The AI Daily Brief: Artificial Intelligence News and Analysis - Samsung Gauss and the AI Phone Wars

Episode Date: November 8, 2023

Samsung announced a set of AI models they're calling Gauss as well as plans to put parts of the model on their next flagship Galaxy phone. NLW looks at what that means in the larger battle to build the AI phone of the future. On the Brief: YouTube announces new AI features plus some big AI fundraising rounds. Interested in the opportunity mentioned in today's show? [e-mail] 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 looking at Samsung and the AI mobile phone wars. Before then on the brief, YouTube tests new AI features and some massive AI funding rounds show the future of artificial intelligence financing. 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 YouTube, our Discord, and our newsletter. Welcome back to the AI Breakdown Brief, all the AI headline news you need in around five minutes. We've been talking a lot this fall about how one of the major themes is not just the advancement of models, although there is plenty of that as well, but about the integration of AI into the workflows that we actually use, the tools and platforms that we interact with on a day-to-day basis.

While on that front, YouTube is now testing two new AI features that firmly fit in that category of tweaks and improvements to the existing experience rather than something big and revolutionary. The first is a chatbot. It's a conversational AI tool that can summarize a video and answer questions from users. So in the screenshot they shared, they show the chatbot prompting the user, not sure what to ask, choose something. The options then given to the user are summarize the video, tell me more about this topic, recommend related content, or how big is the community on YouTube. Now, the other feature that it has, which is interesting for those thinking about the educational use case of AI, is that the chatbot can also quiz you on a topic if you're watching an educational video.

Currently, this tool is in a very limited experiment. It's only available in English. It's only available on select videos. It's only available on Android devices. And it's only available to premium YouTube subscribers in the U.S. who are over the age of 18. The other experimental tool that they announced, which has a similarly limited distribution right now, is a tool for organizing the comment section of a video. Basically, the AI takes all of the comments and organizes it into common themes or what they

call topics. Now, especially for larger creators who have hundreds and hundreds of, hundreds or even thousands of comments, this feature could provide a way to, A, get a better sense of what the community is actually engaging with and cares about, and B, actually engage with the most relevant comments. Even as a smaller creator, this is certainly something that I can see being really valuable. But again, falls into that category of not totally transformative, just a really great updated integration that shows how AI can be useful in what we're already doing. Now, of course, at this point, it feels like every couple weeks or at least every month,

there is some new AI feature update on YouTube and in the larger Google suite, so I think that we can expect to just see more of this as time goes on. Next up, we move to another big tech giant meta, who has updated their policies around AI in advertising for the coming election season. Meta Global Affairs President Nick Clegg tweets, In the new year, advertisers who run ads about social issues, elections, and politics with meta will have to disclose if image or sound has been created or altered digitally, including with AI, to show real people doing or saying things they haven't done or said.

In the blog post, they expand exactly what they're talking about. They want disclosure when something was altered or created to, quote, depict a real person as saying or doing something that they did not say or do, depict a realistic looking person that does not exist, or a realistic looking event that did not happen, or alter footage of a real event that happened, or depict a realistic event that allegedly occurred, but that is not a true image, video, or audio recording of the event. Basically, they want to minimize confusion around what's real and what's not, particularly in the sensitive areas around the elections. Now, at the same time, they do want to make clear that they're not targeting AI as a useful tool

with things like image size adjusting, cropping, color correction, image sharpening. It's just about the substance of what's actually going on in the image or video. Now, interestingly, even as Meta is releasing its own new generative AI advertising products, they're actually blocking political advertisers from using those AI ad tools. In a note, Meta said, as we continue to test new generative AI ad creation tools in Ads Manager, advertisers running campaigns that qualify as ads for housing, employment, or credit or social issues, elections or politics, or related to health pharmaceuticals or financial services, aren't currently permitted to use these generative AI features. We believe this approach will allow us to better

understand potential risks and build the right safeguards for the use of generative AI in ads that relate to potentially sensitive topics and regulated industries. Now, currently this suite of generative AI tools is only available to a limited group, but they're on track to be rolled out to all advertisers globally by next year. Now, speaking of politically sensitive, another big theme that we have been tracking here at the AI breakdown is the increasingly tight export restrictions on U.S. chip manufacturers selling, particularly to Chinese companies. The first of these rules went into place in October of last year, and around the one-year anniversary of those rules, the Biden administration introduced a further set of rules designed to close loopholes from the first.

According to a story from Reuters, these rules are already having an impact. According to Reuters' sources, Baidu has started to shift their chip buying from NVIDIA to, To alternatives like Huawei, the company apparently ordered 1,60910B Ascend AI chips, which were developed specifically to be an alternative to Nvidia's A100. Giving a sense of the scale of the business and why Nvidia has been concerned about these rules, one of the sources said that the order's total value was around $62 million. Now, Nvidia for its part has said that because there is so much unmet demand around the world, these export restrictions wouldn't have an immediate impact,

but they would, of course, over the long term, given how big a market China is. Speaking of big markets and AI, even as consumers get comfortable with new generative AI tools, many companies are positioning for the use of AI in the enterprise. One of those companies is IBM, who has debuted a new half billion dollar enterprise AI venture fund. Now, in terms of what the fund is going to invest in, IBM has said that they will invest across all stages and that they won't set any sort of target number of annual investments or capital deployment timeline. To get a sense of what they're interested in, they recently invested in Hugging Faces Series D. And according to Axios, quote, IBM is particularly interested in startups

focused on tools for specific verticals like healthcare or on a specific business process as long as the startup doesn't compete too much with IBM's own business. Axios also says that IBM is saying that it's going to avoid investing in companies that are direct competitors to startups, which can be a real challenge given how fast things are evolving in the AI space. Now, according to IBM, one of the transitions and trends that they're seeing is that although there has been a ton of AI experimentation at the enterprise level, there haven't really been clear updates that involve a return on investment. It sounds like IBM is trying to invest in companies that are catering to businesses in ways that actually achieve that ROI. Another half-billion-dollar fundraise comes from Alif Alfa,

and that round is chocker block full of big traditional firms as well. Alif Alfa is a German startup that positions itself as a direct alternative to OpenAI and ChatGPT. This $500 million funding round includes companies like Bosch, SAP, and Hewlett-Packard Enterprise as investors. Now, as CNBC points out, quote, A big part of what Alif Alpha is pushing for with its technology is a concept known as data sovereignty. The concept that data stored in a certain country is subject to that country's laws. In a European context, this means requiring that the data that powers these AI models

is grounded in Europe rather than in the U.S. CNBC continues, this concept has gained traction with high-profile European politicians and lawmakers as they seek to ensure that they aren't at the mercy of the U.S. when it comes to data storage and processing. There is definitely a sense between Aleph Alpha and Mistral that Europe really doesn't want to be left behind and to have this entire wave just be U.S. dominated as the first wave of big tech was over the last 20 years. Now, of course, whether these concepts of data sovereignty are actually compatible with that

remains to be seen, but at least these companies are getting the capital to actually compete. Stability AIs Amad Mostock wrote a little bit about the funding landscape and also intimated that they have an announcement about funding coming soon as well. Amad writes, We closed strategic funding ourselves last month, announcements soon, and are bringing in more strategic versus classical fundraising right now. This is fast becoming the norm in scale AI companies because everyone now realizes that generative AI is truly transformative, and so the strategic importance is massive, yet real expertise is rare and tough to find, harder than GPUs. Next year is likely to be more insane as we go from early adopter and realization this is big

to real competitive pressure and top-line potential. The move from early adopter to mass adoption is faster than I have ever seen, and it is tough for capital to flow to fix this. There isn't enough talent, GPUs, knowledge available even if you spend billions as many already are. I think that we will see less of the mega rounds for compute as that has already been overdone. The reality is that data is more important. For us, we focused on distribution in real partnerships versus huge clusters as we focus on accessible, optimized models, and tuning them well. I think you will see expert teams do the same, leveraging real competitive advantages from larger companies when you consider the pace of likely adoption.

Basically, Amad is pointing out that a big part of the next round of funding is going to be actually in the incumbent companies and traditional players, not just venture capitalists, which will likely have an impact on how the industry evolves. It remains a hugely interesting space as even the funding dynamics show. However, for now, we will wrap the brief here. Next up, the main AI breakdown. Hey, guys, before we get into the main part of the episode, I wanted to put out a call for a few different types of people I am trying to hire right now. The TLDR is that I have a

pretty exciting AI education-related project that I'm not really ready to share more details of just yet, but which suffice it to say I am incredibly excited about. On that front, I am looking for two types of people. The first is developers and UI slash UX designers, and the second is content producers, basically technical and non-technical builders. If you are either of those things, and you want to learn more about what we're building, send me a note at jobs at breakdown. Network, and please share examples of what you've built or created. So that can be websites or apps in the case of developers or UIUX designers, or it can be content that you've produced in the case of content creators.

Again, that's Jobs at Breakdown.network. Looking forward to hearing from you, and now on with the show. Welcome back to the AI breakdown. There are a lot of different quote-unquote AI wars happening right now. There is the race around foundation models. There are the battles between open source on the one hand and the closed source companies on the other, many of whom are connected to the arguments of the AI safety folks. But the AI war that we are exploring today is one that is just starting to emerge. And that is the battle to be the first to get AI models running locally on the hardware devices

that dominate so much of our lives. Now, the specific context for this news is an announcement from Samsung of a new generative AI model called Samsung Goss, and the report that this is likely to find its way onto the next generation of their galaxy devices, the S-24. So let's talk first about Goss itself. Goss was announced as a part of a Samsung conference in South Korea on Wednesday. The way that Niki Asia summed it up was that the model can write emails in Korean and English, summarize reports and translate documents, and create images in the forms of photos and paintings. Now, much more interesting than the technology itself, which seems on part of the technology, with and not necessarily super distinct from any other LLM that we've seen is the fact that

Samsung plans to put it directly on its smartphones. Now, Samsung has a 20% global share of the smartphone market, meaning that this is a huge player, and when it comes to the potential for AI integrations on devices like this to mainstream AI, you have to be paying attention to a company that has that sort of reach. Now, a couple other interesting details about Goss itself. It is apparently actually three models brought together, Goss language, Goss Code, and Goss image. The model was apparently developed by the research arm of Samsung, Samsung, research, and is being used by employees inside the company. At the Samsung AI forum, they said that the name, quote, reflects Samsung's ultimate vision for the models, which is to draw from all the

phenomenon and knowledge in the world in order to harness the power of AI to improve the lives of consumers everywhere. Now, one of the big questions is how much of this will be able to run locally on the device. So far, reports on that are pretty vague. For example, the Verge writes, parts of Samsung's Goss model can run locally on the device, and Samsung executives tees last month that the company will start adding generative AI to core functions of mobile devices starting in 2024. They also said in a press release that Goss could, quote, enhance the consumer experience by enabling smarter device control when integrated into products. Okay, so at this point, we're not going to get a lot more details on that, but what makes this story worthy of digging into

is what it represents in terms of this competition to get AI integrated with devices, and specifically to have AI that can run locally on devices without having to go to the cloud. This has seemingly been a big challenge when it comes to Apple's AI strategy. Apple is well known for its preference for keeping things on device as much as possible and out of the cloud, which it views through the lens both of performance and of privacy. Given that today's frontier models can't really run on phones, that's been part of the challenge when it comes to Apple figuring out what it wants to do in the space. However, from all the reports that we're getting, it seems like however much that challenge

remains, it's not one that Apple is willing to wait too much longer on. At the beginning of September, the information reported that the group inside Apple focused on AI was now spending millions of dollars per day training their AI models, even if they weren't sure exactly yet how they were going to be implemented. Now, in October, we got more reports on these efforts as well. Bloomberg reported that Apple was going to spend up to a billion dollars a year to catch up on AI, and that basically the mandate internally was to find ways to integrate it into everything. So as a, for example, AI integrated into Apple Music would create a, create auto-generated playlists. AI integrated into Xcode would assist developers building on iOS.

Bloomberg summed it up as Apple's big plan to bring generative AI to all its devices. Now, one interesting subnote about this is that Bloomberg had reported for the first time that the external-facing image that Tim Cook and Apple were trying to project of Apple not feeling behind on AI was, according to their sources, mostly just spin. wrote Bloomberg's Mark German. Chief Executive Officer Tim Cook says that Apple has been working on generative AI technology for years. But I can tell you in no uncertain terms that Apple executives were caught off guard by the industry's sudden AI fever and have been scrambling since late last year to make up for lost time. A person with knowledge of the matter told Bloomberg,

quote, there's a lot of anxiety about this and it's considered a pretty big miss internally. Now, after this came out, another set of sources said that while there was anxiety internally, it wasn't about feeling behind currently, because Apple is never the first mover. They're the ones who get things right and mainstream it. Instead, those, Those sources said that the real anxiety is coming from a lack of confidence in the particular team who's in charge of AI inside the company actually being able to deliver. Now, the story got yet another update when Apple held its Halloween-themed event at which they announced their new line of IMAX and MacBook Pros.

Now, more important than the devices themselves were the Apple Silicon chips that were running inside them. The new M3 line of chips, particularly the M3 Max, represented the first time that Apple had specifically spoken to the AI market as it made a piece. hitch for its products. Indeed, Inc. summed it up with the M3, Apple is finally talking about AI. Whereas during previous Apple events like the Worldwide Developer Conference, they hadn't actually used the words artificial intelligence. That changed at this event. For example, when discussing the M3, the company said that, quote, the neural engine is up to 60% faster than in the

M1 family of chips, making AIML workflows even faster while keeping data on device to preserve privacy. They also said that increased memory capacity on the M3 Mac supports, quote, workflows previously not possible on a laptop, such as AI developers working with even larger transformer models with billions of parameters. As ink sums up, Apple isn't just talking about the features it builds into its own products that take advantage of its own capabilities, but is also explicitly positioning its new MacBook pros as a tool for developers building AI products. That's a pretty significant change.

Now, all of this was reinforced in a Q&A with CEO Tim Cook last week, week, where Cook said that Apple was, quote, investing quite a bit in AI, and that while, quote, I'm not going to get into detail about it because we don't really do that, you can bet we're investing. We're investing quite a bit. You will see product advancements over time where those technologies are at the heart of it. Now, on that front, outside of just the iPhone, it appears that another area where AI is going to come to the fore with Apple devices is around their Apple Watch, their AirPods, and other wearables. According to Bloomberg, next year's Apple Watch is going to have some significant upgrades.

For example, those watches are reported to have a new sensor which detects when a person's blood pressure is elevated, a new system to detect sleep apnea, an AI health coach, and the company is even working to turn its AirPods into over-the-counter hearing aids. Alex Cohen tweets, I know we're all bombarded with news, right? But holy crap, how are we not all talking about this? New Apple Watch can detect early signs of hypertension, diabetes, and sleep apnea. This is an insane step forward for wearables. Now, of course, it is not just the big existing incumbents that are thinking about AI and devices, but some of the newer players as well. One of the selfie projects that has gotten the most chatter are reports that former Apple head designer Johnny Ive, and OpenAI Sam Altman, have been spending a fairly significant amount of time together exploring a new AI powered product.

The Verge says that the project aims to, quote, create a more natural way to interact with AI. And apparently the collaboration is looking for up to a billion dollars in funding from SoftBank, in order to get this new device, whatever it turns into, off the ground. Writes The Verge, according to three people familiar with the plan, the duo are aiming to create a device that provides a more natural and intuitive user experience to interact with artificial intelligence. They've taken inspiration from how the touchscreen technology on the original iPhone

helped revolutionize our interaction with the mobile internet. SoftBank CEO Masayoshi Sun is offering funding for the effort and has reportedly pushed for chip design company, ARM, which Sun holds a 90% stake in, to play a central role. Now, these reports are pretty nascent, and it also sounds like nothing has yet been committed to when it comes to the form factor or even the approach to this device. Really, it's sounding like it starts from the inspiration that the rise of artificial intelligence is going to create an opportunity for some fundamentally different type of interface.

Now, speaking of that, Sam Haltman has also invested in a company called Humane, which is itself exploring a fundamentally different approach to devices in the AI era, and which is planning on releasing a lot more information about their first device tomorrow on November 9th. I will of course be back with a report on that after that presentation comes out. In many ways, this is just scratching the surface on the AI device space. We're starting to see startups like Tab, which is a device that you wear on your body or around your neck, and which keeps track of everything of import so that you can use it as an active collaborator and the rewind pendant, which is something very similar. So maybe when it comes right down to it, it's not really

so much the AI phone wars, but the AI device wars. That's the relevant framework that we need to be paying attention to. In any case, it seems like the next year is shaping up to be a very vibrant time, not only in AI software, but about the hardware device extensions as well. I will, of course, keep an eye on it and let you know about all of the relevant things here on the AI breakdown. So thanks one more time for listening or watching, and until next time, peace.