The AI Daily Brief: Artificial Intelligence News and Analysis - Apple AI? Apple Didn't Say "Artificial Intelligence" Once at WWDC

Episode Date: June 5, 2023

Those expecting a big AI announcement at Apple's WWDC were disappointed today, however those waiting for a mixed reality headset have much to celebrate. On this episode, NLW discusses how Apple VisionPro might converge with AI, and where AI did show up (even if not named) in the presentation.  Before that on the Brief: drama at Stability AI, an AI breakthrough in the writers strike, and 3900 job losses from AI last month. 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/

On today's AI breakdown, we're looking at Apple's worldwide developer conference where they made mention of AI not a single time. Before that on the brief, progress on the AI writer's strike, stability AI deals with some drama, and AI accounts for 3,900 job losses in May. The AI breakdown is a daily video and podcast about the most important news and discussions in AI. Like, subscribe and share, and go to Breakdown.network for more information. What's going on, guys? Welcome back to the AI breakdown brief. all the AI headline news you need in five minutes or less. Remember, you can get this in the morning as a newsletter as well. It's at the AI breakdown.Behive that's spelled B-E-E-H-I-I-V.com.

Today we kick off with a story that is sure to make headlines everywhere. A new research report from Challenger Gray and Christmas says that last month, employers laid off about 80,000 people across the U.S. Of those, 3,900 were from AI, just under 5%. that made AI the seventh biggest cause of job losses last month. But you have to think it's going to be the number one most discussed cause of job losses. Now, this is significant in the sense that people often think about AI jobs impacts as something that might be farther out. But statistics like this suggest that it's happening right now.

Next up, a potential deal in the Hollywood writer strike? Recently, writers in Hollywood have been on strike for a number of reasons of which AI was won. Writers had sought guarantees from studios that they wouldn't be replaced by, AI, and instead all the studios were initially willing to do was hold a meeting once a year to discuss new developments in the technology with them. The fear that writers had is that they would be marginalized by studio executives seeing AI as a way to cut costs. So imagine instead of hiring writers in a traditional way, you have chat GPT write a script, and then have writers who are being paid on a contractual basis, just punch it up or something like that. Now, over the weekend,

we saw some progress in these strikes with the Directors Guild of America, announcing a tentative deal with the studios. As part of that deal, the DGA said that they had received a, quote, groundbreaking agreement confirming that AI is not a person and that generative AI cannot replace the duties performed by members. On the one hand, some saw this as a very positive development, but on the other, some people were looking at the devil in the details. David Allen Mack writes, I suspect this wording, generative AI cannot replace the duties performed by members will come back to haunt them. Replace? No, but augment, be a mandatory collaborator? Why not? Now, impact on jobs is one of the AI fears that some people say is more pressing than the

extinction conversation that has been getting so much airtime lately. Recently, AI researchers and scientists like Jeffrey Hinton and Joshua Benjillo have been making headlines for their dire warnings about what AI might do if left unchecked. In a recent interview, however, Kyun Jung Cho, another prominent AI researcher, says that this discourse is kind of leading us astray. He said that these quote-unquote, doomer narratives are distracting from the real issues, both positive and negative posed by today's AI. Now, Cho is known in this community for his foundational work on neural machine translation,

which helped lead to the development of the transformer architecture that was used in GPT and chat GPT, and so he knows a little bit about what he's talking about. Indeed, Ben Gio was one of his former supervisors. Cho warned of overly clarifying a narrative of hero scientists and also had pretty harsh words for the effective altruist movement. He said, I am very aware of the fact that the EA movement is the one that is actually driving the whole thing around AGI and existential risk. I think there are too many people in Silicon Valley with this kind of savior complex.

Now, interestingly, Gary Marcus, who just testified alongside Sam Altman in front of Congress, and who is very concerned about a number of different issues that relate to AI, has also come out loudly with a similar argument in the last couple weeks, saying that we're effectively spending too much time on the existential risk questions and not enough time on the more pressing and immediate questions. Now, even as all this gets debated in the media and in policy circles, companies in the AI space continued to proceed at a blistering pace. However, over the weekend, one of those companies' Stability AI was set back just a little bit by some drama around a Forbes article, which argued that its CEO had a habit of misrepresentation, fabrication, exaggeration, and

even outright lying. Now, Amad, the CEO in question, almost immediately published a long list of responses to the article, and I think wherever you land on it, it shows the type of thing that is likely to happen the more that these companies are in the public spotlight. Rightly, wrongly, everyone involved is going to be under intense, intense scrutiny. Finally, let's close the day on one of the more exciting visual tools we've seen for a while. This is called Google Style Drop, and effectively it's an approach to text to image that can use a reference image to create a set of other images all in the style of that first image. So some of the reference examples they have are a watercolor image of flowers,

an old sort of 90s line graphic design, and Vincent Van Gogh's starring night. The results were StyleDrop is prompted to create an object in the style of that reference image are pretty remarkable. Now, this is all powered by Muse, which is a text-to-image generative vision transformer. The summary of the research paper says, StyleDrop works by efficiently learning a new style of fine-tuning, very few trainable parameters, and improving the quality via iterative trading with either human or automated feedback. Better yet, StyleDrop is able to deliver impressive results, even when the user supplies only a single image specifying the desired style. An extensive study shows that for the task of style tuning text-to-image models,

style drop on muse convincingly outperforms other methods, including Dreambooth and textual inversion on Imogen or Stable Diffusion. This tool is likely to be extraordinarily useful for designers who are trying to keep things in a particular aesthetic. And that's basically any designer who's creating anything more than just a single one-off image. From a creative side, imagine a graphic novel that needs all the images to conform to a particular style or mode or, vibe. And then think about brand designers who have to keep everything within the guidelines or the

particular aesthetic of an enterpriser company. While this remains just in research right now, it's one of the tools that I'm more excited about than just about anything else I've seen for the last couple of months. And I mean that from the standpoint of a practical I want to use it right now kind of way. Anyways, guys, that is it for today's AI breakdown brief. If you're enjoying it, please like, subscribe and share and turn on your notifications. And I will be back soon with the main AI breakdown. Apple just had its biggest developer event of the year, and it didn't say AI or artificial intelligence even once. Welcome back to the AI breakdown.

Well, today, obviously, all eyes were tuned in on Apple. It was the Worldwide Developer Conference, which is Apple's biggest annual developer event. And really, it's not just for developers. It's for anyone who's interested in Apple's new product lines and whatever big announcements they happen to have. And this one was anticipated to be big. but for different reasons than some other developer events we've had over the last couple weeks. There was, of course, Google I.O. as well as Microsoft Build, and at each, those companies use their event as a chance to make major AI-related announcements. Google had so many AI

announcements I was able to do a whole video counting them down, but the one that stood out was, of course, Google modifying its core search experience to incorporate generative AI in a more fluent way. They're now rolling out their AI-powered search experience in beta, and in a lot of of ways it represents the single biggest change we've seen to the core of internet search in 20 years since Google first launched. Meanwhile, over at Microsoft, they had as well a number of huge announcements. One big one was that they were making browse with being available to all chat GPT users, meaning that the average baseline free chat GPT experience would now be connected to the internet, but for most people, the biggest announcement from that Microsoft event was that Windows was getting

AI natively in the form of co-pilot. Microsoft is basically trying to position Windows as the first AI native operating system. So then coming into WWDC, there were two huge things that people were wondering about when it came to Apple. The first was, would Apple enter into the AI and specifically the generative AI race? CNET pointed out that in the past, Apple has adopted a wait and see approach. The iPad, for example, wasn't the first tablet, but in the estimation of many, it was the best, and certainly it's been the most category defining. CET also pointed out that Apple hasn't dived into a phone that can fold. When Apple CEO Tim Cook was asked about the topic on the May earnings call, he said, I do think it's very important to be

deliberate and thoughtful in how you approach these things. And there's a number of issues that need to be sorted. AI is being talked about in a number of different places, but the potential is certainly very interesting. Now, one of the big challenges for Apple specifically is that it's built a really strong brand around being the most privacy-centric of all of the big tech companies. And there are many parts of AI that are inherently privacy invasive, in terms of model training and where that data comes from, which potentially put some limits on Apple in the short term. Anyway, coming into the event, one of the big questions was whether Apple would finally dive headfirst into this AI race. But of course, the even more anticipated aspect of this WWDC was Apple entering

into the headset space for virtual reality, augmented reality, and mixed reality. There really was this interesting context where it seemed pretty clear that Apple was going to announce a new competitor to something like the Oculus Rift, but that the broader context where everywhere else is is talking about AI. In a freed from Axios writes, Apple's new headset will land in an AI-crazed world. Apple is used to hogging the tech spotlight, she writes any time it has a major announcement, and its widely expected unveiling of a mixed reality headset on Monday would normally do just that. Yes, but, the long-awaited new product is arriving at a moment of industry frenzy over AI that's putting Apple in the awkward position of trying to change the subject. So what did we get? Well, we did, in fact, get the long

anticipated Apple mixed reality device. It's called the Vision Pro, and we're going to get all into the specifics of that in just a moment. Now, what we also didn't get was any mention of AI basically at all. In fact, the word AI or artificial intelligence was nowhere to be found. That said, the story wasn't quite that clear, so let's talk a little bit about the visual. Pro, and then we'll talk about where AI did find its way into this presentation. So as you can see, they're really talking about a new category, not just a new type of device. They're calling it spatial computing, and it's a world in which the computing experience integrates seamlessly with the real world, or at least that's the goal.

Now, part of Apple's unique take on that is that this is not a device that you look at. It is a device that you look through. It can project its versions of screens. You can toggle it to be more or less immersive, but even when it's in its main, most immersive setting, it will recognize if there's a new person in the room and will then actually take you out of that experience so you can engage with them. This, of course, is a way for them to make it not a disconnecting experience with the real world. A big part of the use case is, of course, entertainment. All of their demo videos show people who are using it to watch movies on exactly the size screen

that makes sense for the room or to create an immersive environment around themselves on a plane. And it also does plug in with game controller so that gaming experience, can be had through the Vision Pro as well. That said, it's also for work. It plugs in with Macs and keyboards and effectively allows people to create an entire immersive set of screens or screen-like experiences all around them. One of the most important new aspects of it is that it's not controlled by a controller. You navigate it using eye motions, using hand gestures, and using your voice. So how are people reacting to it? Well, a lot of people think, as cool as it might be, it still just looks kind of dumb? Stephen Lubka tweets billions and they couldn't make people

wearing the headsets not look stupid. Another common negative response was that it felt sort of out of sync with where markets and society are. Gaborger box rights, rates are 5% plus. Credit card APRs are over 25%. Inflation is high. People can't afford homes. Apple? New face television will be $3,49. You get it sometime in 2024, maybe. Yet for those who were skeptical, there are also a lot of people who were impressed. Venture capitalist Chris Freelich writes, I agree that it's the most ambitious product they've ever created. It melds VR and AR and changes everything about this space. You saw a lot of people who had statements like this. Alex Valadis writes, Apple just killed Meta's AR and VR dreams. Their initial offering already feels

miles beyond anything I've seen from Oculus. Now, interestingly, Metaverse was a word that Apple also didn't say, even though the implications of the Vision Pro for Metaverse applications are pretty clear. The big thing that's really clear is that Apple is treating this like an entirely new category. And if you had any doubt about that, just look at Tim Cook's one tweet about it. Welcome to the era of spatial computing with Apple Vision Pro. You've never seen anything like this before. Now, some people were really ready to write this off before it even began. Professor Scott Galloway said that he thought it was going to be one of the biggest tech failures in history and was going to finally put a nail in the coffin of the idea of this headset-based

paradigm for virtual reality or augmented reality. Did he feel the same way after the event? Well, given that he tweeted, ordered an Apple headset, also canceled my vasectomy, no longer needed. Hashtag prophylactic, I think that he did indeed. And unfortunately for Apple, investors seem to kind of agree with Galloway. Apple was down on the day more than the NASDAQ index, although I don't know how much we should read into this. Leading up to the event Apple had been bid up to all-time highs, and so to some extent this might have been a buy-the-rumor-sell-the-news type of event no matter what Apple said. People certainly like the fact that Apple said it was partnering with Unity because Unity stock

jumped 17% on that announcement. Now, if you're a careful watcher of this show, you will know that Unity just opened its AI creation beta. Think text to game, text to world. And so might there be more room for the convergence of Apple's Vision Pro and AI than it initially seems, it's totally possible. And in fact, with that, let's switch over to what the discussion of AI was, or better put, wasn't. A couple days ago, developer Nate Chan wrote, I like the theory that Apple won't mention artificial intelligence at all, and instead

patent a term like Apple intelligence and use it repeatedly during WWDC. Hell, he might even use the term Apple reality instead of augmented reality. Apple steers the industry. Now, after watching the whole event, Nate retweeted himself and said they actually did it. Zero. uses of the term AI. But for that, it wasn't like AI was nowhere to be seen. It just wasn't called that. In fact, if you were listening closely, you might have heard the word transformer a number of times. iOS autocorrect, for example, which has always been powered by machine learning, was now referred to as being transformer powered. iOS dictation was also transformer powered. Now let's dig into this idea of autocorrect being transformer power just a little bit more.

A couple months ago, Nvidia's Dr. Jim fan wrote, million dollar idea LLM keyboard. Every time I type on my phone and autocorrect makes a stupid mistake, it screams LLM. This is literally next word prediction. We should be typing 10 times faster. Input methods need serious upgrades. The LLM doesn't have to be big and can be optimized to run locally to reduce latency and keep privacy. It also needs no prompt engineering or instruction tuning.

Combined with methods like swipe type, LLM keyboard could in principle render full sentences with an unbroken thumb movement. We'd finally be able to type at the speed of our stream of consciousness. Jim follows up today watching WWDC Apple actually implements the LLM-powered keyboard I mentioned three months ago. The native transformer model runs inference on device and learns your typing patterns. Akash Gupta also picked this up saying, breaking, Apple enters AI. Apple announced transformer-based autocorrect that use LLMs to help you write messages even faster. He said it makes a serious problem easier.

Apple's current autocorrect was stiff and nothing compared to chat GPT. Now Apple has solved a real problem, typing on mobile is hard, with a chat GPT-like technology. Another example that had this sort of AI integration was live voicemail transcription. This is a new feature where you can get a live transcription when someone calls you to see if you want to pick up. That means in his estimation that, quote, the battle against spam has completely changed. Now still, the more interesting integrations with AI come somewhere in the future.

One of the things that a number of people picked up on was that this was the first device to have a 3D camera built in. Now, the way that Apple positioned it was as a totally differentiated way to capture wonderful memories and re-experience those things over and over again in the future, but there are clearly more implications of a technology that can capture photos and videos in 3D. Nick St. Pierre, who's done a lot of experimenting with Neural Radiance Field, says 3D camera LFG, this Nerf tech? So will we start to see people experiment and integrate this 3D camera capacity with some of the

text of video and other AI-powered video creation tools that are coming online? seems totally possible. Other people zoomed in even farther and just started to think about having this device as standard plus an eventual Apple LLM. Sully Omar writes, calling it now, Vision Pro plus Apple LLM. You'll be able to do work by just speaking naturally. It's literally Jarvis in real life. His point is that if you view this as Apple's major plan, as perhaps to use their terms the next generation of spatial computing, they're kind of working on the harder problem first, which is, of course, the new device that gets people comfortable with this type of computing experience that's no longer a phone or a computer. Once that's done, integrating a large

language model seems pretty trivial. Lastly, one more thing that is an AI, but does matter to Sam Altman. Another of Sam's project is called WorldCoin, which is an attempt to create a cryptocurrency that can be used for the whole world. They have concerns around fair distribution and basically want to give everyone on the planet some amount of WorldCoin. Their solution has been an iris scanning device that people in the Bitcoin space find incredibly dystopian and terrifying, but Apple's new Vision Pro uses iris scanning technology as its main privacy tool to turn on the device. In fact, Maya Zahavi tweeted, the entire Apple pitch seemed to me to be a direct threat on WorldCoin. Imagine an I-vision photo of your retina secured with a zero-knowledge proof in the secure enclave.

Voila, a safer, more trusted solution for verified credentials on the iOS. Or, as Kermin-Cole put it, lull, WorldCoin just got out shipped with IrisCan by Apple. So that is the story from here. For those of you who wanted an Apple entrance into the generative AI battle, you will have to sadly wait. But perhaps we got a preview of some of their plans and where this all might intersect in the future. That's going to do it for today's AI breakdown. Let me know in the comments what you think about this particular announcement. Is Vision Pro what you expect?

Do you think people are going to actually use it? Do you see an integration with AI in some of the ways that I shared? And of course, if you found the AI breakdown today useful, please like, subscribe, and share. Check out the podcast and the newsletter. And until next time, peace.