The AI Daily Brief: Artificial Intelligence News and Analysis - Wait, Apple is Putting Google Gemini AI on iPhones?!

Episode Date: March 18, 2024

Apple is negotiating with Google to integrate Gemini AI into iPhone features, signaling a major strategy shift for the tech giant. Plus Nvidia's GTC conference kicks off. 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, Apple is in talks with Google to bring Gemini to the iPhone. Before that on the brief, it's Nvidia GTC week and people are excited. 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 around five minutes. Today we are kicking off with what some have dubbed the Woodstock of AI, the Instagram. invidia GTC conference. Indeed, they're calling it the conference for the era of AI, so they are not backing down from the hype. The event is expected to bring tens of thousands of people in person, as well as

hundreds of thousands of people participating online, and the most anticipated part of the event is a keynote from CEO Jensen Huang scheduled for Monday, where he's anticipated to premiere the company's new B-100 chips. Bloomberg writes, the rumor is that the B-100 will be the company's first multi-dive product, a breakthrough in technology whereby a larger design is partitioned into smaller chiplets for superior performance. And crucially, it's supposed to have almost double the power of Nvidia's current generation H-100. The powerhouse used to train many of the world's most capable LLMs. Bloomberg continues, Traditionally developer conferences target the engineers and computer scientists building software

and applications. The resounding theme of each I've attended this year has been AI, and NVIDIA won't disappoint on that front. What's remarkable is that Invidia will unveil a new hardware component that's attracting a level of excitement usually reserved for consumer products from big tech. They also add the color that when Jensen Huang gave his first GTC keynote back in 2009, 15 years ago, there were fewer than 1,000 people in the ballroom, whereas this year there's expected to be 10,000, and between then and now, Nvidia's shares have gone up more than 24,000%. In anticipation of the event, Wall Street analysts are definitely getting excited. Price targets for Nvidia stock have jumped between 850 and 900, up to 1,000 or even 1100.

As for Nvidia employees, they're clearly excited. Jim Fan writes, Can't wait to see Jensen's GTC keynote tomorrow the biggest Nvidia Festival of the year. Make sure you stay till the end. Our newly founded Gear Lab has something special to share, Winky Face. Bilawal C2 points out that there's also pretty incredible speakers at the event. He shared a list of the sessions that he's excited to attend, all with some pretty big hitters. I'm sure we will be back here tomorrow to give you the update on what Jensen announces during his keynote.

But for now, let's move on to our next topic, which is actually a follow-up on something from last week, Elon Musk and XAI have followed through on Elon's Twitter promise and open-sourced GROC. The Verge writes, on March 11th, Elon Musk said XAI would open source its AI chatbot GROC, and now an open release is available on GitHub. In a blog post, the company writes, we are releasing the weights and architecture of our 314 billion parameter mixture of experts model, GROC 1. They write, this is the raw-based model checkpoint from the GROC1 pre-training phase, which concluded in October 2023.

This means that the model is not fine-tuned for any specific application, such as dialogue. The weights and architecture are being released under the Apache 2.0 license, which means that this is available for commercial use. Matthew Berman sums up the excitement of the community when he writes, Elon Musk's Grock LLM was just open-sourced. It's uncensored, massive, and completely open source. Spite is a powerful motivator. Lull. Moving over into a very different dimension, the U.S. military is this week convening more than 40 countries to discuss the responsible use of AI in military applications. This builds on a political declaration led by the U.S. military that was signed by these partner nations last

year and is a way to try to get specific about what this means. Breakingdefense.com writes, 13 months after the State Department rolled out its political declaration on ethical military AI, representatives from the countries who signed it will gather outside of Washington to discuss next steps. Said a senior State Department official, we've got over 100 participants from at least 42 of the 53 countries. We really want to have a system to keep states focused on the issue of responsible AI and really focused on building practical capacity. Now, at this event, they are going to talk about literally every military application of AI. That means, of course, things like unmanned weapons and battle networks to back office systems

for cybersecurity, logistics, maintenance, personnel management, and more. Wright's breaking defense, this isn't just a talking shop for diplomats the state official emphasized. Interestingly, the State Department, at least from a narrative perspective, is very focused on how AI cannot just improve the way that we kill each other, but how to make military operations more humane. The state official said, we see tremendous promise in this technology. tremendous upside. We think this will help countries discharge their international humanitarian law obligations, so we want to maximize those advantages while minimizing any potential downside risk.

Meanwhile, another big political event is happening in South Korea at the Summit for Democracy, which is part of a U.S.-led initiative to focus on the erosion of democracy around the world. At that event, South Korean President Yun Soguel singled out artificial intelligence and specifically the fake news and disinformation from the technology as a major threat to democracy, specifically pointing the finger at countries like Russia and China who are using it to extend their propaganda campaigns. I mentioned this before, but it is very notable to me how much even as individual governments talk about regulating AI, military and geopolitical establishments are just adopting it at an incredibly rapid pace. Anyways, guys, that is going to do it for today's AI breakdown brief.

Next up, the main AI breakdown. Welcome back to the AI breakdown. We've talked extensively on this show about the laggard that is Apple's strategy when it comes to AI. Most specifically, however, we've been talking about recently how it seems like it is finally trying to catch up to put into practice its very own slant on things. For example, we've recently seen Apple actually using AI speak in marketing when it comes to things like their M3 chips and their new laptops. Obviously, the biggest shift that we've seen in Apple strategy over the last few weeks is their move to kill their electric vehicles project, Project Titan, despite the fact that they've spent over $10 billion in 10 years on it and shifted many of those

resources over into generative AI. All of those moves are why it comes as a surprise to some that Bloomberg and others now are reporting that Apple is in talks with Google to have Gemini power certain AI features on the iPhone. This is of course not confirmed. It's based on Bloomberg sources, but those sources characterize this as an active negotiation to let Apple license Gemini to power certain iPhone features this year. This of course would build on another deal between the two companies, where Google has paid Apple billions annually to have Google as the default option on the Safari browser on iPhone, but it sounds like it's not as clear cut as just an extension of that deal. Writes Bloomberg, the two parties haven't decided the terms or branding of an AI agreement or finalized how it would be implemented.

According to these same sources, Apple had also held talks with OpenAI about a similar relationship. This would obviously be a huge deal for Google. Apple has more than two billion devices in active use right now, which is obviously just a major distribution channel for Google's AI. Interestingly, there have been really different responses to this news. Markets are loving it. Alphabet shares were up as much as 7.4% as the markets opened to this news. Apple was also up 2.2%. A. B.I. Senior Industry analyst Mandeep Singh writes, Apple's potential use of Google Gemini for generative AI inferencing further lowers the risk of any near-term market share loss for the latter in search.

We believe Apple's current deal with Google, which pays 20 billion in traffic acquisition costs on iOS devices, gives the latter an incumbent advantage over Gen. GenAI-based search competitors, including OpenAI, which rely on Bing for real-time links to web pages. Wedbush's Dan Ives writes, this potential Google strategic partnership is a missing piece in the Apple AI strategy and combines forces with Google for Gemini to power some of the AI features Apple is bringing to market. For Apple, this will give them the foundation and blueprint to double down on AI features. Darrow Bassano writes, 2024 is the year Apple faced its limitations, first giving up the dream

of competing with Tesla and EVs, and now conceding it can't compete with Google and Open AI in generative AI. This means iOS users end up winning as we get actual cutting-edge features and not Siri warmed over. Robert Scoble tried to give a more nuanced take that showed why this could be a good deal for Apple. He writes, Apple gets to win with new Siri. Its own models are highly controlled, I hear trained on synthetic data, so we'll give accurate answers, but only on specific topics. If a question pops up that goes outside of that pre-done data set, it'll punt to Google. Sorry, Apple's own models can't answer that question, so we went to Google Gemini. Here's the answer. This way, answers that have hallucinations can come from Google and Apple can wash its hands of those.

Just my theory in talking to a bunch of people about this. Huge news shows that Apple can still come in and change the industry. That said, when it comes to people who are in the AI space, the reaction was pretty much exactly the opposite. Nate Chan pointed to a piece from February about how Apple had acquired 32 AI startups last year and wrote, someone explained to me, what was the point of this then? AI Breakfast writes, Did Apple completely drop the ball when it came to developing in-house AI?

100 billion R&D war chests and nothing to show for it? Outsourcing to Google's Gemini of all things. Wild. Growing Daniel writes, it's kind of crazy Apple needs to call Google in to compete on AI. Elon was able to spin up Grock pretty quickly, but Apple with 70 billion in cash needs to outsource. Abika's CEO Binu Reddy writes, wait what?

Apple can't spin up their own LLMs? Also, are Apple and Google technically competitors or do they simply collude to run the most powerful tech duopoly in the world? More on that duopoly idea and the potential regulatory implications in just a moment. Brian Romley writes, Apple and Google are in talks to integrate Google's Gemini into the iPhone, Bloomberg has reported. If true, it will be the historic mark of the full and long decline of Apple after Steve. It is the work of a money counter and not a brilliant founder.

Now, as Bindu brought up with this idea of a tech duopoly, one of the big issues here could be an antitrust issue. Bloomberg again writes, a partnership between the two Silicon Valley giants would likely draw the eye of regulators. Google's current deal with Apple for search, is already the focus of a lawsuit by the U.S. Department of Justice. The government has alleged that the companies have operated as a single entity to corner the search market on mobile devices. The pair has justified the arrangement by saying Apple believes Google's search quality is superior

to rivals and that it's easy to switch providers on the iPhone. The arrangement between Apple and Google is also under fire in the European Union, which is forcing Apple to make it easier for consumers to change their default search engine away from Google. Fortune said something similar, writing looks like Apple will need a partner to go all in on generative AI, but big antitrust issues await. Now, some have even gone farther. There is, of course, one more take, represented here by Beth Jzos, who writes, Sam Altman scared Apple and Google enough where they decided to team up and compete. We live in incredible times. Now, according to Bloomberg, even though the talks are in active negotiations, it's likely that the earliest we might hear about a deal would be June at

the WWDC conference. Bloomberg also caveats, it's possible that the companies don't reach an agreement or Apple ultimately chooses to go with another generative AI provider like OpenAI. Or Apple could theoretically tap multiple partners as it does with search in its web browser. Of course, the other possibility is that Apple just does its own thing, right? Before everyone was talking about this, the AI Twitter sphere was all talking about Apple's MM1 model. Tom's guide writes, At the core, MM1 is a new method for training multimodal models using synthetic data, including images and text. The researchers behind MM1 claim their new method speeds up performance and reduces the number of follow-up prompts to get a desired result. Being able to improve prompt

understanding and get the desired output with his little interaction with the AI as possible, he's perfect for consumer tech, especially in Siri, which will be used by a wide group of people with varying degrees of technological prowess. Dr. Jim Fan from Nvidia writes, we live in such strange times. Apple, a company famous for its secrecy, published a paper with staggering amounts of detail on their multimodal foundation model. Those who are supposed to be open are now way less than Apple. MM1 is a treasure trove of analysis.

They discuss lots of architecture designs and even disclose that they train on GPT4V-generated data. They provide exact scaling law coefficients to four significant figures, MOE settings, and even optimal learning rate functions. I've not seen this level of details from a big tech white paper for a very, very long time. Apple's so back. Brandon McKenzie from Apple also tweeted about this, so it's clear that they're not being as cagey as they normally are. So really, with all this news, we're pretty much still out in the woods when it comes to exactly what Apple's strategy is. I'm sure that we'll get clearer as the year goes on, but for now, it's just a lot of speculation. Anyways, guys, that is going to do it for today's AI breakdown. Until next time, peace.