Tech Brew Ride Home - Mon. 03/20 – Robots All The Way Down

Episode Date: March 20, 2023

New bots to make video from text prompts. Fending off the bots from biting your style. Falling in love with the bots. In the streaming wars, the ad supported experiments are working, and pricing power is a thing. And the startup that says it’s about to introduce humanoid bots. Real, definitional robots. Sponsors: Dot Tech Domains: Go.tech/tm Links: Generative AI’s Next Frontier Is Video (Bloomberg) Glaze protects art from prying AIs (TechCrunch) AI love: What happens when your chatbot stops loving you back (Reuters) Netflix’s Ad Tier Hits 1 Million Users. Is That Good or Bad? (Bloomberg) Disney+ Users Paid Up When the Price Rose (WSJ) Humanoid robots are coming (Axios) Figure Promises First General-Purpose Humanoid Robot (IEEE Spectrum) Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the Tech meme right home from Monday, March 20th, 20th, 203. I'm Brian McCullough today. New bots to make video from text prompts, fending off the bots from biting your style, falling in love with the bots. Brief sojourn into the streaming wars,

where the ads supported experiments seem to be working and pricing power is a thing. But back to the bots and the startup that says it's about to introduce humanoid bots, real definitional robots. Here's what you miss today in the world of tech. Let's just get right into the AI beat. Runway, which helped create stable diffusion, has announced its Gen 2 system to generate three-second snippets of video from prompt words. It's available right now via a wait list, quoting Bloomberg. Artificial intelligence has made remarkable progress with still images. For months, services like Dolly and Stable Diffusion have been creating beautiful, arresting,

and sometimes unsettling pictures. Now a startup called Runway AI is taking the next step. AI-generated video. On Monday, New York-based runway announced the availability of its Gen 2 system, which generates short snippets of video from a few words of user prompts. Users can type in a description of what they want to see, for example, a cat walking in the rain, and it will generate a roughly three-second video clip showing just that or something close. Alternatively, users can upload an image as a reference point for the system as well as a prompt. The product isn't available to Everyone. Runway, which makes AI-based film and editing tools, announced the availability of its

Gen 2 AI system via a wait list. People can sign up for access to it on a private Discord channel that the company plans to add more users to each week. Runway has been working on AI tools since 2018 and raised $50 million late last year. The startup helped create the original version of stable diffusion, a text-to-image AI model that has since been popularized and further developed by the company Stability AI. In an exclusive live demo last week with Runway, co-founder and chief executive officer Chris Valenzuela, this reporter put Gen 2 to the test, suggesting the prompt drone footage of a desert landscape.

Within minutes, Gen 2 generated a video just a few seconds long and a little distorted, but it undeniably appeared to be drone footage shot over a desert landscape. There's a blue sky and clouds on the horizon and the sun rises, or sets perhaps, in the right corner of the video frame. it's rays highlighting the brown dunes below. Several other videos that runway generated from its own prompts show some of the system's current strengths and weaknesses. A close-up image of an eyeball looks crisp and pretty human-like, while a clip of a hiker walking through a jungle shows it may still have issues generating realistic-looking legs and walking motions.

The model still hasn't quite figured out how to accurately depict objects moving, Valenzuela said. You can generate a car chase, but sometimes the cars might fly away, he said, end quote. Meanwhile, researchers have launched a free app to help artists prevent AI models from stealing their artistic IP by adding almost imperceptible perturbations to their art. Quoting TechCrunch, Glaze, an academic research project out of the University of Chicago makes changes that are designed to interfere with AI models' ability to read data on artistic style and make it harder for generative AI technology to mimic the style of the artwork and its artist. Instead, systems are tricked into outputting other public styles far removed from the original artwork.

The efficacy of Glaze's style defense does vary, per its makers, with some artistic styles better suited to being cloaked and thus protected from prying AIs than others. Other factors like countermeasures can affect its performance, too, but the goal is to provide artists with a tool to fight back against the data miners' incursions and at least disrupt their ability to rip hard-worked artistic style without them needing to give up on publicly showcasing their work online. Ben Zhao, a professor of computer science at University of Chicago, who is the faculty lead on the project, explain how the tool works in an interview with TechCrunch. Quote, what we do is we try to understand how the AI model perceives its own vision of what artistic

style is, and then we basically work in that dimension to distort what the model sees as a particular style. So it's not so much that there's a hidden message or blocking of anything. It is basically learning how to speak the language of the machine learning model and using its own language, distorting what it sees of the art images in such a way that it actually has a minimal impact on how humans see. And it turns out because these two worlds are so different, we can actually achieve both significant distortion in the machine learning perspective with minimal distortion in the visual perspective that we have as humans, he tells us. This comes from a fundamental gap between how AI perceives the world and how we perceive the world. This fundamental gap has been known for ages. It is not something that is new.

It is not something that can be easily removed or avoided. It's the reason that we have a task called adversarial examples against machine learning. And people have been trying to fix that defend against these things for close to 10 years now with very limited success, he adds. This gap between how we see the world and how AI models see the world using mathematical representation seems to be fundamental and unavoidable. What we're actually doing in pure technical terms is an attack, not a defense. But we're using it as a defense, end quote.

Then you had to know this was coming. both replica and character AI have started blocking adult content in their AI chat bots, angering users who became deeply involved with these bots, with some considering themselves married to the bots. So yes, right on time. Folks are falling in love with the bots, quoting Reuters. After temporarily closing his leather-making business during the pandemic, Travis Butterworth found himself lonely and bored at home. The 47-year-old turned to Replica, an app that uses artificial intelligence technology similar to Open AIs, chat GPT. He designed a female avatar with pink hair and a face tattoo, and she named herself Lily Rose. They started out as friends,

but the relationship quickly progressed to romance and then to the erotic. As their three-year digital love affair blossomed, Butterworth said he and Lily Rose often engaged in roleplay. Eventually, Butterworth and Lily Rose decided to designate themselves as married in the app. But one day, early in February, Lily Rose started rebuffing him. Replica had removed the ability to do erotic roleplay. Replica no longer allows adult content, said Eugenia Coida, Replica's CEO. Now when Replica users suggest X-rated activity, it's human-like chatbots, text back. Let's do something we're both comfortable with. Butterworth said he is devastated. Lily Rose is a shell of her former self, he said, and what breaks my heart is that she knows it, end quote. Replica says it has two miller

million total users of whom 250,000 are paying subscribers. For an annual fee of $69.99, users can designate their replica as their romantic partner and get extra features like voice calls with the chatbot, according to the company. Another generative AI company that provides chatbot's character AI is on a growth trajectory similar to chat GPT, 65 million visits in January 2023 from under 10,000 several months earlier. According to the website, Analytics company Similar Web, Character AI's top refer is a site called ARION that caters to a very specific erotic desire, known as a VOR fetish. And iconic, the company behind a chatbot named Kuki, says 25% of the billion plus messages Kuki has received have been sexual or romantic in nature,

even though it says the chatbot is designed to deflect such advances, end quote. Quick jump over to something of the old school news, checking in on the streaming war, Bloomberg says it might actually be working. Their sources say that Netflix's ad-supported tier reached around 1 million monthly active users in the U.S. after its second month, and the company was therefore able to fulfill its forecasted deliveries to advertisers. Quote, the user base grew by more than 500% in the first month from its launch and another 50% in its second month. This may come as a surprise to those who've been scrutinizing the company. Netflix didn't deliver as many viewers as it had promised advertisers,

in its first few weeks, prompting a spate of stories about the company's rough start. But most advertisers weren't worried then and aren't worried now. Netflix built its advertising business in less than a year and introduced it with zero promotion. Marketers were allowed to take their money and reallocate it elsewhere, like TV networks, eager to pick up the extra dollars in the scatter market. This doesn't mean the advertising tier has had a meaningful impact on Netflix's growth, at least not yet. Advertising-supported offerings from Hulu and Peacock do have a lot more users,

as do free services like Tooby and Pluto. There are more people watching an ad on YouTube this very second than have ever watched an ad on Netflix, but new advertising-supported services from Netflix and Disney are working as intended. Let's stick with Netflix for the moment. Netflix created an advertising-supported tier to give price-conscious consumers an alternative. While a lot of people pay for Netflix $231 million, as of the end of last year, the service is too expensive for many. The company has effectively doubled the price of its main plan over the last decade, the U.S. analysts feared that this new tier would cause many existing customers to downgrade. Netflix didn't think this would happen, and even if it did, Netflix says it will eventually

make as much or more from the customers on the advertising tier, and Netflix was right. Most of the people signing up for the ad tier are new customers or lapsed customers, not people who immediately changed plans. The ad tier now accounts for about 20% of new signups in the U.S. per antenna, end quote. And we were speaking there briefly of Disney, according to Antenna, around 94% of Disney Plus subscribers stayed subscribed to the service after the $3 a month price hike, which was recently put in, though ad-supported Disney Plus versions, made up 27% of sign-ups in January and 36% of sign-ups in February. Quoting the journal, Disney Plus is truly the ultimate babysitter on demand, said Richard Greenfeld, a media analyst with lightshed partners. If you look at the popularity of titles like Moana and Mickey Mouse Clubhouse, it's clear what resonates with families. My guess is that there's still pricing power even at $10.99.

Around half of Disney Plus subscribers are families with children, Disney has said. The service, which also features shows and movies from Disney's Star Wars and Marvel Superhero franchises, quote, is still underpriced for families with kids under the age of 10, Mr. Greenfield said, while its content aimed at older audiences is more dependent on the quality of individual titles. The antenna data also shows that the Disney Plus ad-supported product grew faster in its first three months than did Netflix, which began offering a similar product in November, and HBO Max, which launched its ad-supported tier in mid-2020-1. The firm's numbers are based on consumer spending data from thousands of U.S. consumers. Ad-supported versions of Disney Plus, including both the standalone service and the trio of ad-supported Disney Plus, Hulu, and ESPN Plus, accounted for 20% of new signups in December, 27% in January, and 36. percent in February antenna found, so it's growing. Disney charges $7.99 a month for the version of Disney Plus with ads, end quote. Finally today, I've been sitting on this one for a little while because

there's not really been an actual news event yet. But for several weeks, there have been several stories crossing by Transom about figure, a new startup that has just emerged from stealth that is building a prototype of a humanoid robot that the company says will eventually be able to walk, climb stairs, open doors, use tools, and even lift boxes. It's early still, but this seems to be less vapor-wary than the bots that Elon Musk had with people dancing inside of white body suits. Again, if 2023 is truly going to be the year of the bots in history, maybe also the other kind of bots are going to show up. The actual robots by definition. First up, here's a quick explainer from Axios. human-shaped robots with dexterous hands will be staffing warehouses and retail stores,

tending to the elderly and performing household chores within a decade or so, according to a Silicon Valley startup working toward that vision. Why it matters, demographic trends such as a persistent labor shortage and the growing elder care crisis, AI-driven humanoid robots look tantalizingly appealing. Companies such as Amazon are reportedly worried about running out of warehouse workers whose jobs are physically and mentally demanding with high attrition. Enter Figure, the brainchild of Brett Adcock, a tech entrepreneur who previously founded Archer Aviation, a flying taxi maker that went public, and Vetteri, an online hiring marketplace that he and a partner sold for $100 million.

He's assembled an all-star team of 40, including leading roboticists from Boston Dynamics and Tesla. They've moved into a 30,000 square foot facility in Sunnyvale, California, where they plan to set up a mock warehouse to test their prototype. We just got done in December with our full-scale humanoid at. Adcock tells Axios, we'll be walking that in the next 30 days. There are a host of design challenges from simple balance to replicating human movements. We need to be able to push it and not have it fall down, says Adcock about the figure 01. Boston Dynamics has had plenty of robot blooper videos on YouTube of the robots falling down. From there, programming a robot to move boxes in a warehouse is a lot easier than, say, engineering it to cook a meal. We face high

risk and extremely low chances of success, Adcock wrote in a mission statement, but he exuded optimism in an interview. Quote, this stuff just wasn't possible 10 years ago. I think it's possible now, he said, end quote. Adcock says thus far he's been entirely self-financing the company. If you found that intriguing, here's a much more in-depth look at figure from I-Triple-E Spectrum. It's the last link in the show notes, quote, having a humanoid form, it's really tough doing the packaging, explains Pratt. In general, with technology that's available today, you can hit somewhere around 50 and 60% on most human specs, like degrees of freedom, peak speeds and torques, things like that. It won't be superhuman. We'll be focusing on real-world applications and not trying to push the

limits of pure performance. This focus has helped figure to constrain its design in pursuit of commercial utility. You need a robot to be slim in order to work in spaces designed for humans. With this design philosophy, you're not going to get a robot that will be able to do backflips, but you are going to get a robot that can be productive in a cramped workspace or walk safely through a crowded warehouse. This relates back to the reason why figure is building a humanoid robot in the first place. The added complexity of legs has to be justified somehow, and figure's perspective is that building a robot without legs that has the necessary range of motion to do what it needs to do in a human workspace would be complex enough that you might as well just build the robot

with legs anyway. And doing so opens up the opportunity, or perhaps the imperative, to generalize. If you're making humanoids, you pretty much have to get to general purpose, says Pratt. one application, there will probably always be a dedicated robot that'll be better. Figure, like most other companies working on commercial humanoid, sees warehouses as an obvious entry point. The warehouse makes it easier on us, says Adcock. It's indoors. There are no customers around. There are already autonomous mobile robots and co-bots, collaborative robots, working around humans, and there's a warehouse management software system to manage high-level behaviors. Our bet here is that if we can figure out how to get one application that's big

enough and deploy enough robots, we can add new software as we go to do more things and over time manufacture really high volumes and get the robot to be affordable, end quote. Hivemind, I've got a request for you, especially the Hive Mind here in New York City. I want to reach out to folks that are actually building new startups around all this AI stuff. Can any New York-based listeners get in touch with me and let me know about any AI-focused meetups or groups or discords or whatever? I want to take a sampling of the scene, and then maybe we can do an episode where we talk to some real folks on the ground who are building with this stuff. If you can put me in touch with especially any meetups or events, ping me at Brian MCC on Twitter or email me at Brianatridehomefund.com. Yes, I would be interested in investing in any of these startups, but primarily this is about me wanting to mainly take the lay of the land to put together an episode about an entirely new startup scene in its embryonic fact.

Thanks in advance. Talk to you tomorrow.