AppleGPT? Apple Has Been Secretly Working on a ChatGPT Competitor

0:00Today on the AI Breakdown, we're discussing new reports about Apple's generative AI efforts. Before that, on the brief, some really interesting statistics about how millennials, Gen Z, and other generations are viewing artificial intelligence. 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 newsletter, and our YouTube channel. Welcome back to the AI Breakdown Brief, all the AI headline news you need in five-ish minutes or less. Now today, obviously, the biggest news is that Apple is apparently working on an LLM system and an Apple GPT, chat GPT type competitor.

0:40But that is the subject of the main AI breakdown. And so where we are starting today is with a new report that's all about generational attitudes towards AI that has some really interesting things to say about what Gen Z and millennials think about the future of entertainment. Now, this report was put together by Dan Fromer. Dan is the founder and editor-in-chief of The New Consumer, and he just put together a massive report on what he calls the three things that everybody's talking about right now, AI, Ozempic, and the economy. So a few key trends. First of all, confirming what we already know, AI is absolutely everywhere.

1:14The number of quarterly earnings calls that mention AI has tripled. The percentage of articles that mention AI has also tripled or more. 90 % of US consumers 15 and up have heard of AI, and already half of millennials and Gen Z say that they are, quote, very familiar with AI. That number is 30 % for Gen X and only 10 % for boomers. And now in terms of specific tool awareness, ChatGPT is in the lead, followed by Bard and Bing, followed by Firefly, Dali, GitHub Copilot, Jasper, StableDiffusion, and MidJourney. Unsurprisingly, a lot of people are worried about how AI might impact their jobs. 28 % said they are worried that AI will replace them within five years.

1:53What's more, the people who are the most familiar with AI are the most worried about getting AI'd. People who make more say that they are already using AI tools at work. For example, 57 % of people who make between$100 ,000 and$150 ,000, 61 % of people who make between$150 ,000 and$200 ,000, and 63 % of people who make over$200 ,000 say they're already using AI tools at work. Some of the things that I found most interesting were around attitudes towards art and entertainment and creativity. Dan asked the question of whether different generational cohorts considered AI art to be art. Among boomers, 59 % said no, while only 17 % said yes, whereas millennials and Gen Z were much more likely to see AI art as art.

2:3453 % of millennials said that yes, AI art was art versus 33 % who didn't, and 45 % of Gen Z said it was art versus 39 % who didn't. Now, these two are perhaps the most interesting numbers. When asked, will AI become the dominant form of entertainment, 41 % of Gen Z and millennial consumers at least somewhat agreed that 20 years from now, most movies, art, music, and books will be created by AI. What's more, 39 % of those same generations at least somewhat agree that 20 years from now, the best movies, art, music, and books will be created by AI. I find this super interesting because on the one hand, it's not surprising to me that people would assume, even if cynically, that AI is going to take over the entertainment sphere.

3:16The fact that two-fifths of people think that the best stuff will be created by AI shows, I think, a serious strain of optimism. Now, finally, this one is really important. So much of our discussion is about AI tools as productivity enhancing or worker replacing, but Dan had the insight to ask people whether they'd actually use these tools for fun. When asked, have you used AI tools for fun or entertainment? 10 % of boomers said yes. 34 % of Gen X said yes, but 61 % of millennials and 64 % of Gen Z said that yes, they had. Investor Winnie Lau tweeted about this saying, build things that are fun and magical first.

3:55Love Dan Frommer's consumer trends report today that highlights exactly this. Most Gen Z and millennial consumers say they've used AI tools for fun. Seeing it put here really reinforces how much our conversation has not focused on that aspect of it. Lastly, there does appear to be a real generational separation when it comes to how worried versus how excited AI makes people feel. When asked that question, by far the most dominant answer for Gen X and boomers was that AI made them feel worried. 53 % responded worried, 28 % responded scared, 21 % responded stressful, and only 23 % and 17 % respectively responded hopeful or excited.

4:33When it came to Gen Z and millennials, 34 % said they were excited, 34 % said it made them feel hopeful, 28 % said it made them feel happy, 22 % said it made them feel amused, whereas 31 % and 18 % respectively said that it made them feel worried or scared. So a really interesting study, good on Dan for producing something so fascinating. Now, speaking of the use of AI in entertainment, a Reuters report today went deep into the way that South Korea's biggest record label is using AI to make K-pop and other artists on their label more accessible to a global audience. Basically, what this label is doing is that they're recording artists singing in multiple languages, but then mixing that with native speakers of those languages to make it feel more natural.

5:15They're effectively breaking down speech into its different constituent parts and reconstructing a combination of the singers as well as the native speakers to produce something that allows them to put out songs in multiple languages at the same time. For example, in May, the company put out a track from singer Minnat in six languages simultaneously, Korean, English, Spanish, Chinese, Japanese, and Vietnamese. Again, a super interesting look at how AI might break down linguistic barriers when it comes to global entertainment. Now, one of the flip sides, of course, of this positivity surrounding AI for entertainment and art is the question of whether it's infringing upon the rights of artists that models were trained upon.

5:54Well, in that, a San Francisco judge has kind of struck a blow. U.S. District Judge William Oreck said in a hearing yesterday that he was actually inclined to dismiss most of a lawsuit that had been brought by a group of artists against generative AI companies, including Stability AI and Midjourney, basically saying that they hadn't done a good job actually providing facts around the alleged copyright infringement. Oryx said that they should be able to provide a lot more given that they had access to Stability's relevant source code, and said as part of his decision, quote, I don't think the claim regarding output images is plausible at the moment because there's no substantial similarity between images created by the artists and the AI systems.

6:29As Reuters puts it, the hearing provided a first glimpse of how judges may treat a wave of lawsuits that accuse companies of misusing vast swathes of material to train their AI systems. Speaking of things done by humans today that might be done by not humans in the future, the New York Times has reported about a new Google tool internally known as Genesis, which is effectively an AI journalist. Apparently, Genesis has been demoed for executives at the New York Times, the Washington Post, and the Wall Street Journal, and was positioned by Google as an assistant for journalists rather than a replacement.

6:59Still, it sounds like the executives weren't convinced. According to the New York Times, quote, some executives who saw Google's pitch described it as unsettling. Two people said it seemed to take for granted the effort that went into producing accurate and artful news stories. Now, Google was quick to say that this was not about replacing journalists. A spokesperson said, quite simply, these tools are not intended to and cannot replace the essential role journalists have in reporting, creating, and fact-checking their articles. Instead, the idea, she says, is to provide options for headlines, to change writing styles, and things like that.

7:30Finally today, given that we saw about 40 % of millennials and Gen Z think that AI might write the best entertainment in the future, we'll end with some words from James Cameron, who very much doubts that AI can write a good script. Interviewed by CTV News, the Avatar and Titanic director said, I just don't personally believe that a disembodied mind that's just regurgitating what other embodied minds have said about the life that they've had, about love, about lying, about fear, about mortality, and just put it all together into a word salad and then regurgitate it, I don't believe that's ever going to have something that's going to move an audience.

7:59You have to be a human to write that. I don't know anyone that's even thinking about having AI write a screenplay. Let's wait 20 years, he said. And if AI wins an Oscar for best screenplay, I think we've got to take them seriously. Anyways, guys, that is going to do it for today's AI Breakdown Brief. The interesting discussions about the cultural implications of AI just continue to grow. Now, for you guys, if you are enjoying the AI Breakdown, please, if you are watching this on YouTube, hit that subscribe button. A huge percentage of you who watch these videos actually aren't subscribed yet, and I want to make sure you are up to date with the most relevant news.

8:31For those of you who are subscribed or who are listening on the podcast, I really appreciate you spending some time with us each day. And of course, I'll be back soon with the main AI breakdown. Welcome back to the AI breakdown. Today, we are talking about something that is, on the one hand, one of the most anticipated pieces of news, and at the same time, still hugely, hugely important. What we're talking about, of course, is the report from Bloomberg that Apple has been quietly behind the scenes working on its own LLM platform and chatbot, but apparently doesn't know what it wants to do with it.

9:04This comes from a Bloomberg scoop published yesterday called Apple Tests Apple GPT Develops Generative AI Tools to Catch Open AI. So here are the important details. First of all, Apple has built something that it calls Ajax. This is apparently a framework for creating large language models. And on top of Ajax, Apple has also created a chatbot that some people internal to that company have dubbed Apple GPT. Now, despite CEO Tim Cook's public comments downplaying any AI efforts, according to people with knowledge of the matter, this AI effort has become an extremely important internal project. One really key detail, I think, is that the people who were interviewed as part of this story say that Apple's AI work is trying to address potential privacy concerns.

9:50This would track with other Apple efforts that have consistently tried to position it as the most privacy-friendly big tech company. Now, it also sounds like from the reporting that despite all these tests, Apple really isn't sure exactly what they want to do with this technology. Despite that lack of clarity, markets simply didn't care. Apple's market cap went up about$67 billion in the minutes following the report, showing just how hungry these public markets are for AI competition right now. So how is the community reacting to this, though? One point that people are making is that this is very on-brand for Apple.

10:25Apple tends not to be the first of things. Instead, they try to be the best of things. As Robert Scoble points out, Apple always is late. Others have started to log their requests for what this might become. Bilawal Siddhu writes, Apple, for the love of God, please make Siri better. It's 2023 and the competition is miles ahead. I can barely use it to set reminders, let alone have a conversation. Yes, I want my own AI Jarvis, but I'll settle for parity with Google Assistant. AI developer H.J. writes, most desired features for Apple GPT. One, offline capabilities, an offline mode that allows for basic interaction with the AI even when there's no internet connection.

11:02This would allow people to use the app in areas with poor or no internet connectivity. Two, contextual awareness, improved understanding of user context based on device usage. This would allow for better, more personalized assistance, such as suggested actions based on user habits or current location. Three, multilingual support, the ability to understand and respond accurately in multiple languages, providing assistance to users from different linguistic backgrounds. Four, integration with other apps, the ability to seamlessly interact with other applications on the iPhone to perform tasks such as sending emails, making appointments, setting reminders, and playing music, among other things.

11:35Five, interactive learning mode, a feature that allows users to teach the AI about their specific preferences, improving its ability to provide tailored assistance. Six, emotion detection. While this is a complex and ethically nuanced feature, detecting the user's emotional state could help the AI tailor its responses to better suit the mood or needs of the user. Robert Scoble tried to identify gaps that he sees in ChatGPT as areas that Apple could potentially fill in. He says, my thinking is that there are a lot of holes when you travel a lot and force yourself to use GPT for everything. It can't find me a coffee in SJC airport.

12:07It completely made up a couple of menu items in an Austin restaurant. It didn't help us enough with the subway system in New York. It can't tell you what the smoke in the sky is if something in your city is burning. It can't tell you anything interesting about real-time when is the fireworks show tomorrow night and where should we stand. Basically what Scoble is asking for or seeing the opportunity in is what he calls real-world AI. But perhaps one place to look for hints about what Apple may be thinking is how they are using this internally. Going back to that Bloomberg piece, they report that Ajax was first created in 2022 as a way to unify machine learning development within the company.

12:40So far, they've made small AI-related improvements to products including Search, Siri, and Maps that are based on this Ajax system. And when it comes to this chatbot, what people have called Apple GPT, it was initially apparently rolled out at the end of last year by a very small engineering team. Now, that rollout was quickly halted over security concerns, but subsequent to that has been extended. Interestingly, it appears that it is not open to all Apple employees and requires specific approval to get access. So far, according to Bloomberg sources, Apple employees have been using it in the product prototyping and design process.

13:13However, and this is important, Bloomberg writes, quote, Apple employees say the company's tool essentially replicates BARD, ChatGPT, and Bing AI and doesn't include any novel features or technology. The system is accessible as a web application and has a stripped down design not meant for public consumption. As such, Apple has no current plans to release it to consumers, though it is actively working to improve its underlying models. Bloomberg continues, Apple is still trying to determine the consumer angle for generative AI. It's now working on several related initiatives, a cross-company effort between its AI and software engineering groups, as well as the cloud services engineering team that would supply the infrastructure for any significant new features.

13:50While the company doesn't yet have a concrete plan, people familiar with the work believe Apple is aiming to make a significant AI-related announcement next year. So what do I think about what Apple's most likely path here is? First of all, let's talk about whether they would launch a competitor to ChatGPT. The short answer for me is that I find it unlikely. Apple isn't so much in the business of standalone software experiences, particularly web-based software experiences. If you think about it, where would this live? Apple.com is effectively a store. It's not a content website. So it would represent something pretty fundamentally different for the company to create a consumer-facing web app, which leads me to think that if they were to do that, it would have to be predicated on having a really, really different approach.

14:35So far from this reporting, it doesn't seem like they have some differentiated feature or approach that would justify this shift. A place that seems far more likely is a radical overhaul of Siri, and specifically Apple taking on the idea that AI could be used to create the ultimate personal assistant. Other competitors in the space are clearly making a big effort on this sort of AI personal assistant experience. Just a couple days ago, Microsoft announced the pricing for their co-pilot feature, and at$30 a month, it seems that they have a pretty strong assessment of how important it's going to be to consumers.

15:08So is one path for Apple's AI to be the power center for their entire set of hardware devices and software systems? A souped-up AI-powered Siri that can navigate between iOS, macOS, and other future hardware platforms? That's certainly something that would be in line with what Apple's doing now and where their core strengths are. Of course, bridging from that, the place where it seems most important to have a natural language input AI-powered interface is of course around the Apple Vision Pro. The Apple Vision Pro represents the most serious attempt yet at a totally different paradigm of computing that doesn't involve screens, it doesn't involve typing, it doesn't involve pointing and clicking.

15:48And given that the natural way to control this system would be speaking commands, having a deeply integrated Apple native AI system to control this just makes a lot of sense. However, to the extent that we are looking at an Apple entrance into the generative AI space that isn't just about impacting their hardware platforms, the key differentiation that feels possible to me is about private access to personal data. As AI Breakfast writes, imagine a language model that has access to years worth of your iMessage exchanges, emails, photos, and health data. Over the years, people have gotten comfortable with Apple's focus on privacy.

16:26They do have incredible access to information about us, information that we've shared passively and actively. If you look at the development of enterprise AI and enterprise LLMs, one of the key trends seems to be companies spinning up their own models so that they can train it on proprietary data without having to worry about leaks and platforms like OpenAI's GPT inappropriately or accidentally accessing their proprietary information. It is entirely possible to me that there is a personal consumer equivalent, where what people really want out of a personal assistant AI isn't some generic third-party platform that they sign up for and give information to, but something that sits and integrates deeply with the places that already have a lot of information about us.

17:08To the extent that that is a key AI-based consumer experience of the future, there's no one better positioned than Apple to take advantage and actually build that tool. Now, of course, ultimately, my guess is that we get some combination of all of this. I can't imagine a world in which a native Apple AI isn't powering their devices like Apple Vision Pro. I can't really imagine a world where AI doesn't totally reshape how users navigate iOS and macOS and all the other Apple devices even beyond the Vision Pro. And finally, to the extent that everyone has their own little Jarvis in their pocket, their own personalized AI system, I think no one is better positioned to bring that into reality than Apple.

17:44Obviously, all of this is for now some ways out. And so for the moment, we are left to just fun speculation. Thanks for indulging in a little bit of that with me today. Until next time, peace. Thank you.