The Daily: Who is actually using GenAI, what they’re using it for, and Nobel physics and chemistry prizes go to AI pioneers

On today’s podcast episode, we discuss what GenAI platforms people will be using next year, the main ways people are using the technology, and what to make of AI researchers who were just awarded the Nobel prizes in physics and chemistry. Join host Marcus Johnson, along with analysts Jacob Bourne and Grace Harmon, for the conversation.

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

Episode Transcript:

Marcus Johnson (00:00):

See how this episode is sponsored by TikTok for Business? Very mindful, some might say. Using their measurement solutions to reveal the true impact and performance of your ads. Seems very demure. Learn more about better measurement on TikTok at tiktok.com/business.

Jacob Bourne (00:20):

I mean, they'll still be relevant, especially for the workplace and for certain use cases, but I think when people can access the technology through things like Meta's AI assistant on social media, for example, I think that will diminish the need to really go into a standalone platform.

Marcus Johnson (00:41):

Hey gang, it's Thursday, October 17th. Grace, Jacob and listeners, welcome to Behind the Numbers Daily: An eMarketer podcast made possible by TikTok. I'm Marcus. Today I'm joined by two people. We start with one of our analysts who writes for our technology and AI briefings. She's based in California. It's Grace Harmon.

Grace Harmon (01:02):

Hi guys, nice to be here.

Marcus Johnson (01:03):

Hello, Grace. And we also have with us one of our technology analysts. He's also based in that same state. We refer to him as Jacob Bourne.

Jacob Bourne (01:12):

Hello, nice to be here as well.

Marcus Johnson (01:14):

Hey, chap. Today's fact, so it is specifically for you guys. That's your hint. What is the largest tree in the world?

Grace Harmon (01:23):

Redwood.

Marcus Johnson (01:24):

Bang. It is indeed. It's the General Sherman, which is the flavor of Redwood and it lives in Sequoia National Park in California. Have you guys been to see this or these?

Jacob Bourne (01:36):

I have, but it was a long time ago. But it's impressive, very impressive.

Marcus Johnson (01:39):

Grace.

Grace Harmon (01:40):

I went to see the Redwoods, I think in the Redwood forest when I was a small kid and they were monstrous then. I'm sure they're a little more reasonable size now.

Marcus Johnson (01:49):

275 feet, for folks using the metric system, 84 meters. For reference, that's nearly as tall as the Statue of Liberty from the bottom of the base to the very top, so that's how tall they are, very tall. Tallest tree in the world. Also, 36 feet in terms of diameter, so 12 yards wide so to speak, and not only that, over 2000 years old, the General Sherman was born around the same time as Julius Caesar. I can't wait. I'm kind of looking forward to seeing them now.

Grace Harmon (02:23):

They're pretty spectacular.

Marcus Johnson (02:24):

They sound amazing.

Jacob Bourne (02:25):

Awe-inspiring for sure.

Marcus Johnson (02:27):

Yeah. Anyway, today's real topic, how many folks are actually using Gen AI and what for? In today's episode, first in the lead, we'll cover Gen AI usage. Then for other news, we discuss some AI breakthroughs that just won Nobel Prizes. We start of course, with the lead, and as I mentioned, we're talking about how many people are using Gen AI and what for. So Jacob, I'll start with you. Based on the numbers that you've seen, how many folks are using Generative AI?

Jacob Bourne (03:01):

Yeah, well, some recent survey data from OpenText found that 39% of consumers use Gen AI at least weekly, so quite a bit there. A recent McKinsey survey found that 91% of the employees who responded to the survey said they use it for work, which is enormous. According to a Harvard economist, 40% of Americans ages 18 to 64 have used it at least once, which is actually a faster rate of adoption than with the internet or personal computers. So pretty stunning. I think one interesting thing about adoption though is that despite the fact that most of the major AI companies are located in the US, the biggest users are not in the US. Actually, a Boston Consulting Group survey found that India was the largest user of ChatGPT, followed by Morocco and then the United Arab Emirates.

Marcus Johnson (03:52):

Give me that number of people using it at work, did you say, or at companies.

Jacob Bourne (03:55):

Yeah. McKinsey did a survey and found that 91% of the respondents employees using it said they use it for work.

Marcus Johnson (04:02):

Interesting. Because 92% of Fortune 500 companies are using AI products according to Axios. Based on the numbers that you've seen, Grace, what'd you make of how many people are using Gen AI and the numbers Jacob cited as well?

Grace Harmon (04:17):

Well, I think there's going to be different applications. Like he just said, it doesn't necessarily mean that at every workplace they're using ChatGPT's chatbot, but there are a lot of new AI applications coming out. On the sales front, we've been writing a little bit for the briefings about the new digital twin technology, Gen AI ad targeting, things like that. So there's a much broader way that it can be applied in the workplace other than just using it to organize emails or send a pitch email, things like that, or figure out how to design your website. So I think that's part of what's helping to lead adoption at the workplace is that we're figuring out how to apply it in a much broader range.

Marcus Johnson (04:57):

Yeah, we do have some numbers on this. Our forecasts are monthly usage, so different from the weekly ones that Jacob mentioned, but we have a hundred million Americans or a third of everyone on the internet using Gen AI. In two years time we expect that to get to 41% of folks. And then when you look at terms of people who are using it for work, we think that about just over a third, 37% of Gen AI users are using it at work. That goes up to 43% next year. One of the other breakdowns I thought was interesting folks, was by generation, always just a tale of two different species. When you look at what's going on across the ages, close to half of folks under the age of 45 are using Generative AI, that's crossing 50% next year for people 45 and older adoption closer to 20 to 25%. So a big discrepancy there as well.

Grace Harmon (05:54):

Specifically in the workplace?

Marcus Johnson (05:56):

That one, no. It's a good question. That one is a broadly broad use of Gen AI.

Grace Harmon (06:01):

Interesting.

Jacob Bourne (06:01):

I mean, I think for the over 45 crowd, it's lower, notably lower, but still pretty high for new technology I think.

Marcus Johnson (06:10):

Yeah, probably because I think for Gen X was closer to 31% of them using Gen AI versus 15% for Boomers. So we spoke about how many people are using it broadly, but if we look at AI tools that people are favoring over others, the highest web traffic, this is from back in March in terms of AI tools, the highest web traffic was ChatGPT, probably a surprise to no one, 2.3 billion visits. That's 18 times more than second place, Gemini, which comes from Google and then that's also 50 times higher than third place Poe, according to the World Bank. Jacob, what do you make of current platform use and what's that likely to look like over the next year or so?

Jacob Bourne (06:57):

Yeah, and I think the most interesting thing here to me is how high Poe, that platform you mentioned is on the list because Poe is really just a clearinghouse of all the leading Generative AI models, both chatbots and image generators, and so it shows that what I think we're going to see more of and that's that users want to see variety. They don't want to be locked into one platform necessarily because I think the different models are good at different things, and so Poe gets you access to some of the leading models in one place.

(07:27):

I think the other thing we're going to see in a year's time is just the impact of the rise on device Gen AI. Things like Apple Intelligence is going to make the standalone Gen AI platforms a little bit less relevant. I mean, they'll still be relevant, especially for the workplace and for certain use cases, but I think when people can access the technology through things like Meta's AI assistant on social media for example, I think that will diminish the need to really go into a standalone platform. Now that said, I think the third thing that's going to happen is that these standalone platforms are going to become easier to access because they're going to be available on a wider variety of devices, even in vehicles, for example, and I think we're going to see people being able to access the models just via voice commands versus mainly going into an app, which I think just means overall we're going to see more adoption, more variety of different tools being used and for a wider variety of use cases.

Marcus Johnson (08:27):

Yeah, I mean ChatGPT is well out in front currently, according to the company OpenAI, they say about 200 million people are using its AI chatbot ChatGPT on a weekly basis. They say that's double last year's total. [inaudible 00:08:40] was pointing that out. I mean Grace, in a year's time, do you see it being a similar ratio in terms of OpenAI and ChatGPT's lead or can you see other players getting closer in terms of second, third, fourth?

Grace Harmon (08:52):

Well, I don't think there's any doubt about ChatGPT's market dominance. I would say going back to the number you had said, those figures were from March and it isn't that everything has changed drastically.

Marcus Johnson (09:01):

Good point.

Grace Harmon (09:02):

ChatGPT is still at the top, but since then, for one, we had the Made by Google event and we saw a lot of big changes for Gemini and a lot of advantages that they've added for on-device. And Marcus, you're exactly right, that the more ways that you can provide users to use that AI, the more likely someone is going to be to use it. I know that Anthropic expanded its Claude app globally a lot more this year, which is just expanding the number of ways that someone can use it. If you don't have to stop and pull out your laptop to be able to use a chatbot, it's going to be a lot easier, a lot quicker to use it. And then the number of partnerships, especially with smartphones, especially with devices that AI companies can get is a really big deal, boosting another boost for ChatGPT like you said, is Apple Intelligence.

Marcus Johnson (09:46):

Big part of this is just awareness and ChatGPT is benefiting hugely from brand awareness. 63% of Americans had heard of ChatGPT for one reason or another. That's compared to 40% who knew about Google's Gemini, 32% who knew about Microsoft's Co-Pilot or Samsung Galaxy AI. That's according to a September YouGov numbers.

Grace Harmon (10:12):

And ChatGPT is also just synonymous with Gen AI, I'd say.

Marcus Johnson (10:16):

Yeah, at this point, absolutely. Definitely got the first mover advantage. It must be very hard to unseat them. I mean this is kind of getting to what you were saying a little bit, Grace, in terms of people carrying their smartphones around and wanting AI infused into that device to help them with certain things. 60% of consumers now consider AI features important when choosing their next smartphone, and that includes 21% who say AI features are very important. That's also from some YouGov numbers from September. Grace, to you, what are the main ways folks are using Gen AI?

Grace Harmon (10:53):

In general, I know that looking back, especially at Jacob's former report, that image creation and then chatbots are two of the top ways. I do see video generation tools being something people are really interested in and they could take off. We just aren't seeing companies actually come out with the tools they've promised. OpenAI said that it was going to come out with its own video generation tool that would create 60 seconds of video in February or March, and we still haven't heard of peep. It's not available to the public. I think it's been used in some demos at film festivals, but that's it. Adobe came out with some video generation tools, but the capabilities are really limited. It's between 2 and 10 seconds.

(11:33):

So compared with image creation, with text creation where there really isn't limits with how detailed of a photo you can make, how long of a text you can create, I don't see how well video can take off until you can actually create something that's more than a clip for Instagram or an extension, background for a YouTube video. But yeah, text and image are still number one, and then coding capabilities just still aren't that reliable.

Marcus Johnson (12:01):

Right.

Jacob Bourne (12:02):

Yeah.

Marcus Johnson (12:02):

They seemed to be one or two photo and video tools, seem to be some of the most important AI powered smartphone features for US consumers. YouGov again said voice assistance up top 30% that were searching the web, taking actions and stuff, but then second, just behind that with 25% was photo and video tools. That's object removal, photo and video editing. Then right behind that in joint third with 22% was writing tools and advanced search as well.

Jacob Bourne (12:32):

Yeah, I mean I think the video generation when it's fully deployed is going to be massive. I think the two biggest limitations right now, one is just there's ethical concerns about deploying it due to the issue around deepfakes. The other issue is just that it's really expensive computationally and the power it takes to make a video generator publicly accessible is just very expensive. But in terms of the main ways that people are using it right now, I mean I think the breakdown here is how people are using it in their personal lives and how they're using it at work. I think content creation and just general search are big on both fronts. I think in terms of people's daily lives, I think using chatbots to get advice on navigating social situations is big, accessing healthcare information, and we're also seeing that these chatbots are really serving as companions for people, which I think has both positive and disturbing implications to it.

(13:29):

As far as the workplace usage goes, I mean it really spans a number of industries and use cases. I mean things like accounting, marketing campaigns, coding assistance for software development is big, though there's a lot of range in terms of how effective people think it is. Custom service chatbots, automated investing is getting big now. We did another podcast on that previously, and I think that the more technical use cases are also gaining ground to things like drug discovery, material discovery, cybersecurity, but then of course as the flip side of people using it to generate malware. The list goes on and the list is going to continue to expand as the technology improves and as people just figure out new ways to use it.

Marcus Johnson (14:15):

You mentioned people using Gen AI to talk to as companions. There was some data from Common Sense Media asking what teens are doing with Gen AI, and they were surveying 13 to 18 year olds here, and 18% said they were using it for personal advice, so that's nearly 1 in 5. However, that was the ninth main reason that they're using it, but still a pretty significant close to 20%. Number one was homework help. Just over half, 53%, and second, 42% was boredom. Right behind that translation, right behind that brainstorming with 38%.

Jacob Bourne (14:58):

I don't know if I would call boredom a use case, but it's interesting that that made the list.

Marcus Johnson (15:01):

These kids. Mainly it seems to be about saving people time and making life more convenient, some more YouGov data. 63% of folks agreed that AI will save them time. 44% said they want AI on their smartphone to make life easier, which I guess is what Apple is hoping, well, Apple Intelligence is hoping to do with this rollout.

(15:24):

Let's end by looking at what some of the most interesting ways people will be using Gen AI for in the very near future. One that jumped out to me, I thought this was really interesting. Adobe Analytics, the study of 7 in 10 people believe that using Gen AI to produce images of themselves wearing a product can boost their confidence when making a purchase. So it got me thinking, what if Gen AI could basically create an entire retail site where you are wearing every single product? What does that do?

Jacob Bourne (15:52):

The virtual trial.

Marcus Johnson (15:53):

[inaudible 00:15:53]. Exactly, yeah. What else comes to mind when you think of what people might be doing with Gen AI in the very, very near future?

Grace Harmon (16:00):

Well, I'd first like to say with startups, there are definitely companies that are providing services for retailers where you can, maybe not with a catalog of clothes as large as H&M or Fashion Nova or something like that, but there's definitely services for you to offer full try-on services to your customers.

Marcus Johnson (16:17):

Yeah, this is so fascinating.

Jacob Bourne (16:19):

There are three key areas I think to look at. One is just Gen AI being used for scientific discovery, including designing experiments, coming up with scientific hypotheses. I think it's going to increasingly be used in that way. There's also this notion of swarm AI agents, so this notion of [inaudible 00:16:39] AI, but you have a network of them that's working behind the scenes to get things done, not just one, so they're kind of working together, collaborating with each other. And then the third is this fusion of digital twins and Gen AI basically taking existing simulation AI and adding a semantic component to it, I think is going to end up making our models of complex systems, whether it's economic systems or the Earth's climate or what have you, or manufacturing plants. Just make it more realistic.

Grace Harmon (17:11):

I think there's also this delicate balance where the dream with AI application in a lot of instances is that you can ask it to do something really complex and not have to do anything, but maybe there's a better application where you are giving it these enormous tasks, but keeping that oversight on. I just don't see at any point in the really near feature as having AI where it doesn't need any oversight. You mentioning with OpenAI's Swarm framework that just came out. When I was screening through the code, there's part of the code details that there will be that the AI can fire itself, it can hire new AI agents, but also that it has its own internal oversight board so it monitors its own [inaudible 00:17:56] concerns. So there's the risk of creating an echo chamber in there, and I just don't know... That doesn't work to me.

Jacob Bourne (18:03):

And Grace, I think it's interesting because these are some of the fears that some AI researchers voiced when ChatGPT first came out, and it's just fascinating to see how fast we've gotten to the point where these things are now being developed and are soon going to be with us. AI agents that oversee themselves, I mean, that's kind of a big red flag to I think a lot of people.

Grace Harmon (18:26):

Yeah. Well, I mean, OpenAI was careful to be like, this is not an OpenAI product. This is on GitHub.

Marcus Johnson (18:33):

Yeah, it does seem like Jacob, you mentioned it's difficult to predict what's going to happen in the very, very near future. It does seem like we want it to be able to do this and be able to do that, but most of the time it's people using technology to do old things in new and slightly more convenient ways, and that Adobe survey in terms of what people are expecting from Gen AI in the retail world, it seems to speak to exactly that. People were most excited about the potential of Gen AI to help them filter products based on their needs, 40% of people. Then it was design a custom product, right behind that, summarize product reviews, chat-based customer service, so things that they're already doing, but can you help me do it a little bit more efficiently?

(19:16):

Speaking of Gen AI helping out in the science realm, let's go to that for our fourth quarter of today's show. Today in other news. Just one story. AI gets its Nobel moment. Story one, as I said, AI gets its Nobel moment. Alison Snyder of Axio. She explains that in one instance, British-Canadian Geoffrey Hinton and American John Hopfield were awarded the Nobel Prize in physics for their work on AI in the 70s and 80s. That showed us a new way to use computers through artificial neural networks. Ms. Snyder points out that the Nobel Prize is often awarded for research done decades ago after its impacts can be clearly assessed as having the greatest benefit to humankind.

(20:00):

In another instance, the prize in chemistry went to Google DeepMind CEO Demis Hassabis, DeepMind director John Jumper and University of Washington professor David Baker for developing an AI system that cracked one of biology's toughest problems predicting the structure of a protein. Jacob, we'll go to you first. What was your initial reaction to this news?

Jacob Bourne (20:19):

Yeah, I think there's two main takeaways here. First, I think this prize, especially the one given to DeepMind for its AlphaFold AI that can predict protein structure is kind of a nod to AI itself, not just the people who built it. It's kind of indirectly saying, kind of awarding the capabilities of this particular model. The other big takeaway I think, is that Geoffrey Hinton got this Nobel. Now this is someone who used to work for Google, who resigned so that he could warn the world about a potentially super-intelligent AI that could be catastrophic, do catastrophic damage to humanity, and I think there's a lot of people out there who think that's a silly notion, but I think by giving him a Nobel is very intentional in terms of putting a spotlight on that concern and giving it some credence.

Marcus Johnson (21:11):

Yeah. Grace, how about for you?

Grace Harmon (21:11):

Well, I was also interested, I guess with the latter prize about how they were applying AI, which was really careful and they were really smart about it. They weren't using AI to make the hypothesis. They weren't asking AI to create the question of what they were trying to solve. They were using it to design better experiments based on what the AI was able to find. So I think it was just really smart human work off of what the AI was able to do, so it seemed kind of about the human ingenuity in knowing what to use the AI for and what it can do, and then also knowing what humans can do better. I'd also say, David Baker, go Huskies. I'm a University of Washington alum.

Marcus Johnson (21:54):

Oh, hello. Well, that folks is what we have time for today. Thank you so much to my guests. Of course, thank you to Jacob.

Jacob Bourne (22:01):

All right, thanks for having me, Marcus.

Marcus Johnson (22:03):

Yes, sir. Thank you to Grace.

Jacob Bourne (22:04):

Thank you for having me.

Marcus Johnson (22:05):

Yes, indeed. Thank you to Victoria. She edits the show. Stuart runs the team. Sophie does our social media. And thanks to everyone for listening in. We hope to see you tomorrow for the Behind the Numbers Weekly listen. That is of course, an eMarketer video podcast made possible by TikTok that you can watch on Spotify and YouTube or YouTube, or you could of course listen to it the usual way.