The Daily: GPT-4 is here, Snapchat's new AI chatbot, and can VR influencers save the metaverse?

On today's episode, we discuss Snapchat's new AI chatbot, Meta's generative AI plans, and what's new about GPT-4. "In Other News," we talk about whether VR influencers can save the metaverse and how drone delivery company Wing is looking to make last-mile delivery more efficient. Tune in to the discussion with our analysts Jacob Bourne and Jasmine Enberg.

Subscribe to the “Behind the Numbers” podcast on Apple Podcasts, Spotify, Pandora, Stitcher, Podbean or wherever you listen to podcasts. Follow us on Instagram

Made possible by

Tinuiti

Analyze, engage, and influence your audience with Meltwater’s data-rich suite of social and media intelligence solutions. More than 27,000 organizations around the world use Meltwater to make better business decisions faster. Learn how at meltwater.com.

 

Episode Transcript:

 

Marcus Johnson:

Hey gang. It's Monday, March 20th. Jasmine, Jacob, and listeners, welcome to the Behind the Numbers Daily: an eMarketer Podcast, made possible by Meltwater. I'm Marcus. Today, I'm joined by two folks. Let's meet them. We start with our principal analyst who covers everything social media based out of California. It's Jasmine Enberg.

Jasmine Enberg:

Hi Marcus. Hey, everyone.

Marcus Johnson:

There she is. We're also joined by one of our connectivity and tech briefings analysts, also based out of California. It's Jacob Bourne.

Jacob Bourne:

Hey Marcus. Hey Jasmine.

Marcus Johnson:

Hey buddy. So today's fact, Jasmine, this is for you.

Jasmine Enberg:

Yay. You promised me one. I'm so excited.

Marcus Johnson:

I did. Okay. Here we go. So, well, first of all, we're going to start with a bit of a geography lessons generous explanation. With the difference between Scandinavia and the Nordics, people can only use them interchangeably. I've definitely done that in the past. So the Nordic region consists of Denmark, Norway, Sweden, Finland, and Iceland, as well as the Farrow Islands, Greenland and Orland?

Jasmine Enberg:

Correct.

Marcus Johnson:

Did I say that right? Okay.

Jasmine Enberg:

Yes.

Marcus Johnson:

A tiny collection of islands between Sweden and Finland. Scandinavia is a subset of those countries, including just Norway, Sweden, and Denmark. Right, Jasmine?

Jasmine Enberg:

Correct. And I also use them interchangeably, so you're not alone. I am wrong though.

Marcus Johnson:

Like the rest of us. Some language in the regions can be understandably similar. For example, Swedish and Norwegian are very closely related. Many words are almost identical or at least understandable. However, even though Finland is often lumped in with Scandinavia, which should be wrong, although it is the Nordics, the country's language shares nothing in common with its Scandinavian friends. Finnish is part of the Uralic language family, of which Hungarian or even Estonian are the closest relations. I'm saying that with a question mark because I'm hoping Jasmine can confirm.

Jasmine Enberg:

You're a hundred percent correct.

Marcus Johnson:

Yes. Where's the bell? Quick.

Jasmine Enberg:

I did not know it was called, as you say, Uralic languages.

Marcus Johnson:

U-R-A-L-I-C. I may have pronounced that dreadfully wrong.

Jasmine Enberg:

I've been calling them the Pheno-Agrian languages, but I also could be wrong. But yes, the point is Finnish is completely different than Swedish, Danish, Norwegian, and it's closest to Hungarian and Estonian.

Marcus Johnson:

Finally, a fact of the day that's actually accurate. Jasmine is from Finland, which is why we're bringing this up, but she speaks basically every language possible. All right, good. Finally, got one, right, V. Check us out. And you took four years. Anyway, today's real. I said V, V has nothing to do with the fact of the day. So when they're wrong, it's entirely my fault. Today's real topic, what social media chatbots might look like, and the arrival already of Chat GPT-4.

So folks in today's episode, war first cover what a social media chatbot might look like and what to make of Chat GPT-4's arrival in the lead. Then for enter news, we'll discuss whether influencers can save VR and how far alphabet owned wing can take drone delivery, how far it can take drone delivery. Pun very much intended. You are welcome. We're off to a bad start. Anyway, we're talking about social media chatbots, what they might look like. We'll start with Snapchat. Snapchat is introducing a chatbot powered by the latest version. I say latest version, I guess it's powered by the old version now, right? Of open AI's Chat, GPT. So not the four. Will it automatically get updated to the four or that has to be a whole new process, Jacob?

Jacob Bourne:

I think that we can be sure that they're going to want to up use the latest GPT technology, yes.

Marcus Johnson:

Okay. When this was announced, I don't think four was out, but it will be based on open AI's Chat GPT something, probably the latest one. And the Snapchat powered chatbot is called My AI. Alex Heath of the Verge explains that it will be pinned to the apps chat tab above Conversations with Friends. It will initially only be available to Snapchat plus subscribers who pay $4 a month. The goal is to make the bot available to all of Snapchat's 750 million worldwide monthly users. Snapchat explains that my AI can recommend birthday ideas, plan hiking trips, suggest recipes, or write poetry. Jasmine, we'll start with you. What's your take on Snapchat's new AI chatbot powered by Open AI's GPT?

Jasmine Enberg:

Well, I think it was smart for Snap to roll it out first to Snapchat plus subscribers. It makes it a little bit more exclusive. It could also make more people want to sign up for Snapchat plus. But Snap says that it's not meant to be a search engine, and that's a hard sell for me, especially given all the news about Gen Z now using TikTok and Instagram for search and TikTok even potentially moving into search ads. And I don't think Snap can afford to miss out on that Gen Z trend.

Marcus Johnson:

I think that they've done a good job of positioning this in the sense of this is a smaller scale and is focused on fun. Alex Heath of the Verge was saying, the design suggests that my AI is meant to be another friend inside of Snapchat for you to hang out with, not a search engine. You can name it, customize the wallpaper for your chat. That angle I think is interesting. Can more people see it? Not like a search engine? That might be a bit bit more difficult. But Jacob, we talked previous episode talking about Chat GPT and generative AI about these more contained use cases and how they might be more beneficial for adoption because you know what you're putting into it, as opposed to putting the entire contents of the internet into this thing, you are knowing what you're putting in into the chatbot, so hopefully you know what you're going to get out of it.

And this Chat GPT that's inside Snapchat says it will adhere to the company's trust and safety guidelines, which has got to be a huge plus for both the company and what could come of this, and also for people using this. What's your take on Snapchat's new AI chatbot?

Jacob Bourne:

I mean, in theory it's great. They're adapting this technology for a specific purpose. They said that they equipped with safeguards, which is really important given that there's a young user base for Snapchat. The problem is that the Washington Post actually tested the bot and found as part of the testing, they role played as if they were teenagers, and that they found some pretty disturbing responses from the bot that's not going to fly for real time deployment. And basically the bot offered advice like how to hide the smell of alcohol on your breath, if your parents were going to delete the app, how you could continue to use it under the radar. And then most disturbingly, the bot apparently gave advice about how a minor can have a relationship with an adult. So pretty disturbing things. And Snapchat did provide disclaimers that this might happen.

I think the problem here is that when it comes to minors using these platforms, even a little slip up, it could pose a huge risk for these companies. So I think it really shows that there's a lot of pressure to adopt generative AI, but for social media especially, I think things could go really wrong pretty quickly, especially when minors are...

Marcus Johnson:

It's a great point. I mean, you mentioned Snapchat catering to a younger user base, so safety, extra importance. Snapchat's tried to get ahead of this saying, "Harmful and misleading content issues may occur," but still there is definitely rolling the dice to a certain extent with putting this type of artificial intelligence into its product. Product heavily used by younger folks. Let's move to what matters up to in the world of AI. They also added their voice to the chorus of generative AI announcements, noting that they are pulling teams together from across the company to create a new product group focused on building delightful experiences, is what they're calling them around generative AI. CEO, Mark Zuckerberg said, "In the short term will focus on building creative and expressive tools. Over the long term, we'll focus on developing AI personas. We are exploring experiences with text, think chat in WhatsApp and Messenger and images, think creative Instagram filters and ad formats, and with video and multimodal experiences."

If you have no idea what he just said, you're not alone, Jacob. What is he talking about? Decipher this for us. What the hell are meta planning when it comes to AI?

Jacob Bourne:

Well, I think the details for how it's going to integrate it into social media is pretty vague so far. We know that Meta overall is making a bit of a pivot from the Metaverse focus to generative AI just in keeping with the industry trend that we're seeing. I think the thing to know is that AI chatbots provide an immense amount of entertainment value, and I think that's something that these social media companies really want to harness for their platforms. I think how it's going to work in reality, I think it's going to take some fine-tuning. I mean, for starters, integrating at Chatbot into social media is basically allowing users to interact with the chatbot. And really social media at it at its heart is really about users interacting with each other. It kind of is a bit of a divergence from that original purpose of the platforms.

I think the other area that we might see these companies like Meta moving is using the chatbots for content creation itself, which I think could be a huge time savings, especially for users who have multiple social media accounts. I think what we might see happen is this focus on automation could initially enhance engagement, but if you think about it, again, social media is released about people interacting with each other. So too much artificial content could also distract from this user engagement in terms of people really wanting to express themselves and hear what others have to say. I think it's going to be... and it's going to pose some advantages and a disadvantages for the platforms.

Marcus Johnson:

One thing I suggested, so testing text-based AI tools on WhatsApp and Messenger. And so they're saying Meta could eventually offer to businesses in areas like sales and customer support, some kind of a Chat GPT styled conversation bot. So you'd be able to communicate with a business or a business's chatbot through something like WhatsApp, a platform that they haven't really monetized at all thus far. Jasmine, what did you make of Meta's plans for generative AI? Very loose plans?

Jasmine Enberg:

Well, in terms of what you just brought up, incorporating the chatbots into Messenger and WhatsApp, messaging commerce isn't big here in the US, but it is elsewhere. And a lot of companies that use WhatsApp in particular already have their own chatbots on the app that they use for things like customer service. I think it could be a really compelling use case, but more broadly speaking, I think it's impossible to talk about Meta's AI ambitions outside of the context of the massive layoffs and restructuring that were just announced. And Jacob, you started to allude to this. I think Meta has realized that AI is where the threat and the potential opportunity is right now. It's not backtracking on its metaverse ambitions, it is downplaying them, but it knows it needs to focus on this now to be able to remain competitive and fix the challenges, the short term challenges that it has with its business and usage.

Marcus Johnson:

So thinking about what Snapchat's doing with my AI chatbots and then Meta saying that they've got some plans in the works, what could a useful social media chatbot even look like? Is there an example of what one could potentially do that might be pretty useful in the short term?

Jasmine Enberg:

So following on that WhatsApp trend for businesses, the chatbots that exist right now are pretty basic for the most part. I think that if Meta could build or any app could build a more human-like experience for brands and customers to use, that would also automate some of the community management and customer service workflows, and it wouldn't take away from the social experience. I think that would be a pretty useful social media chatbot.

Jacob Bourne:

I mean, building on what Jasmine just said, I think the way that LinkedIn is proposing to use it really I think shows a good initial use case for it. And it's basically to help users enhance their social media profiles. I think it's in a way right now what we're seeing is generative AI is coming at an inopportune time for social media in a way. We're seeing low levels of public trust in these platforms. And I think that while generative AI tools like My AI could help boost subscriptions, it could also further erode that already waning trust among the public if it exacerbates these content moderation issues that people are already concerned about.

Marcus Johnson:

All right. Final thing we've got to talk about in the lead is Chat GPT-4. So Open AI's Chat GPT came out about six months ago, and there's already an update with open AI already announcing Chat GPT's successor GPT-4. So let's talk about that quickly. What's new? Well, Chat GPT-4 can respond to images providing recipe suggestions from photos of ingredients? It can write captions and descriptions. It can find available meeting times for three schedules. It can take a hand drawn mock up of a website and generate code for a functional site in a matter of seconds, notes sharing GAFA of Recode, it can pass a simulated legal bar exam with a score that would put it in the top 10% of test takers. The old GPT scored in the bottom 10%. In a live demo, it generated an answer to a complicated tax query. There was no real way of checking to see whether the answer or verifying to see whether the answer was correct.

The new model process is up to 25,000 words. That's about eight times as many as Chat GPT, it's predecessor, and as the BBC notes, GPT-4 will initially be available to GPT plus subscribers. They're the folks who pay $20 a month for premium access. Jacob, Chat GPT-4 is already here. What do we think of it?

Jacob Bourne:

Well, I think you outlined all the ways in which it's performs a lot more lot better compared to its predecessor GPT 3.5. I think the other thing to know is that it's new, but we've already been introduced to it even though we didn't realize it. Microsoft's Bing AI actually runs on GPT-4, so that gives you an indication of how it works and its performance. I think in addition to being more powerful, the image text functionality I think is part of the direction we're generative AI is going to be going over the next year. We're going to be seeing more of this multimodal functionality where you can use audio, images, video, text interchangeably and get output based on it. I think the fact that we're seeing GPT-4 release just a matter of months after Chat GPT's initial public debut means that we can expect a rapid fire succession of more advanced AI models to be coming out over the next year and beyond, not just from Open AI, but from rivals as well.

I think the main thing there is that these models and the updates are very expensive for these companies to produce. I mean, each one costs millions of dollars and that's on top of the millions of dollars it costs to make it publicly available to users. And the thing is that these tools right now are being provided at relatively affordable costs, which don't necessarily reflect the true costs to make them and deploy them. I think the thing for people to know is right now the profitability of companies like Open AI is still uncertain. And so these tools are being launched publicly for users to gain productivity benefits from. They're potentially remaking entire industries, and they're based on an uncertain economic foundation.

Marcus Johnson:

Within a month of its release, around a hundred million people had used Chat GPT. That was obviously the first one. So we'll see how many people adopt this type of generative AI after this new release. All right. With us, we've got time for the lead time, of course, for the halftime report. Two takeaways from the first half, one from each will start with Jasmine.

Jasmine Enberg:

I think we're in a period of a gold rush toward generative AI, and it's still very early days, and there's a lot of limitations to a lot of the tools that are out there. And I think there's going to be a lot of trial and error and a lot more iteration before we get to a place where it's a real utility for a lot of people and industries.

Marcus Johnson:

Jacob?

Jacob Bourne:

I mean, I think that social media companies at this point should tread pretty lightly in terms of adopting generative AI for their platforms. I think among industries, they face some of the highest risks to potential fallout from its use, and there's a lot of pressure to adopt it, but I think that they should make the initial use cases pretty narrow and see how things go.

Marcus Johnson:

All right, that's all we've got time for, for the lead. Time for the second half of the show today. In other news, can VR influencers save the Metaverse and drone delivery company Wing looks to make last mile delivery more efficient. Story one, can VR influencers save the Metaverse asks Mimi Swabby of the BBC because it's not in great shape at the moment. Meta's VR universe, Horizon Worlds reported 300,000 users in February of last year, 2022. The Wall Street Journal thinks that number has shrunk. Meta was hoping for 500,000 users of Horizon Worlds by the end of 2022, but it was trending towards 200,000, going in the wrong direction. The Wall Street Channel's report said most users stop logging back into Horizon after the first month. And only 9% of user-generated worlds, 9% of user-generated worlds are ever visited by over 50 people. The BBC core notes, some experts think influencers like Tech Manju, Tigress X and Thrill could be the key to boosting the Metaverse's profile. But Jasmine, can virtual reality influencers save the metaverse?

Jasmine Enberg:

No.

Marcus Johnson:

Okay, moving on.

Jasmine Enberg:

No. Well, no one can save the Metaverse because it doesn't exist yet. I think the better question, and what this article is actually really talking about, is whether influencers can get people and their audiences excited about and on board with the idea of the metaverse. And the answer to that question is yes, to an extent because audiences follow their favorite creators wherever they go, and creators can sway their perception of different topics, businesses, platforms, products, you name it. But there's still a lot of hurdles, even if you get influencers on board with the metaverse and they get their audiences on board with it. And those are things like device adoption and the optics of it being so closely tied to meta, though I think that's kind of unfair to both parties.

Marcus Johnson:

Story two, Alphabet owned drone delivery company Wing looks to make last mile delivery more efficient. Rice Insider, intelligence as senior retail analyst, Zach Stanbo, he notes that Wing expects to handle millions of deliveries for millions of people at a lower cost per delivery than ground transport by mid 2024, according to Wing's, CEO, Hal. Well, by acting like an efficient data network, not a traditional transport system, Zach explains that Wing drones will pick up, drop off, travel and charge in whatever pattern makes the most sense for the whole system. Think ride-sharing and how the cars move about to meet different retailers and restaurants peaks in demand across cities. The most interesting sentence, Jacob, in this article is what and why?

Jacob Bourne:

So the heart of the news here is that Wing is developing its Wing delivery network to function, "More like an efficient data network than a traditional transportation system." So I think on its face, I mean, this seems like a really great approach. I mean, you have the efficiency that's going to allow the drone delivery surface to scale and in scaling, that's what's going to make it economically viable. Now the thing to remember here is that these traditional transportation systems, that it's kind of comparing its approach to have decades of experience dealing with the chaotic real world issues that delivery surfaces have to face. And one of the things... and of course those challenges have not a lot to do with algorithm efficiency. I think one of the things that these drone delivery services are really going to face is number one, an extremely tightly controlled air regulation. So basically the US aerospace regulation that is going to pretty much limit how far and to what extent these drones can travel. There's issues like drones potentially colliding with power lines and starting fires. There's issues of privacy, drones flying overhead people's properties, for example.

While algorithm efficiencies is valuable, it doesn't necessarily get to the heart of those problems. And I think those are the problems that they're being delivery networks is really going to help to tackle in order to make it viable.

Marcus Johnson:

Well, the thing that jumped out to me, drones will pick up, drop off, and then it continues, travel, charge, et cetera. Hadn't really thought much about the returns potential. I think that that's pretty fascinating. You go into your backyard, drone comes down, you give it the thing, and off it goes, potentially, at least. Our forecasting team expect there to be 66,000 drone delivery units in the US this year and nearly double that next year, 110,000. That's what we've got time for for this episode. Thank you so much to my guests. Thank you to Jasmine.

Jasmine Enberg:

Thank you, Marcus. Thank you everyone.

Marcus Johnson:

Thank you to Jacob.

Jacob Bourne:

Thanks, Marcus.

Marcus Johnson:

Thank you to Victoria who edits the show. James, who copyedits it, Stewart who runs the team. Well, thanks to everyone listening in. We'll see you tomorrow for the Behind the Numbers Daily: an eMarketer Podcast, made possible by Meltwater.

"Behind the Numbers" Podcast