The AI Daily Brief: Artificial Intelligence News and Analysis - What AI Backlash at YouTube Can Teach Other Companies

Starting point is 00:00:00 Today on the AI Daily Brief, what other companies can learn from the backlash around YouTube's use of AI to upscale people's videos? Before that, in the headlines, an interesting partnership between meta and mid-journey. The AI Daily Brief is a daily podcasted video about the most important news and discussions in AI. Hello, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, Blitzy, super-intelligent, and high touch. And as always, to get an ad-free version of the show, head on over to Patreon.com slash AI Daily. Welcome back to the AI Daily Brief Headlines edition, all the daily AI news you need in around five minutes. We kick off today with an interesting partnership announcement. Meta has struck a

Starting point is 00:00:45 new deal with Mid Journey. On Friday, Meta's new chief AI officer Alexander Wang posted on X, today we're proud to announce a partnership with Mid Journey to license their aesthetic technology to our future models and products, bringing beauty to billions. This technical collaboration between our research teams is part of our effort to team up with the best companies in the industry whose work and expertise complements our own. We are incredibly impressed by Mid-Jurney. They've accomplished true feats of technical and aesthetic excellence, and we're thrilled to be working more closely with them.

Starting point is 00:01:13 To ensure meta is able to deliver the best possible products for people, it will require taking an all-of-the-above approach. This means world-class talent, ambitious compute roadmap, and working with the best players across the industry. Mid-Jurdy founder David Holes wrote, Bringing sublime tools of creation and beauty to billions of people is squarely within our mission, excited to partner with the Titans of Industry to make this happen.

Starting point is 00:01:32 However, he also confirmed, quote, We remain an independent community-backed research lab with no investors working on a staggering array of ambitious projects focused on bringing about humane futures where we are all Mid Journey. Join us. Now, this one is super interesting. Mid Journey has always had a pretty fascinating place in the AI Image Generation Pantheon and that it is just really aesthetically tasteful.

Starting point is 00:01:54 I've said before that my first conversion experiences into the world of AI were largely with Mid Journey. I would imagine dreamy scenes from history like Ernest Heminghaming way in Paris in the 1920s. As their models have evolved, they've continued to get better and more tasteful and more interesting and more visually arresting, although they haven't evolved in some of the ways that other tools have, which have left them in a different use case column for me and many others. When you see an image like a cover image or a thumbnail on one of my videos, it's almost invariably created by ideogram. Sometimes it's from OpenAI, which you can usually tell from

Starting point is 00:02:25 the weird yellow tint, but I would say that the vast majority of time it is, in fact, ideogram. Ideogram has the best fidelity to text, most closely follows my instructions, and ends up being really good for the day-end, day-out business sort of use cases. When I need to do something visually arresting, though, something where the main determinant of success is beauty or just visual interestingness, then I'm turning to mid-journey. This has led people to wonder what meta is going to do with the company. The verge suggested that, in their words, meta is going to stuff mid-jorney AI images into your feed, which some think is terrible, but which others rightly point out, that since Facebook

Starting point is 00:02:58 is already inundated with AI images, leveling up the aesthetic quality with a Mid-Journey partnership might not hurt. The other question is whether the deal arises out of a failed aquahire attempt. It was already reported that Meta had attempted to buy out Gen Video Startup Runway, so we know they've had their sites on various AI multimedia companies. The partnership, however, suggests that Meta isn't just looking for talent, but also trying to get their hand on the best tech to build visual AI products. Responses really run the gamut. Some saw this as a knock on meta skill. AI commenter Spore writes, Mid Journey is phenomenal and this is a great partnership for both parties, but man, another point in favor

Starting point is 00:03:31 of being bearish on meta. Gaming startup founder Assad Darr suggested instead, though, that the move was more about urgency, commenting meta licensing mid Journey tech is smart. Creative AI is still undervalued, but it's going to define how we share, market, and communicate online. Tony Wong thought it was an example of the natural evolution of platforms. He wrote, meta plus Mid Journey equals platforms now outsource taste itself. We've seen this playbook before. just stream music, they algorithmic musical discovery. Netflix didn't just host shows they became the curator of what stories get told. TikTok didn't just show videos they rewired attention spans. Now meta embeds AI taste into the infrastructure. Like all things with AI, this is basically

Starting point is 00:04:12 a Roarshack test on how you feel about AI and the company's involved. So whether this is making AI slop look prettier, or an example of how Zuckerberg comes back swinging when his back is against the wall is mostly going to be in the eye of the beholder. One other story about meta, however, the company is upsizing their infrastructure through a $10 billion cloud deal with Google. Sources speaking with the information say that the two tech giants have agreed to a six-year deal that will see META spend more than $10 billion. Those sources said that the deal largely relates to meta securing their AI inference needs. Now, this month has seen META shuffle around their data center plans as they figure out how to

Starting point is 00:04:45 fund one of the most ambitious buildouts in the industry. The company has forecast around $70 billion in KAPS this year, and another big increase in 2006. Earlier this week, it was reported that META had closed a $29 billion private credit deal, marking an interesting shift from big tech firms self-financing their infrastructure buildout. And now we have META positioning themselves as a large renter of inference as well as a data center operator in their own right. Meta also has agreements in place with Amazon and Microsoft, but the Google deal is a significant jump in capacity. Now, ultimately, this makes sense. When Zuckerberg was out recruiting the superintelligence team, he made a big point that the best

Starting point is 00:05:18 talent wants access to unconstrained compute. Ever since then, Meta's move seems to designed to set up near-infinite resources for their big AI push. News from another major player in the Foundation Model game, XAI, the company has officially open-source GROC 2.5. The model, which is now two generations behind, has become available on Hugging Face. Elon Musk said that the previous generation model GROC 3 will be available in around six months.

Starting point is 00:05:43 The release gives us the first opportunity to examine GROC's architecture. The model has 268 billion parameters arranged in a mixture of experts' design with eight total experts and two active at a time. It has a maximum of 131,000 tokens of context. Interestingly, AI engineer Tim Kellogg criticized the release, noting that the custom license has some anti-competitive terms. He queried, why did it take so long to open this up?

Starting point is 00:06:05 There's nothing interesting here. Now, to the extent that there is something interesting for some, it's really just XAI setting the tone by deprecating their old models into the open source world. AI researcher Sebastian Rochka commented, it's quite cool that the open source community gets the real full-size model that was used in production. as opposed to the light or spin-off version.

Starting point is 00:06:23 Separately, Elon Musk has announced plans to launch an AI software company called Macro Hard. Aside from enjoying the pun relative to Microsoft, Musk commented, In principle, given that software companies like Microsoft do not themselves manufacture any physical hardware, it should be possible to simulate them entirely with AI. We'll see whether anything comes of it, but the idea does echo OpenAI's concept of Level 5 agents that are capable of doing all the work of an organization. Lastly today, a fundraising update from Anthropic, the company is nearing a deal to raise up to

Starting point is 00:06:52 $10 billion in a jumbo-sized round. The last reporting on the rumored term suggests Anthropic was looking to take in $5 billion at a $170 billion valuation. So in a matter of a few weeks, Anthropic has it doubled their target and apparently had no trouble at all filling the round. Bloomberg reports that iconic capital is leading the round, with firms including TPG, lightspeed venture partners, Spark Capital and Menlo Ventures participating. The Carter Investment authority and the government of Singapore Investment Corporation have also reportedly participated in talks as Anthropic opens their cap table to sovereign wealth funds for the first time. Now, if you're keeping track, this seems to be the round where Anthropic grows beyond the capacity

Starting point is 00:07:27 of traditional venture capital. They'll come close to tripling the $61.5 billion valuation they achieved at their last round in March, and have also tripled the scale of fundraising for this particular round. At this rate, Anthropic will overtake Open AI as the most valuable AI lab in a year. And Benedict, pointing out the public market appetite, said, if Anthropic IPOed, it would rip straight to $500 billion. Indeed, and by the way, I think that is an arbitrage that people are not taking enough advantage of, but that is a topic for another day. For now, that's going to do it for today's AI Daily Brief headlines.

Starting point is 00:07:58 Next up, the main episode. This episode is brought to you by Blitzy, the Enterprise Autonomous Software Development Platform with infinite code context. Blitzy uses thousands of specialized AI agents that think for hours to understand enterprise-scale code bases with millions of lines of code. enterprise engineering leaders start every development sprint with the Blitzie platform bringing in their development requirements. The Blitzy platform provides a plan, then generates and pre-compiles code for each task. Blitzy delivers 80% plus of the development work autonomously while providing a guide for the final 20% of human development work required to complete the sprint.

Starting point is 00:08:32 Public companies are achieving a 5x engineering velocity increase when incorporating Blitzie as their pre-I-Development tool, pairing it with their coding co-pilot of choice to bring an active STLC into their org. Blitzy is providing a limited time, 30-day free proof of concept for qualifying enterprises. The team will provide a 5x velocity increase on a real development project in your org. Visit blitzie.com and press book demo to learn how Blitzie transforms your STLC from AI-assisted to AI Native. That's BLITZY.com. If you are a regular listener, you will have heard about superintelligence agent readiness audits at this point. But I wanted to tell you today about the full suite of agent readiness.

Starting point is 00:09:13 products that go beyond just the initial readiness report. Over the last six months, Super Intelligence has built out an entire agent planning suite. We help you move from discovery to planning to implementation. After you've completed your agent readiness audits, we help you double-click on your most important use cases with what we call our use case planning reports. These reports are going to help you understand what sort of technical preparation you need to do to be ready for a use case, what challenges you might face in implementation, and whether you should be thinking about building, buying, partnering, or some combination.

Starting point is 00:09:43 After that, you can even get a spec document in what we call our technical blueprint that gives either your developers or the developers of the partner you work with what they need to build exactly the agent that you're looking for. If you want to learn more about superintelligence agent planning suite, we've built a custom GPT to answer your questions. Just go to bit.ly slash super super agent. That's bit.l.ly slash super super agent, all one word. And if you have any questions, the agent can even help you book an appointment with our team. marketers spend endless hours building segments and journeys for campaigns. High Touch just announced a new round of funding to change that with their latest product, AI decisioning. Instead of manually deciding which message goes to which customer, AI decisioning

Starting point is 00:10:28 deploys agents that act like a personal marketing specialist for every single user. These agents learn from your own data warehouse, work across your existing tools like Salesforce marketing cloud, brazen iterable, and continuously optimize for the outcomes you care about, like maximizing lifetime value. It's fully transparent, enterprise secure, and already in use by teams at brands like PetSmart, whoop, Weight Watchers, and Funrise. So if you're ready to move past static campaigns

Starting point is 00:10:52 and into the age of AI-driven marketing, check out High Touch's AI Decisioning at HighTouch.com. That's hightouch.com. Welcome back to the AI Daily Brief. Today we've got an interesting topic. On the one hand, we're covering a new AI-related backlash at YouTube after the company started implementing AI in ways that they weren't exactly clear about, angering a number of creators, but we're pairing that with some new rules of the road from Netflix

Starting point is 00:11:19 around Gen AI and asking broadly what company's strategy should be when it comes to AI and media. Now, I think background context for all of this is that we are in this interesting and frankly sort of weird moment, where on the one hand, we are seeing in general a lot more AI backlash. This has been present ever since the beginning of ChatGBTGBT, but it's one of those summer moments where it gets a little bit louder, and where the critics are jumping on a few recent happenings to really amp up the anti-AI narrative. At the same time, when you go and spend time with anyone in sort of normal society, normal work, normal youth, normal school, AI has just so quickly become ubiquitous and commonplace that it

Starting point is 00:11:59 takes a lot of the heft, frankly, out of some of those critiques. No one who's interacted with AI or uses ChatchipT needs to be so. sold on how powerful and useful this technology is. And in that context, the conversation necessarily shifts away from whether this is going to be a thing and more how we do it well. And today we're getting into a couple different instances of how to do it well or perhaps how not to do it well. Over the last few days on all the big social networks, i.e. X, threads, etc., a ton of posts started to come up like this one from Kiwi sagas. PSA for artists on YouTube. YouTube appears to be using AI to upscale your content and generate higher definition videos on their platform.

Starting point is 00:12:37 They continue, the left images are my original upload on TikTok, and the right is YouTube's version of the same content. You can even see the AI errors. Now, by the way, if you were not watching this, it's probably a little bit better watching the video on either YouTube or Spotify. But basically, we've got two images from this creator. The first in each case is a little lower fidelity, a little more blurry, and then the second is a much crisper image. Now, it would be one thing if it was just a perfect improvement in the sense of higher resolution, but this is AI and that means there can be weird artifacts, and as you can see in the second image, it's not ultimately exactly the same. It's a little bit different.

Starting point is 00:13:12 Jamrock Hobo writes, Looks like we can't showcase artwork on YouTube shorts anymore without it being automatically subjected to a horrible AI sharpener. The website is conditioning users to let actual AI content go unnoticed. Once again, we have the same thing, where you see a piece of art, and then an artificially AI-enhanced version of it. Pretend Mirrors writes, YouTube is applying post-processing to all shorts so they look more like AI.

Starting point is 00:13:36 This is intentional and malicious. When their generative AI shorts feature launches, users will be less likely to discern what's AI slop and what's not, and the audience will think AI video is just as good. So what we have here is an introduction of the concern. It's not just that creators are annoyed about YouTube doing this without their permission, it's that they think that it's going to train people basically

Starting point is 00:13:56 to not be able to tell the difference between AI generations and human-generated videos, with a host of attendant consequences. Sam Aberime writes, Please be aware of what's going on. If everything starts to look a little off and a little fake, it's much easier for bad actors to not only push through actual disinformation, but also attempt to discredit factual video sources. Again, reinforcing this idea that the problem is not just the permissions,

Starting point is 00:14:20 but instead that closing the gap between AI and non-AI video has consequences when it comes to miss and disinformation. Over the weekend, numerous outlets picked up the story, including the BBC, who wrote a piece called YouTube secretly used AI to edit people's videos. The results could bend reality. The central question in the piece, as expressed in the subhead, as AI quietly mediates our world, what happens to our shared connection with real life? The BBC writes, in recent months, YouTube has secretly used AI to tweak people's videos without letting them know or asking permission. Rinkles and shirts seem more defined. Skin is sharper in some places and smoother in others. Pay close attention to ears and you may

Starting point is 00:14:54 notice them warp. These changes are small, barely visible without a side-by-side comparison, yet some disturbed YouTubers say it gives their content a subtle and unwelcome AI-generated feeling. There's a larger trend at play. A growing share of reality is pre-processed by AI before it reaches us. Eventually, the question won't be whether you can tell the difference, but whether it's eroding our ties to the world around us. It got enough buzz that YouTube had to respond. The team YouTube account wrote, We hear your concerns and want to clarify. The enhancements on some YouTube shorts are from an experiment using traditional machine learning to improve video clarity, not Gen A.I. YouTube's creator liaison, René Richie, wrote,

Starting point is 00:15:30 No Gen AI, no upscaling. We're running an experiment on select YouTube shorts that uses traditional machine learning to unblurred, denoise, and improve clarity in video during processing, similar to what a modern smartphone does when you record a video. YouTube is always working on ways to provide the best video quality and experience possible, and will continue to take creator and viewer feedback into consideration as we iterate and improve on these features. However, for some, the explanation isn't good enough. Samuel Woolley, who is the chair of disinformation studies at the University of Pittsburgh, said, you can make decisions about what you want your phone to do and whether to turn on certain features.

Starting point is 00:16:02 What we have here is a company manipulating content from leading users that is then being distributed to a public audience without the consent of the people who produce the videos. Willie also argued that the choice of words feels like, in his description, a misdirection. I think using the term machine learning is an attempt to obscure the fact that they use AI because of concerns surrounding the technology. Now, the BBC points out that to some extent, this is just a new version of an old conversation. They write, 30 years ago, there was similar hand-wringing about the havoc Photoshop would reap on society. Decades later, we have conversations about the harms of airbrushing models and magazines and beauty filters on social media.

Starting point is 00:16:34 Perhaps AI is more of the same, but it puts these trends on steroids. I don't think ultimately that this is, in and of itself, a huge, huge deal. Even some of the creators that were quoted in these articles, ultimately say, yeah, look, they're a company, they're trying to improve video quality, I get it. YouTube's been a good platform to me. But I do think that it brings up important questions around consent and control, classic questions of the relationship between platforms and their creators, and obviously exemplifies the fact that AI, despite its increasing ubiquity, is going to be a cultural battleground in some ways. Now, speaking of that, another company that has been wrestling with its approach to AI has been Netflix. You might remember back in July on an earnings call,

Starting point is 00:17:12 Netflix CEO Ted Sarandos talked about how they had used Gen AI in one of their original programmings in the Argentina market. Sarando said, We remain convinced that AI represents an incredible opportunity to help creators make films and series better, not just cheaper. There are AI-powered creator tools, so this is real people doing real work with better tools. This year, we had El Atonata. It's a very big hit show for us from Argentina. In that production, we leveraged virtual production in AI-powered VFX. And in fact, that VFX sequence was completed 10 times faster than it could have been completed with traditional VFX tools and workflows. And also, the cost of it would just not have been

Starting point is 00:17:46 feasible for a show in that budget. So that sequence actually is the very first Gen AI final footage to appear on screen in a Netflix Inc original series or film. At the time, we discussed the fact that Netflix was one clearly trying to say that they're not using AI tools to replace people. Instead, AI is a new tool set for their people. And two, that this isn't even just a cheaper way of doing something that would otherwise have been done with a more traditional process. This is an example of where, if it weren't for Gen AI, the entire sequence would have had to be scrapped because it would never have been in the budget in the first place. In other words, a total net additive force.

Starting point is 00:18:18 Even before that, Netflix was no stranger to the AI conversation. Back in 2024, it got in a little bit of hot water when it was accused of using AI-generated images in a true crime documentary. Futurisms The Bite wrote, resorting to the tech to generate pictures of a real person, especially of somebody who's still in jail, and will only be eligible for parole around 2040, should raise some alarm bells.

Starting point is 00:18:38 This isn't inventing a fictional narrative for the sake of entertainment. This is tinkering with the fabric of reality itself to manipulate a true story that actually happened. Well, this week, the company tried to get out ahead of these types of issues by publishing a set of guidelines for Gen. A.I. Use in its content production production. Posted to their partner help center, the piece is called Using Generative AI in Content Production, and is meant as a resource for their partners. They write, this guidance helps filmmakers, production partners, and vendors understand when and how to use GenAI tools in production. It also offers a practical tool for assessing and enabling confident GenAI use when producing content for Netflix. The company lists five guiding principles.

Starting point is 00:19:15 First, the outputs do not replicate or substantially recreate identifiable characteristics of unowned or copyrighted material or infringe any copyrighted protected works. Two, the generative tools do not store, reuse, or train on production data inputs or outputs. Three, where possible, generative tools are used in an enterprise-secured environment to safeguard inputs. Four, generated material is temporary and not part of the final deliverables. Five, GenA.I is not used to replace or generate new talent performances or union-covered work without consent.

Starting point is 00:19:43 Basically, they say, if all five of those are met, quote, socializing the intended use with your Netflix contact may be sufficient, but if you answer no or unsure to any of them, you got to go get ridden approval. It's clear that one of the big things that they're thinking about is some of those issues that we were just discussing in the context of YouTube. Netflix writes, audiences should be able to trust what they see and hear on screen. Gen AI, if used without care, can blur the line between fiction and reality or unintentionally mislead viewers. In that same post, they also share a use case matrix with the red-light, yellow-light green light approach to help people understand some examples. An example of a full green light would be something like using Gen A.I for ideation only,

Starting point is 00:20:21 i.e. mood boards and reference images. The rationale for it not needing escalation is that it's low risk and non-final. For something where a company was using Gen A.I. to generate background elements like signage or posters that appear on camera. It's in that in-between category. They write incidental elements may be low risk, but if story relevant, please escalate. Meanwhile, the full red lights are things like using Gen AI to create final character designs or key visuals, or using Gen AI for talent replication, such as reaging or synthetic voices. Now, importantly, Netflix isn't saying that those things are completely off the table. They're saying that they have to be reviewed and signed off on by Netflix, not just by a third-party production company.

Starting point is 00:21:00 Now, to many, this is just incredibly common sense. Cristobal Valenzuela, the co-founder CEO of Runway writes, really nice to see this public Netflix guide for using AI and content production. It makes it easier for their partners to understand how to adopt. This is the way. And in general, I think we're going to see a lot more of this. I would anticipate that in 2026, there are less debates around the ifs of AI and more focus on the hows, and frankly, I think that will be all to our benefit.

Starting point is 00:21:25 But of course, in some ways it still feels incredibly toe-dipping relative to what the ultimate impact of AI on entertainment is going to be. To get a preview of where that might be headed, check out the new company Showrunner. Variety describes it as a tool for user-directed TV shows, and basically the idea is exactly what it sounds like. You type in a few words, and you can create entire scenes or entire episodes of a TV show, either from scratch or based on an existing IP. It was created by Edward Sachi. Variety writes, Sachi's hypothesis is that AI, instead of simply being a tool for cheaper special effects, represents a new entertainment medium, one that more closely resembles video games. He said,

Starting point is 00:22:00 using AI purely as a VFX tool is a little sad, and continued, the toy story of AI isn't just going to be a cheap toy story. Our idea is that Toy Story of AI would be playable with millions of new scenes all owned by Disney. Now obviously with that vision, the company is in talks with all the IP holders, and has investment from Amazon, among others. Whether it works remains to be seen. Even Sachi said maybe nobody wants this and it won't work. But I would expect a lot more experiments like this in the years to come.

Starting point is 00:22:25 For now that that's going to do it for today's AI Daily Brief. Appreciate you listening or watching as always. And until next time, peace.

The AI Daily Brief: Artificial Intelligence News and Analysis - What AI Backlash at YouTube Can Teach Other Companies

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.