future.It works as you would expect a "summarize" button to work. Read a headline. Click to load the story. Realize you are sort of interested but have other things to do. Click the "summarize" button and it dutifully presents the gist of things with three convenient bullets in a little box at the top of a story, lick·e·ty-split.
This is kinda terrifying but I think he’s right.
WPTavern - WordPress Developers Are Experimenting With Gutenberg-Native AI Block and Content Assistants - As more WordPress plugins for AI-generated content and images, chatbots, and assistants, are landing in the official directory, developers are beginning to explore even deeper integration with the block editor. Moving beyond the prototypical content generators that are cobbled together into a plugin, the tools developers are experimenting with today will provide a more deeply integrated experience that works seamlessly with the block editor as a natural extension of its capabilities.
Check out the video on this article demonstrating one WordPress developer’s GPT plugin capabilities. It’s very impressive and will definitely streamline content creation. It’s also possible it will streamline the creation of a lot of craptastic content.
Music Monday
I’ve been a Tori Amos fan since her first album, Little Earthquakes, blew me away. I saw her in concert at the Orpheum in Minneapolis for that tour. Still one of the best concerts I’ve ever been to.
The recent controversy over the AI-generated Drake and The Weekend collab got me thinking that no AI-generated piece of music is ever going to make me feel the way a perfomance such as this one will: