The AI Daily Brief: Artificial Intelligence News and Analysis - Is This The First AI-Powered Operating System?

Episode Date: September 22, 2023

Microsoft announced that the complete Windows 11 operating system was going to have access to their Copilot AI assistant. Before that on the Brief, YouTube launches a slate of tools for AI creators. TAKE OUR SURVEY ON EDUCATIONAL AND LEARNING RESOURCE CONTENT: https://bit.ly/aibreakdownsurvey ABOUT THE AI BREAKDOWN The AI Breakdown helps you understand the most important news and discussions in AI.  Subscribe to The AI Breakdown newsletter: https://theaibreakdown.beehiiv.com/subscribe Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@TheAIBreakdown Join the community: bit.ly/aibreakdown Learn more: http://breakdown.network/

Today on the AI Breakdown, we're talking all about Microsoft's big AI co-pilot announcement yesterday. And before that on the brief, even more product announcements this time with YouTube. The AI breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.network for more information about our Discord, our YouTube channel, and our newsletter. Welcome back to the AI Breakdown Brief, all the AI headline news you need in around five minutes. Later today, our main episode is going to be about all the Microsoft announcements, and we kick off the brief with another set of just-announced AI products this time from Google. Yesterday, Google held its Made on YouTube event,

and basically all of the big announcements had to do in some way with artificial intelligence. The TLDR of all of this is that they are trying to make it easier for people to create content directly for YouTube. What does that look like? Well, a bunch of different things. First of all, there is a new app they're calling YouTube Create, which is a editing and production app for mobile creators. You have to think that this is going directly after TikTok and Capcom. And while this is an AI itself, it has a bunch of AI features that really give it life.

One of those is called Dreamscreen, which gives creators the ability to create AI-generated images or video for the background of their shorts. Now, of course, right now, people could use other tools to go make those images or videos and then bring them into another creation suite, but this simplifies that process down to one single place and a native prompting system. But what if you don't know what to create content about? Well, they have a new feature for that as well that they're calling AI insights. They write, YouTube Studio will tap generative AI to spark video ideas and draft outlines to help creators brainstorm.

The insights will be personalized for each channel and based on what audiences are already watching on YouTube. We've been testing a version of AI-powered tools in YouTube studio with creators, and more than 70% of those surveyed said it's helped them develop and test ideas for videos. Now, this is interesting to me as someone who obviously creates a ton of content for YouTube. Not so much because I'm bereft of ideas. Obviously, the artificial intelligence field keeps us well-stocked in terms of things to talk about. about, but I'm quite keen on the idea of YouTube cross-referencing not only what my audience is consuming, but what other audiences that are related are consuming as well to share what it thinks might be good topics. Especially for videos that aren't strictly about AI news, I can see this being incredibly

valuable. Okay, so we've got YouTube Create the new mobile app. We've got Dreamscreen, which is the AI generated image or video backgrounds for shorts. And we've got AI insights again, which they call tools to spark inspiration. But like a TV salesman of old would say, that's not all. Once you've got your idea and your dream screen background, what about the perfect music? Yes, you guessed it, YouTube has an AI-assisted search for that as well. They write, simply type in a description of your content and AI will suggest the right music at the right price. Now, finally, the one that I'm perhaps in the long run most excited about of any is a new AI-powered dubbing tool that will allow creators to dub content into other languages. YouTube acquired a company called aloud and this is that technology

coming to bear. Now, there are a lot of tools out there, many of which you might have seen on Twitter or in my newsletter that are bringing generative AI to language translation. But again, the value of having it natively in YouTube in the creation suite, especially if it works at a comparable level to those other tools, is just absolutely huge. We talked earlier this week about how the updates that Google announced for Bard weren't so much transformational and the type of thing that was going to get the techies on Twitter really excited, but were the types of things that were just highly functional developments that were going to make AI much more useful in a day-to-day kind of way. This set of YouTube announcements is absolutely in that same spirit, with none of it being some remarkable technological

advance, although automated dubbing is still pretty amazing, even if lots of companies are doing it now, but instead just better integration into the actual production suite that is likely to impact a huge number of creators. Now, staying on Google for just a moment, there's been a really interesting back and forth in the news this week around whether or not they want to ditch Broadcom as their chip supplier. Basically, the information wrote an article a couple days ago called to reduce AI costs, Google wants to ditch Broadcom as its TPU server chip supplier. The scoop argued that Google has extensively discussed dropping Broadcom as the supplier of its AI chips as early as 2027 in order to save up to 30% and subsequently fully design its

tensor processing units in-house. Now, in the wake of this news, Broadcom's stock price went down 4.3%, which is a pretty significant drop to be prompted by just a single report. Now, of course, this shows how much the market values the artificial intelligence business and is pricing in where companies fit in that matrix. After the report, Google said on Thursday that it did not see any change in its relationship with Broadcom, which helped pared back those losses. So ultimately, this was sort of a Nothing Burger story, although I'd be surprised if the information's reporting is wrong, that there had at least been these conversations internally,

but it is a good reminder of just how significant AI is to the current market environment. Now, one of the things that is likely to happen over the next couple years is that in addition to building out new software capacities on their own, a lot of the giants of software today are going to be acquiring startups to bring new AI-related services in-house. An example of that is that Salesforce is acquiringairkit.aI, which is a tool for building AI customer service agents. Now, Salesforce is well acquainted with AirKit's founders as they've also previously sold a different company to Salesforce. And again, more than this specific announcement itself, what's interesting, I think is this trend that we're likely to see of growing consolidation in the market.

Now, part of the reason that companies might want to be consolidated into the giants is that competing with them is going to be extremely difficult. As a for example, you've heard me experiment with a couple different voice cloning tools on this podcast, including play.h.t. As well as 11 labs, but iOS 17 now has a native voice cloning tool as well. To do the voice cloning, you have to go through about 20 to 30 minutes of completing different phrases, similar to how some other services are trained, and then Bingo Banking. go, you get your voice clone. I haven't tried this out yet, but I certainly will be doing so, and I will report back here about how it compares to the state of the art from companies like

11 Labs. Moving over to the policy side of things, AI continues to be a hot-button topic in Washington, with the latest news being that John Thune, the Senate's number two Republican, the Senate Minority Whip, is set to unveil a new AI legislative package that he calls light touch and is explicitly designed to offer an alternative to what he assumes or argues is likely to be a heavy-handed plan from Senate Majority Leader Chuck Schumer. Now, the legislation is not out yet, but Politico writes that it would, quote, require companies to assess the impact of artificial intelligence systems and self-certify the safety of systems seen as particularly risky.

Now, one of the things that makes this interesting is that even though it is explicitly presented in opposition to a forthcoming, as yet unseen bill from Senator Chuck Schumer, it still has bipartisan support, at least from Democrat Senator Amy Klobuchar. Others including industry executives, such as Ryan Higman and I. IBM have lauded the bill saying it strikes a, quote, very moderate balance. Now, someone who might be unlikely to find the contents of that bill particularly appealing is Connor Leahy. Connor tweeted yesterday, I had a great time addressing the House of Lords about extinction

risk from AGI. They were attentive and discussed some parallels between where we are now and non-nuclear proliferation efforts during the Cold War. Connor says, I think there should be a complete moratorium on development of AI's using unprecedented levels of computing power. Every time we do the next bigger AI run, we don't know what's going to pop out of the other side. Someday it's going to be Russian roulette, and if you ever find yourself playing Russian roulette, I have a recommendation. Stop. But the AGI companies are not stopping, and this is where

the government comes in. Now, as to Conner's recommendations for what specifically government should do, he had three pillars. One was liability for developers and users. Second was a cap on computing power. And third was a global AI kill switch. Now, of course, there was a lot of discussion around this, and frankly, a lot of disagreement. But one interesting part to help understand where Connor sits, is a back and forth he had with Jack Clark from Anthropic. Clark writes, what happens to GPT4 if your cap is implemented? Connor responds, ideally I would want it to be rolled back and deleted out of proper precaution, but I am open to arguments around grandfathering in older models.

Now, the EACC's just found this to be absolutely insane, an exemplary of how out of whack, to them at least, AI safety advocates have got. Accelerate Harder said, is this the most deranged, domer take? Gabriel Garrett says this is such an unhinged take of his, it's insane. Why are Dumer's getting these enormous stages in front of governments? The argument, of course, here being that GPT4 is self-evidently not problematic, and so if they're trying to roll back even that, can we really trust their takes on what we should be concerned about?

Now, one new person who has waited into the AI and X-risk conversation is Artifact and Instagram co-founder Kevin Sistram. In a conversation at TechCrunch Disrupt, he likened it to former technology revolutions and basically said that people always get up in arms this way because they can't imagine the benefits of the future after which these technologies have been. been fully adopted. He said people are superpowered because of these technologies, and I think that's much more likely to happen. So for now, friends, the debate rages from the stages of tech conferences, to the halls of the House of Lords, to Discord's, Twitters, and yes, even YouTube videos and podcasts.

Thanks as always for listening or watching. Next up, the main AI breakdown. Hey guys, one more quick thing before we get into the main episode. If you subscribe to the newsletter, you've seen this, and you might have heard it on an earlier episode, but right now I am getting information from you guys, the listeners about what you are looking for in terms of AI educational resources. A bunch of you have filled out the survey already and it's so helpful. But if you would take the about one minute and go to bit.ly slash AI breakdown survey, I would love to know what type of online courses you might need, what you're trying to learn more about, whether you'd be interested in a community of learners. I'm getting really close to

making some decisions about what we're going to do next and I really want all of your input. Again, it'll take about one minute and you can find it at bit.ly slash AI breakdown survey. Thanks so much. And now on with the show. Welcome back to the AI breakdown. What an absolutely massive week for AI product announcements. And as you've heard me discuss a couple times, this week wasn't really a story about new technology, although there is one exception to that.

Instead, it was a week that showed that the next phase of artificial intelligence is likely to be an integration phase, in which AI tools start to find the technology. their way into the workflows that we already use, into the software that we already use, and in so doing likely reaches a much larger audience than it does today. Now, the big announcement yesterday was that Microsoft is expanding its co-pilot AI assistant across the entire Windows operating system. In their announcement post, they write, we are entering a new era of AI, one that is fundamentally changing how we relate to and benefit from technology. At Microsoft, we think about this as having a co-pilot to help navigate any task. We have been building AI power.

co-pilots into our most used and loved products, and today we take the next step to unify these capabilities into a single experience we call Microsoft Copilot, your everyday AI companion. Copilot, they say, integrates the context and intelligence of the web, the user's work data, and whatever they happen to be doing at that moment on their computer to provide a better assistant experience. Basically, this is a big step towards an AI-powered operating system, and it's coming really soon. This is going to begin rolling out as part of a free update to Windows 11 that starts on Tuesday, September 26th. Now, this was not all they announced at this event. They also discussed some new computers, including updates to their Surface line. But as Tom Warren from the

verge said, so this was a co-pilot event with the side of Windows and Surface. The initial response was fairly positive. Linus Ekinsdam tweeted, always on AI everywhere in Windows. This is the tightest operating system integration with AI that we have ever seen. Microsoft is showing how it's done, how many rapper startups will struggle now. Now, there are two parts about this that I think are pretty interesting. First of all, one of the things that I've seen people get more excited about recently than anything else is Open Interpreter. Open Interpreter is a terminal-based interface that's effectively like having Code Interpreter from ChatGBT, run on your computer accessing all of your files.

Now, part of why people got so excited about Open Interpreter is that I think it showed a vision of a future in which a chat-based natural language control system becomes the dominant way that you interact with your computer versus having to learn a specific set of controls and it clicks in shortcuts. In other words, the idea that as Linus says this is the tightest operating system integration with AI we have ever seen is a really significant precedent for the future. The second piece, of course, that he mentions is this idea of how many rapper startups will struggle now. This has been a big theme throughout the year, where companies that are effectively just building off of someone else's API, often open AIs, are struggling to provide enough value in a differentiated user experience to be a going concern in the long run. In a world where the big tech platforms and the big software companies are all integrating their AI natively, just having a different interface is an insufficient way to differentiate for a startup.

Now, one other AI announcement that was in here was that Dali 3, which was also announced by OpenAI this week, will be integrated directly into Bing. It's another really smart way that Microsoft is leveraging its relationship with OpenAI to try to create enough value to get users to come over to their systems. Now, of course, the other side of the conversation was represented here by Nick St. Pierre, who said, I don't care how many AI features and updates Microsoft announces today, I'm still never switching to PC. And as much as Nick was joking around for the sake of Twitter, you have to think that this is

what market analysts are actually going to be asking themselves. Can Microsoft's differentiation around AI actually change consumer behaviors? On that front, I had to laugh at this thread from Rafael Rivera, who took a screenshot of a demo that Microsoft actually did at this event, in which they were running an LLM locally on one of their new service laptops, and said Microsoft leadership signed off on this Windows app for today's forward-looking AI and hardware event. What slash how the actual app? Basically, it's an extremely awkward, uncomfortable, and ugly-looking window with mismatched fonts,

no clear visual hierarchy. It just basically sums up all of the worst of what people think when they think PCs versus Max. And while ultimately this doesn't really matter, I just thought it was a funny note in the larger conversation. Now, I want to take a step back and talk about just to everything that we have gotten this week, because I think that for as interesting as Microsoft, Microsoft's updates are, and they are interesting, what's even more interesting to me is the extent

to which this really reinforces this idea of a new phase of integrated AI that we are moving into. Going back through the week, on Tuesday we got the announcement from Google that BART is now fully integrated across Google Workspace apps, so we can take advantage of the information in your Google Drive, and your YouTube, your search history, your Gmail, in order to provide better contextual results. I said in that video previewing that that this was a major upgrade not because of some technology underpinning, but just because giving it access to one's personal data is likely to make it much more useful for that person's specific needs. At Amazon's Fall Heart event, we learned all about the new Alexa, which is powered by a custom-built Alexa LLM. This is designed,

once again, to give Alexa much more power to actually understand the home environment that it's working in, and to be able to interact with the user in natural language ways. One of the examples again that they gave was being able to say things like Alexa, turn on the new light we just added, and Alexa being able to parse out what that actually meant. Now, because this is now powered by an LLM, Amazon is naturally worried a little bit more about hallucinations, and so initially this will only be available in the U.S., and they won't roll it out all at once.

Also on Wednesday, we had OpenAI's announcement of Dolly 3. Now, this is one that bridges both technology novelty as well as this integration theme, given that they are promising that Dolly 3 itself is a performance upgrade over things like Mid-Jurney, given that it can handle, as they say, significantly more nuance in detail, and not require the same sort of prompt engineering that you have with other systems because you can just interact with it through natural language.

But it is part of this integration story as well, because it is, as they say, built natively on chat GPT, and thus fluidly plugged into a system that already hundreds of millions of people use. Again, instead of having to set something up in Discord like you do for Mid Journey, you just talk to chat GPT as you normally would, getting the advantage of using it as a brainstorming partner for refining prompts, and then the results of the prompting happens through the chat GPT interface as well. For many observers who are assessing how significant the Dolly 3 announcement was, it is this integration with chat GPT, which is the biggest deal, even relative to any changes in quality

versus what stable diffusion or mid-jurdy can offer right now. So that's an announcement on Tuesday, two announcements on Wednesday, and then on Thursday, in addition to this Microsoft co-pilot event, there was also the big made-on YouTube event, which featured a slew of AI upgrades, including their dream screen feature, which allows people to put AI-generated video and images behind themselves and shorts, use AI to generate new ideas for content or even find the right soundtrack. And all of this is happening in a new dedicated mobile creator experience. Once again, integration, integration, integration. So quite clearly, one of the big themes of all of these announcements is a new phase of AI, which is

not just about racing to new higher performance technology, but about bundling it in user experiences that make it more useful and get it to a wider audience. However, there is another dimension of this week that's worth calling out as well. When Jeffrey Hinton left Google earlier this year, he argued that competitive pressure was creating new dangers as companies that had previously been responsible stewards of AI were now under too much pressure to release things fast regardless of the risks. That warning was echoed by Max Tegmark this week. The MIT professor and founder of the Future of Life Institute was interviewed by The Guardian and said, I felt that privately a lot of corporate leaders I talked to wanted a pause, but they were trapped in this race to the bottom

against each other, so no company can pause alone. Further echoing that theme is an article from the information about Jacopo Pantalione, a senior engineer at NVIDIA, who had been there since 2001, only to leave this July, leaving much economic value to himself on the table, because he realized that the world that he was building wasn't the one that he had wanted to help build. He said this market of machine learning, artificial intelligence is almost entirely driven by the big players, Google's, Amazon's metas. They have the enormous amounts of data and enormous amounts of capital to develop AI at scale. This was not the world I wanted to help build. Now writes the information Jacapo is dedicating his career to studying the unintended societal impacts of AI,

including publishing a book on the topic this month. Its premise? The concentration of power in the hands of tech giants like Google is the real danger of AI, not the human-killing AI future being propagated in the press. So this is an interesting twist, where a person who had been in the belly of the beast of these companies is saying that something needs to give and government needs to get involved, not necessarily because we're all headed towards an inevitable extinction, but because the world is going to be too fully controlled by a small oligarchy of companies. It's hard not to look at the slate of news from this week and not see at least some evidence of that thesis. And yet, at the same time, and this is the damnable thing about the industry, it's also hard

not to be pretty excited about all these new tools and new features that people like me and probably people like you are going to go use to be more productive and to create more things. That's what makes this policy challenge so hard. And that's what makes this such an interesting moment in technological history. Anyways, I'm sure we will get more announcements to go over soon. and I'm sure that it's not only integration progress that we're going to be seeing. Obviously, the next big battle is going to be around Google Gemini, and whether it can actually meet and exceed GPT4 and what that means for the rest of the industry.

But we will wrap it there, and I will just ask if you are enjoying this content, do me a favor. If you're not subscribed yet, subscribe. And if you are subscribed, I would so appreciate it if you would consider leaving a rating or a review. It makes a big difference, and I really appreciate it. Thanks as always for your attention. And until next time, peace. Thank you.