Episode Transcript:
Marcus Johnson:
Hey gang, it's Thursday, April 6th. Gadjo, Jacob and listeners, welcome to the Behind the Numbers Daily and eMarketer Podcast made possible by Cint. I'm Marcus. Today I'm joined by two folks. Let's meet them. We start with one of our senior analysts on the connectivity and tech briefings based out of New York, it's Gadjo Sevilla.
Gadjo Sevilla:
Hi everybody.
Marcus Johnson:
Hey fella. We're also joined by someone else on that very team, connectivity and tech briefing based out of California, is Jacob Bourne.
Jacob Bourne:
Hey Marcus. Hey Gadjo.
Marcus Johnson:
Hey, fella. So today's fact is insane. This might be the craziest facts. We've done like a thousand episodes, this might be the craziest facts yet. Oh man, America, so disappointed. Okay, here we go. People used to mail their children.
Jacob Bourne:
What?
Marcus Johnson:
That's right. You heard me correctly. According to Danny Lewis of the Smithsonian magazine, when the post office's parcel post officially began January 1st, 1913, the new service suddenly allowed millions of Americans to send and receive things. And there were some unintended consequences as some parents tried to send their children through the mail. So just a few weeks after parcel post began and Ohio couple, come on, Ohio, you're not helping, named Jesse and Mathilda Beagle mailed their eight-month-old son James to his grandmother who lived just a few miles away in Batavia. According to the US Postal Service historian Jenny Lynch, baby James was just shy of the 11 pound weight limit for packages sent via parcel post. And his delivery cost his parents 15 cents in postage, although they were kind enough, to ensure him for $50.
This quirky story soon made newspapers, the article reads, and for the next several years, similar stories would occasionally surface as other parents followed suit.
Gadjo Sevilla:
That's crazy.
Marcus Johnson:
This is horrifying. I love that people read this story in the papers and instead of thinking to themselves, what the hell, instead they thought, honey is genius. Get a box. Yeah. We'll send them would
Jacob Bourne:
Definitely be illegal by today's standards.
Marcus Johnson:
I should hope so. Oh my goodness America, get it together. Anyway, today's real topic ChatGPT-4 is here. So what?
So in today's episode, first in the lead, we'll cover the arrival of ChatGPT-4, and then for in other news, we'll discuss 3D holoportation, the hell does that mean, and digital avatars for brands. But we start gents with ChatGPT-4. We covered it quickly in an episode a few weeks ago when we were talking to Jasmine Emberg, who's one of our principal analysts covering social and to yourself, Jacob, when we were talking about social and how Generative AI and AI in general will be fused together with the social media world and ChatGPT-4 just launched and it was hot off the press, so we quickly talked about it, but we wanted to dedicate an episode to talking about it exclusively. And so open AI, they've already announced their ChatGPT successor is called ChatGPT-4.
So what's new? ChatGPT-4, it 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 create code for a functional for a working site in seconds, assuring [inaudible 00:04:27] of Recode was making that point. 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 from a few months ago scored in the bottom 10%. It can supposedly answer complicated tax queries, process up to eight times as many words as ChatGPT. So initially it's available to ChatGPT Plus subscribers. They paid 20 bucks a month for premium access, but ChatGPT-4 is already here. Jacob, I'll start with you. Initial reactions.
Jacob Bourne:
Well, I think from the last time we spoke, I think the biggest thing to know from a commercial standpoint is that GPT-4 now has a plugin that can connect it to the internet, which means that consumers can use it to book flights on Expedia hotels, shop for groceries on Instacart. So I think this is really going to be a game changer for automated e-commerce, but the model is impressive. I think that's a general consensus. It's certainly a watershed moment for Generative AI overall. And I think one psychologist tested the model and showed that it has a verbal IQ of 155, which is off the charts.
Marcus Johnson:
Wow.
Jacob Bourne:
Yeah, and I think the other thing to know in terms of the big picture look at GPT-4 is that it's really causing a commotion among AI experts who have been really surprised by its capabilities, surprised at how fast Open AI is advancing this technology. And even Microsoft recently published a paper saying that GPT-4 has sparks of AGI or Artificial General Intelligence, which of course is the holy grail of AI research. It would be, if fully achieved, would put AI on par with human intelligence. And so that's creating a stir because people, including people in the AI field, are alarmed by how fast it's advancing and are really concerned that society's not ready for it.
Marcus Johnson:
I mean, you mentioned the IQ of 155, did you say? 150?
Jacob Bourne:
135. Yep.
Marcus Johnson:
Einstein had an IQ of 160, which puts you in the top 0.1 whatever percent of intelligent people in the world. I think Stephen Hawking was also at 160, so that's remarkable.
Jacob Bourne:
Just to be clear, it's a verbal IQ store, so it's not in [inaudible 00:06:44]-
Marcus Johnson:
Right. Right.
Jacob Bourne:
But yeah.
Marcus Johnson:
Yes, that's a good point. You mentioned the plugin and people can use this to book flights, things like that. We saw when ChatGPT came out a few months ago that it reached a million users in just five days. Do you see ChatGPT-4, do you see it being used by millions and millions and millions of people, 20, 30, 40% of the population quite quickly? Or is this something still that's being talked about, but it's early adopters only for now?
Jacob Bourne:
Yeah, that's a good question and I'm glad that you clarified that. This upgrade is actually ChatGPT, which is powered of course by GPT-4. It's a good clarification. I think that at this point, the plugin is available for pay users, so it's not a free version that you can use. It would take some getting used to, I think for a mass adoption for people to really be comfortable with ChatGPT booking their vacation. I think people who've already subscribed, I think definitely will be wanting to try that out for sure.
Marcus Johnson:
Okay, gotcha. Initial reactions to GPT-4.
Gadjo Sevilla:
So I think first of all, it's become more coherent version. They're saying that it's a tool that can allow you to have free flowing conversations. You can interact in various languages with the chatbot. So there are some educational implications. They're actually giving examples of what it can do. At the same time, they're saying the factual responses are 40% more likely to be factual, but they're also cautioning that it's more button buttoned down. So there is disallowed content. So there are certain topics that it won't cover or that it won't give you the kind of answers that the previous versions would give you, which could lead to some problems, some rabbit holes there. So I think the biggest thing for me though, is the timing for the release. Google came out with Bard and all of a sudden if anyone was thinking is an arms race an escalation, this clearly shows that there is.
Marcus Johnson:
Yeah.
Gadjo Sevilla:
Yeah. And I wouldn't be surprised if Open AI was just sitting on this and just waiting for the right time to introduce it. So there you go.
Marcus Johnson:
Yeah. One of the things jumped out to me as well for in terms of its capabilities, Mateo Wong of the Atlantic, noting it can write a 60,000 word book from single prompts. So some of the capabilities is just remarkable. Again, it's not perfect by any stretch of the imagination, but some of the things it's starting to be able to do are starting to really get people's attention. You mentioned the arms race, and so over 1000 tech leaders and researchers, including Tesla and Twitter's, Elon Musk and Apple co-founder Steve Wozniak, have urged AI labs to pause developments of the most advanced systems for at least six months warning in an open letter that AI tools present profound risks to society and humanity.
Cade Metz and Gregory Schmidt in New York Times were noting the letter claims that AI developers are locked in and out of control race to develop and deploy evermore powerful digital minds that no one, not even their creators can understand, predict, or reliably control. The letter says the pause would provide time to introduce shared safety tools for AI systems. Gadjo, I'll start with you. What do you think happens as folks call for a halt on the training of powerful AI systems? Do you think there will be one?
Gadjo Sevilla:
Might be some delays if ever, but I don't think you can stop the course of progress. I mean, they do have valid concerns, which I think everyone is thinking, and it's not the first time we've heard a warning saying this is dangerous, it's too quick. But at the same time, we need to realize that in technology, there are companies that have the advantage and there are others that may be six months behind. So I guess the question is who's asking and why? At the same time, it's only a matter of time before the government echoes those calls. We were already hearing the FDC start to give guidelines of what you can and cannot do. Although their plate is pretty full right now, but at any given time-
Marcus Johnson:
Yes. To say the least.
Gadjo Sevilla:
At any given time, they could shift their focus and say, "We need to regulate these things, this AI a lot more stringently", whether or not they have the personnel to manage, that is another story though.
Marcus Johnson:
Yeah. I mean, there's some in the works. In 2021, the European Union, so across the pond, proposed a law designed to regulate AI that might create harm, including facial recognition systems. It could pass this year and would require companies to conduct risk assessments of AI to figure out how their applications could affect health, safety and individual rights. There are a number of proposals for regulations of technology put forward by the US, UK, and EU. The UK has actually ruled out a dedicated regulator for AI. Jacob, there's a lot at stake here. BBC asks cohort citing Stuart Russell, computer science professor at the University of California, Berkeley, and a signatory to the letter saying, "AI systems pose significant risk to democracy through weaponized disinformation to employment through displacement of human skills and to education through plagiarism and demotivation." It seems like the stakes are incredibly high. What do you make of this call to slow things down and the potential for more regulation?
Jacob Bourne:
Yeah. Well, I think the stakes are really high. I don't think a petition letter is going to halt a model training at this point, though.
Marcus Johnson:
Interesting.
Jacob Bourne:
It would really take government intervention and regulators might look at that letter and pay attention to it. There was certainly a lot of controversy around the letter in terms of some fake names originally on there. Certainly some rival companies, CEOs from rival companies were on there. But there were a lot of people in the AI field who signed the letter who are genuinely concerned, and regulators might look at that. I think in the absence of regulation, we're going to continue to see these companies push the needle.
These companies that are competing with each other and already we see Open AI is talking about potentially releasing a GPT-5 by the end of the year, and they're calling the GPT-5 A potential AGI. I'm a bit skeptical of that, but certainly I think the thing to know is that things are accelerating, and depending on how fast they move this year, I think governments might be forced to step in. I think the problem right now is that they don't quite know what to do, and so I think they're sorting that out.
Marcus Johnson:
By AGI, artificial general intelligence?
Jacob Bourne:
That's correct, yes.
Marcus Johnson:
How would that be different?
Jacob Bourne:
Well, Artificial General Intelligence refers to AI reaching the point where it actually surpasses human intelligence, so it would be smarter than us. So it's a pretty high bar in the past. Wow. But yeah, prior to ChatGPT in this whole commercial generative of AI, people were talking about decades to even a hundred years before we would reach AGI, but now with the moves that Open AI is making, Microsoft, Google, Deep Mind, now that trajectory looks like that threshold might be passed sooner rather than later.
Marcus Johnson:
So that's the goal of the touring test is, because the tour test being when you test to see whether-
Jacob Bourne:
Yeah, the touring test is makes AI indistinguishable from a human. I think it's a similar concept, but this is basically say AGI by definition would be smarter than a human.
Marcus Johnson:
Got it. Okay. Yeah. So folks, the touring test is a test to see whether computers, the output can sound like a human, and the closes to that you get the better you score on the touring test. Wow. Okay. So fascinating. So going back to regulation for a second. So Margherita Stancati and Sam Schechner of The Wall Street Journal noted last week that Italy has temporarily banned ChatGPT over data's privacy concerns. Italy's privacy regulator has ordered a temporary ban on Open AI's ChatGPT, saying the AI chatbot improperly collected and stored information. Gadjo, I'll start with you. What are your thoughts on ChatGPT being banned in Italy?
Gadjo Sevilla:
I think that's just the beginning. The concern in most of Europe is, you know, GDPR and also how this technology could be mining data sets with personal information. Although Open AI did say that it's not sharing, it's training data in part, they don't use it to mine personal information, but to provide general answers. Them saying that and the actual use of the tool, you can type anyone's name and get a ton of information right now. So I think with the European laws right to be forgotten in GDPR, they will quickly put a stop to this sort of expansion if it's unregulated. So not surprising at all.
Marcus Johnson:
There's certainly consequences here. So the Italian regulator, so they opening inquiry into Open AI, giving them three weeks to prove it is trying to comply with EU privacy rules or it could risk being fined. And the max under what Gadjo was just saying, GDPR, which is the General Data Protection Regulation, and the EU, the max is 4% of the company's worldwide annual revenue or the equivalent of $22 million, whichever is higher. Jacob, where does your mind go when you think about this ban of ChatGPT in Italy?
Jacob Bourne:
Yeah, I mean, mine just says that lawmakers are in over their heads when it comes to Generative AI. It's a new technology. Even people in the field don't understand it, and there's a lot of issues with it, and some are easier to regulate than others. I think with Italy doing this, they're taking the easy route for regulation and basically they're taking an existing law and they're saying, 'Well, you violated this law in your data collection", and it's an effective ban by doing that. But really, we can be sure that Italy's issues with ChatGPT go a lot deeper than digital privacy, but passing new legislation to tackle Generative AI specifically is a huge undertaking. The EU is trying it right now with its AI Act, but even that, there's hundreds of tech companies right now lobbying the EU to limit the extent, the reach of the AI Act and basically make it so that you can regulate the end use of Generative AI, but you can't regulate the models, which of course, that would defeat the purpose, the whole outcry about GPT-4 and its capabilities and what could be next.
And so it really remains to be seen what the EU is going to do on that front.
Marcus Johnson:
Yeah. Italy using this law to put a stop to ChatGPT, at least for now, it feels like the US governments and prosecutors putting Al Capone behind bars for tax evasion. He definitely did other things, he certainly did worse things than trying to evade paying taxes, but this was a thing that managed to get him. It was not the first time a chatbot's been banned in Italy. The regulator banning Replica.ai in February, and they're saying there's concern for the kids as well. Italy concerned there's no age verification, so children under 13 can use the chatbot exposing them to unsuitable responses. The BBC noting that Bard, Google's rival AI chatbot is only available to specific users over the age of 18 because of those same concerns. That's it for the lead. It's time now for the halftime report.
Gadjo, I'll start with you. Your quick takeaway from the first half, please.
Gadjo Sevilla:
So the quick takeaway from the first half is basically there's an acceleration in adoption and innovation of Generative AI. And clearly some companies do control that flow of information. What that means is there will be pushback, whether from governments or from rival companies or just from even people in the field of AI who are worried about the ethics. And I think it's a conversation we're going to have to continue. And at the same time, there's always this feeling that we're trying to catch up to something that is evolving before our eyes. And I think that's clearly manifested with GPT-4s arrival and all the implications that you can now plug it into bigger ecosystems. And so I think we'll just continue to see that and continue to learn from it.
Marcus Johnson:
Yeah, Jacob?
Jacob Bourne:
Yeah, I think that the biggest thing to know is that we've really entered unchartered territory with the dawn of GPT-4 in terms of the Generative AI landscape. The tech is really advancing much faster than expected, and some of the leading AI experts are really floored by it. Regulators have a long way to go to catch up. So I think with Open AI, looking at GPT-4.5 and five, and even the internal conflict that's going on within Open AI, I think it's going to be a wild year ahead on the general AI front.
Marcus Johnson:
Well, we've got a ton of ChatGPT and Generative AI coverage and how it will affect advertising, the creator economy, retail, et cetera. You can head to insiderintelligence.com for all of that. One such report is called ChatGPT and Generative AI. Our Guide to 2023's Most Talked-About Technology and Where It'll Have the Biggest Impact. That was written by analysts, Yoram Wurmser and Jess Lis. But there are plenty of other reports on Generative AI on insiderintelligence.com, so head there if you're a pro subscriber. Time now for the second half of the show today in other news. What the hell is 3D holoportation? And does every brand need a digital avatar?
Story one. 3D holoportation could be around the corner. Jennifer A. Kingson of Axios writes the proto, formerly known as portal, uses cameras where you are and you can project or beam that image of you in 3D directly into a vending machine size box on the other person's end. Think 3D video calls, if you will, so proto have a lower budget, two foot tall version for $2,000. That's version of the size of the box that you would get projected into on the other side of the core, or a full seven-foot life-size version for a hundred thousand. So you either appear in a tiny box on someone's desk or in a full phone booth sized box, and your image gets projected in 3D. Use cases, colleges using them to beam expert teachers from around the world into classrooms or military families being able to see and talk to a more realistic version of overseas loved ones. But Jacob, the likelihood of this technology taking off is what?
Jacob Bourne:
Yeah, I mean, this represents a piecemeal release of these virtual technologies that will eventually form the metaverse. I think certainly this could have really strong applications for what you just mentioned, education, some people that need to connect with loved ones at a distance. But I think we're unlikely to see mass consumer adoption in the near term. Right now, the overall demand for virtual technologies is pretty low among consumers, largely due just to inflation. I don't foresee people, at least this year, wanting to shell out money for something that they probably don't need in the future if the economy recovers, if and when potentially that this might be something that people will buy.
Marcus Johnson:
Interesting. Story two, digital avatars may prove useful for businesses, and not just in the metaverse writes Ben Sherry of Inc. He notes that AI powered virtual characters can have natural conversations and possess unique personalities. So the example it gives is, imagine you run a staffing company, instead of getting clients to fill out forms describing the kind of employees they're looking for, the candidates could simply talk to a virtual consultant, which could recommend new hires based on the conversations that you have with it. You could go even further and add a charming personality to your virtual consultant, further enhancing the customer experience. But Gadjo, will companies actually invest in these digital avatars?
Gadjo Sevilla:
I think some would. There are a lot of applications where this would be helpful, just helping with customer support, putting a name in a face, even if it's virtual. It's a natural extension of voice chatbots that we interact with already over the phone, online. So I think it's just putting together all these technologies to push the information and the brand that these companies are trying to promote, I guess, in a more widespread and easily accessible way.
Marcus Johnson:
Yeah. Some interesting use cases in World AI, a group developing a character creation system says you could develop a digital avatar to rehearse situations such as terminating an employee. So you could role play, you could use it to practice, or they could use a digital person to host corporate training exercises. That's all we've got time for. Thank you so much to my guests. Thank you to Gadjo.
Gadjo Sevilla:
Thank you.
Marcus Johnson:
Thank you to Jacob.
Jacob Bourne:
Thanks, Marcus. Thanks Gadjo.
Marcus Johnson:
And thank you to Victoria who edits the show. James, for copy editing. Stuart, who runs the team. Thanks to everyone listening in. If you have a few seconds and could leave us a review, we would massively, massively appreciate it. Those reviews help us to keep this podcast going. We'll see you tomorrow hopefully for the Behind The Numbers Weekly Listen, an eMarketer Podcast made possible by Cint.
%20scale(0.100000,-0.100000)%22%0A%20%20fill=%22%23000000%22%20stroke=%22none%22%3E%0A%20%20%3C/g%3E%0A%3C/svg%3E%0A)
