The AI Daily Brief: Artificial Intelligence News and Analysis - The 5 Most Important AI Announcements from Meta Connect

Episode Date: September 28, 2023

Yesterday at Meta Connect, Mark Zuckerberg announces that AI was going everywhere across the Meta suite of apps including Messenger, Instagram, and WhatsApp. NLW counts down the top 5 new features. Before that on the Brief: ChatGPT's ability to browse the internet is back; Biden talks about his forthcoming executive order on AI and a survey of 1600 scientists. Today's Sponsor Netsuite | The leading business management software | Get no interest and no payments for 6 months https://netsuite.com/breakdown 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 counting down the most important AI announcements from MetaConnect. Before that on the brief, chat GPT with Browse is back and Biden prepares an AI executive order. The AI breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown Not Network to learn more 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. One of the most anticipated features that ChatGPT launched this year was the ability to browse the internet to better answer queries. Now, up until that point, ChatGPT had a knowledge cut off in late 2021, after which it simply didn't know things. So if you were asking or prompting

chat GPT for information about something that related to anything that had changed in the last two years, you were kind of out of luck. Browsing the internet obviously gave ChatGPT new capabilities, and frankly in some ways was required as a feature parity kind of. thing, given that Bard and Bing AIs had both integrated that feature right from the get-go. Now, that said, at some point along the way, people figured out that they could use Chat Chapti's browse tools to get around paywalls. That created some fairly significant legal risks for OpenAI, and so for a good amount of time, Chatchapit browse has been turned off. Now, as Sam Altman puts it, we are so back. OpenAI tweets, ChatchipT can now browse the internet to provide you with

current and authoritative information complete with direct links to sources. It is no longer limited to data before September 2021. The updates they say include identifying user agents so sites can control how ChatGPT interacts with them. Now, as for right now, browse is only available to Plus and enterprise users, but it'll be expanding out to all users soon. And for those who sign into their Plus account and are wondering where it is, you have to go over to settings into beta features and enable Browse with Bing before you have access to it. It's the same place that you turn on plugins. Now, from a sheer utility perspective, this is obviously a welcome return. And I'm sure that it has absolutely nothing to do with the timing of Meta's Connect event,

at which, as we will hear, they announced that AI was basically going to live everywhere that Meta does from Instagram to Messenger to Facebook to WhatsApp and beyond. Now, in a world where AI lives everywhere, of course, there need to be probably some new guardrails, perhaps new regulations. Right now in D.C., there are so many different groups clamoring to give their input on how to regulate this new technology field, or why not to. we've had numerous proposals from different groups of bipartisan senators, even preemptive proposals meant to get ahead of what people see as forthcoming legislation that hasn't been introduced yet.

The White House is, of course, in that mix. For the past several months, they've been engaging with executives, as well as civil rights and other types of thought leaders who have input on the opportunities and the challenges of the industry, and it appears that an executive order from the Biden administration is forthcoming. Now, it's not that we didn't know that there was going to be an EO. The White House basically announced that back in July. What's more, we don't really know what the contents are going to be. During a meeting of the Presidential Council of Advisors on Science and Technology,

President Biden repeated the same sort of it has great possibilities and also great dangers language that we've heard throughout from his administration. And so really the only new information is that it appears that this EO is coming in a matter of weeks, not a matter of months. A last note from this meeting is that as part of it, that panel of advisors showcased to Biden a number of different use cases that were positive for AI, such as using it to predict extreme weather, using AI to, quote, create materials that have properties we've never been able to create before,

and to, quote, understand the origins of the universe, which is literally as big as it gets. Now, moving back to the world of actual AI products and startups, one buzzy company called Mistral earlier this year raised a $113 million seed round. Now, this got tongues wagging and bubble accusations flowing, because again, I'm talking about a nine-figure seed round. But people who know the individuals involved in their background at companies like Meta and Google DeepMinds, said it wasn't as crazy as it seemed. Well, that company has released its first model, Mistral 7B, and as TechCrunch puts it, the model was released under the Apache 2.0 license, a highly permissive scheme that has no restrictions on use or reproduction beyond attribution.

Now, many pointed out that at least on first glance, the Mistral LLM looked really promising. The 7B model actually outperformed Metas Lama 2 13B across a variety of benchmarks. Now, over in the world of entertainment where everyone is trying to figure out the implications of the tentative agreement between the studios and writers, a new report from consulting giant Bain & Co has suggested that Hollywood should, instead of replacing those creatives, use AI to reduce cost in other areas. The report is called Tech in Content Production, Will AI Kill the Video Star? And the subtitle said it might help them.

The report writes that when it comes to using AI to replace writers or actors or visual artists, quote, studios should spurn that path. Instead, the report says, quote, they can use technology. to reduce budgets by pulling more of the production process up front and streamlining production and post-production. The savings will enable studios to make more quality content for less. By way of example, they say, think more usable minutes per day of filming and doing half your visual effects in pre-production. That means movies hit theaters or streaming platforms months earlier, some with a 20% reduction in budget or more. Now, I think this is a really sensible evolution

of the conversation. By treating AI like a binary when it comes to the entertainment industry, you just calcify each side of the conversation. What Bain is pointing out is that even the incredibly cost and profit conscious studios can still be leaning into the benefits of AI without undermining the core creatives that make their industry run. It's a really interesting report. It even goes into examples of how it could see cost savings in practice. And I think more than anything else, it reflects an evolution of the conversation,

where we've moved from this being all theoretical, to this being theoretical but applied to contractual negotiations, as in the case of this strike, to being highly-priced practical and specific, and able to actually be put into models that help businesses understand how AI can save them time and money. Lastly, today, another survey around AI attitudes, but this time it's a nature survey asking 1600 research scientists what they think. Now, a couple things that make this more interesting to me than perhaps some of the other surveys that we've gone over in the past. One, in general, you have to assume that this is a slightly more informed group, and because of that, what you see is definitely more nuance and

specificity in both their concerns and their understanding of positive aspects. These researchers, for example, are extremely excited about AI helping them process data more quickly. More than 50% are excited about it saving researchers time and money. Almost 70% are excited about it providing faster ways to process data. Now, similarly, when it comes to the negative impacts, the fears are more precise as well. Around 69% are worried that it will lead to more reliance on pattern recognition without understanding, basically scientists becoming over-reliant on AI. Concerns like entrenching bias or discrimination in making fraud easier also had more than 50% of these respondents concerned. Another really interesting benefit of generative AI specifically,

more than 50% of respondents said that they liked that it helped researchers without English as a first language. We've talked a lot on this show about how breaking down linguistic barriers seems to be one of the areas that AI will disrupt first. Now lastly, in a sign of just how early things remain, when asked how they used large language models currently, between 20 and 30% said things like to help write research manuscripts, to help do research, to conduct literature reviews, to brainstorm research ideas. But by far the most common response with over 40% was, for creative fun not related to my research.

And I think that this is really common. Right now what you're seeing is a ton of people across basically every industry in the world, experimenting with chat GPT or mid-jury or whatever tool it is, just for themselves, for the joy of it, for the interest of it, for the excitement of it. and then starting to have ideas about how they might apply it to their work or their research or their school or whatever context they find themselves in. Those people become the early adopters and their case studies of how they use it become the templates for other people who are doing similar work to them. And that's why this is spreading so fast. Overall, I think it's positive

to see a more nuanced survey, although I will say it doesn't seem like this got deep into any of the big questions of risk. And I would certainly be interested to see what this set of scientists have to say about that. However, for now, that is going to do it for today. AI Breakdown Brief. Next up, the main AI breakdown. Before we get to the main episode, I want to tell you about today's sponsor, NetSuite. Now, I know from interacting with you guys that so many of you are executives, managers, business leaders, entrepreneurs, and all of you are basically trying to figure out how technology is changing the world and how it can change your business. On that journey, I think NetSuite can be a really valuable partner for you. NetSuite gives you the visibility and

control you need to make better decisions faster. It's the software superpower behind so many of the world's most successful businesses. And for the first time in NetSuite's 25 years as the number one cloud financial system, you can defer payments of a full NetSuite implementation for six months. That's no payment and no interest for six months, and you can take advantage of that special financing offer today. NetSuite is number one because they give your business everything you need in real time all in one place to reduce manual processes, boost efficiency, build four and increase productivity across every department. If you are listening to the AI breakdown, you have a keen sense of just how important data is

to any modern business. Having all of your information in a single place can be the difference between making the right decisions and the wrong ones. I think it's great that they've created this offer to make their service more accessible to any business that needs it. If you've been sizing NetSuite up to make the switch, then you know that this deal, no interest, no payments is unprecedented. Take advantage of this special financing offer at NetSuite.

That's netsuite.com slash breakdown to get the visibility and control you need to weather any storm. One more time, net suite.com slash breakdown. Thanks to NetSuite for supporting the show. And now let's get on to the main episode. Welcome back to the AI breakdown. AI product autumn continues unabated. And today we are talking about all of the biggest announcements from Meta's Big Connect event, which was held yesterday.

Now, we are going to count these down five to one for some fun with, a ranked order where I'll tell you the announcement and then why it was a big deal, and we're going to kick it off with an honorable mention, which is all about privacy. Yesterday, META announced a new generative AI privacy guide, which is basically a way of helping people understand better things like how they've changed their models, how interacting with their AIs is different from talking with friends, and what META's new privacy policies are in this whole new AI powered world. One big thing here is that while META says that they have worked hard to, quote,

limit the possibility of private information that you may share with generative AI features from appearing in responses to other people, they've also, quote, built-in commands that allow you to delete information shared in any chat with an AI across Messenger, Instagram, or WhatsApp. They also say that meta-AI can't intrude upon your conversations it has to be summoned, or to use the word that they use, invoked. Hold aside the fact that it's spooky season and invoke sounds like something more for a witch ceremony than a technology, and we'll keep moving on. Anyway, this doesn't get an honorable mention because it's a huge deal in and of itself.

It gets an honorable mention because meta is rating it as high enough to actually have a conscientious policy around, which shows to some extent that consumers are also thinking about it as well. But with that, let's move to our main ranked list. And at number five, we're starting with these social profiles, these characters that have been leaked about and promised for weeks now. Now, basically, these are the sort of character profiles that we had seen in reports where these chatbots have different personalities based on what the user might want to interact with. Now, the goal of these different chatbots is to give especially younger users the ability to customize how they interact with artificial intelligence, even if the model underlying it is the same.

Basically, it brings a layer of personality and customization to the experience that they might not have with an otherwise generic chatbot. Now, what's new and what hadn't been reported before, is that they were partnering with a group of people from all different aspects of popular culture to, quote, play and embody some of these AIs. So, for example, there's Paris Hilton as Amber, detective partner for solving whodunits, Tom Brady as bruh, a wisecracking sports debater who pulls no punches, Snoop Dog as dungeon master, and even Mr. Beast as Zach, the big brother who will roast you because he cares. They say that from these 28 starting points, they'll be rolling out even more characters in the coming weeks. Now, you might be noticing that I put this

pretty low on the list, right? It's number five, even though in many ways it seems like they were positioning this as the most hyped and to be discussed part of the whole announcement. The reason that I'm putting it so low is that I'm just extremely skeptical that these are going to resonate in the way that meta hopes they resonate. I could be completely. wrong, I could be out of touch entirely with what young people want. That wouldn't surprise me a bit. But it feels far more likely that young people treat these like a cringe novelty rather than as an actual new social experience. Now, at the same time, the fact that they might onboard more young people to interacting with an AI chatbot that they might not otherwise have experimented with

could mean that they don't have to become a long-term default feature of meta for them to have been successful. I just remain in the skeptical camp when it comes to this specific implementation. So on that front, we're going to have to wait and see. Now, number four on our list of most important announcements from Meta's event was the announcement of AI Studio. This is basically a set of developer tools for enterprises to actually get their hands dirty and make chatbots for their companies. Meta says that the idea of AI Studio is to give companies the ability to, quote, create AIs that reflect their brand's values and improve customer service experiences. So, of course, one of the things that makes this interesting is the fact that Facebook and Instagram in particular

have become huge e-commerce platforms. People see ads for things that are related to their interests and can buy them right away from that mobile experience and actually use the chat experiences inside those apps as a way to interact with those brands. And indeed, Zuckerberg said yesterday that the main use cases that meta imagines for these AI studio tools and for these custom enterprise chatbots

are e-commerce and customer support. Now, this is a very preliminary offering. It's going to be available only in Alpha and meta plans to only begin really starting to scale the toolkit early next year. Now, interestingly, they said in addition to the release of AI Studio, they're also building what they call a sandbox tool that'll be designed to help companies experiment with creating their own AIs. So an example use case might be a small indie game studio using Meta's platform to build more interactive non-player characters, especially across Metaverse games.

Now, we have already had a chatbot trend for this type of use case of e-commerce and customer support going all the way back to 2015, 2016, but obviously the technology underlying those chatbots is hugely different now. I'm still not sure that this becomes a default way that people interact with their customers, which is why I have it still relatively low down the list at number four. But I do think that it shows how meta is thinking about not just artificial intelligence as a social engagement tool, as with their characters, but as something that can actually help the businesses that make the platform work from a financial perspective right now.

Next up, we have a new feature that's akin to things that we've recently seen from YouTube. Basically, creators who use Facebook and Instagram are soon going to have a veritable bounty of new AI-powered tools. You'll remember that when we talked about YouTube's Meet on YouTube event last week, one of the things that they had was Dreamscreen, which was a new builder of video or image backgrounds for YouTube shorts. Well, Meta is going to have something that's not totally dissimilar, but that's an AI photo editing tool where users can do things like take a photo and restile it. The example they gave is someone who takes a selfie and turns it into a Picasso painting, as well as a backdrop tool, which is effectively an AI green screen that drops in an

AI-generated background to whatever the creator's imagination is. They're also adding a new tool for AI-generated stickers across their chat experiences. So, for example, when someone is interacting on Facebook Messenger and wants to send a, quote, unicorn birthday cake, rather than having to rely on that sticker already being created by someone, they can just generate it on the fly. Now, the reason that I think these tools are more significant, perhaps, than those AI social profiles, is that these are the types of things that I think are more likely to bring some fun, whimsy and creativity back to a platform that perhaps feels like it's missing some of that right now, like Facebook. At the end of the day, the reason that we still use these apps, despite perhaps having

moved on from their heyday for some of us, is that all of our people are still there. They still bridge us to the people in our lives, especially ones that we don't interact with that often through things like our normal SMS systems. I don't know if Unicorn Birthday cake stickers are going to change the tide of getting younger people to interact on Facebook and Messenger, but AI tools that allow people to better express their own creativity with the friends that they already have, feels like an interesting thing to at least explore. I also think that this is representative of a larger trend that we've talked about a lot this fall, where many of the quote-unquote AI innovations aren't really innovations,

but more integrations into the experiences that people are already using. These tools are a great example of that, and I think are likely to be fairly successful for the company, even if in not flashy, not headline-grabbing kinds of ways. Number two is the new Rayban Meta Smart Glasses. Now, Meta has built these types of smart glasses before, but in the previous versions, the main value proposition was almost entirely about videos and imagery, in other words, capturing a picture from your glasses rather than pulling out your phone.

What's different about these, of course, are that they're powered by a new underlying technology in the form of LLMs. They write, We've integrated Meta AI, our advanced conversational assistant on Rayban Meta Smart Glasses and optimized it for a hands-free on-the-go experience. By saying, hey, meta, you can engage with Meta AI to spark creativity, get information and control features just by using your voice. Now, there are a ton of other whizbangs and gadgets and excitements for people who do live streaming. You can now go directly

live from your glasses, which is very cool. But it's this feature which to me seems so relevant. And for those who watch my video about why Chat Chachapit Vision was such a big deal, you'll recognize something similar. In the same way that ChatGPT Vision, its multimodal example, being able to input images from the real world and interact with them, both through text and through voice, meant that it opened up this entirely new world of use cases for the chat GPT mobile app. Meta's glasses are something really similar,

but in a different form factor. Let's watch a clip of Mark Zuckerberg introducing the glasses to get a sense of what I'm talking about. The next generation of Rayban meta smart glasses. These are the first smart glasses that are built and shifted. with meta AI in them. Starting in the US, you're going to get this state-of-the-art AI

that you can interact with, hands-free, wherever you go. We're going to be issuing a free software update to the glasses that makes them multimodal. So the glasses are going to be able to understand what you're looking at when you ask them questions. So if you want to know what the building is that you're standing in front of, or if you want to translate a sign,

that's in front of you to know what it's saying, or if you need help fixing this, this sad, leaky faucet. You can basically just talk to MetaI and look at it and it'll walk you through it, step-by-step how to do it. Now, so far, consumers have proven pretty resistant to this type of tech-integrated smart device, but I wouldn't be surprised if the ability to ask questions about the world around you and get real-time answers is the thing that makes the big difference and actually

breaks this into the mainstream. By one anecdotal example, I have never, ever even slightly. considered any of these products before, and I am very seriously considering pre-ordering these glasses for $299. Now, last up, number one on our list of most important announcements is in some ways the simplest one. It's the fact that meta's AI is being integrated everywhere. Hold aside the cute little characters and social profiles. Hold aside the ability to turn your selfie into a famous painting style. The fact that meta is leaning into integrating AI chat into your existing chat experiences is where I think the real power of this feature lies.

During Zuckerberg's demo, he showed a set of screens from Messenger, WhatsApp, and Instagram that all showed different use cases. In a group chat on Messenger, someone created an image of surfing the clouds. In a family group chat on WhatsApp, someone uses meta to ask for restaurant recommendations in Austin. And in a biking group on Instagram, Meta AI is invoked to use their word to discover extreme mountain biking trails in Colorado. Now, again, I'm not totally sure that any of these use cases, aside from perhaps the image generation, which has no equivalent elsewhere, is really going to become normal or standard for people.

I think that if you're planning a biking trip with a bunch of friends in Colorado, you're going to do extensive research, not just ask an AI assistant in Instagram. But I think that that matters a lot less, at least initially, than the fact that just with a simple at meta-a-i, this AI assistant can be called into the places where people are already interacting day in and day out. I don't think meta needs to know what use cases there are going to be to feel fairly confident that if they give people this technology, they will lead them to the use cases that matter. And so what you have here ultimately in the great AI battle that all of these companies are a part of is the second front, firmly and clearly articulated. The first front in the battle is underlying models, Metas Lama versus OpenAI's GPT versus Google's forthcoming Gemini.

But the second front of the battle is about integration into consumer experiences that people already have. That's where companies that already have a relationship with customers like meta and like Google think they can gain ground even if they happen to be still behind on the foundation models relative to open AI. It's fascinating to watch and a lot more options exist today for people to use this type of AI than they did before. And so I guess at the end of the day, we're just waiting for the next set of announcements that could redefine our imagination once again. However, that is where we will wrap for today's AI breakdown. Thank you, as always, for listening or watching.

If you want to come chat about these issues, come join us on the AI Breakdown Discord. You can find a link at bit.ly slash AI Breakdown. We'll see you there. Until next time, peace.