30% of Young Americans Have Used ChatGPT at Work

0:00Today on the AI Breakdown, some new stats showing how ChatGPT at work is a breakout use case. Before that of the brief, Apple's new model for improving Siri and more than 200 musicians sign a letter asking for their rights to be protected. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.net for more information about our YouTube, our Discord, and our newsletter. Welcome back to the AI Breakdown Brief. All the AI headline news you need in around five minutes, although today's is quite extensive, so we'll see what the length ends up being.

0:35We kick off today with a little bit more information about this new model that Apple has been teasing that they're calling Realm. It seems like it's a key part of their strategy to make a new Siri, which is desperately and by far and away the thing that the AI community is most looking for out of any sort of future Apple AI announcements. Realm stands for Reference Resolution as Language Modeling. And as Tom's Guide puts it, it's designed to run on a phone and make voice assistants like Siri smarter by helping it understand context and ambiguous references. So part of the thing that Realm does is that it can take information from the screen while the person is using the assistant.

1:12Again, from Tom's Guide, it's a visual model that reconstructs the screen and labels each on-screen entity and its location. This creates a text-based representation of the visual layout, which can be passed on to the voice assistant to provide it context clues for users' requests. The authors of the paper said that this new approach, despite the model being smaller, has it outperforming even bigger models like GPT-4 on certain queries. The authors of the paper wrote, We especially wish to highlight the gains on on-screen datasets, and find that our model with the texture encoding approach is able to perform almost as well as GPT-4 despite the latter being provided with screenshots.

1:45Now, of course, one of the things that we're waiting to figure out is exactly what the breakdown of Apple native AI versus Apple partnered AI with companies like Google will be. We've heard reports that Apple is in conversations with Google to put Gemini in the new iOS operating system, and so where exactly those lines will be drawn remains a little bit to be figured out. Next up, the latest open letter to rock the AI world, literally in this case, more than 200 musicians including Billie Eilish, Nicki Minaj, Stevie Wonder, Jay Balvin, along with the estates of Frank Sinatra and Bob Marley have all signed an open letter pleading with tech firms not to develop AI tools that could replace them.

2:21The letter came from a group called the Artist Rights Alliance. It reads, We the undersigned members of the artist and songwriting communities call on AI developers, technology companies, platforms, and digital music services to cease the use of AI to infringe upon and devalue the rights of human artists. They say the assault on human creativity must be stopped. We must protect against the predatory use of AI to steal professional artists' voices and likenesses, violate creators' rights, and destroy the music ecosystem. There are tons of artists that you will know and have heard of on this list.

2:49It is definitely not an insignificant group of people. But what's interesting is that the media is, shockingly, not reporting as nuanced as the letter actually is. Let's get more specific. The letter reads, Make no mistake, we believe that when used responsibly, AI has enormous potential to advance human creativity and in a manner that enables the development and growth of new and exciting experiences for music fans everywhere. Unfortunately, some platforms and developers are employing AI to sabotage creativity and undermine artists, songwriters, musicians, and rights holders. When used irresponsibly, AI poses enormous threats to our ability to protect our privacy, our identities, our music, and our livelihoods.

3:26Some of the biggest and most powerful companies are, without permission, using our work to train AI models. These efforts are directly aimed at replacing the work of human artists with massive quantities of AI-created sounds and images that substantially dilute the royalty pools that are paid out to artists. For many working musicians, artists, and songwriters who are just trying to make ends meet, this would be catastrophic. Unchecked, AI will set in motion a race to the bottom that will degrade the value of our work and prevent us from being fairly compensated for it. So basically, this is the standard complaint from creatives.

3:55They don't want AI trained against their work that is then used to replace them. Ultimately, as we've discussed so many times on this show, this is a conversation that is going to bridge from, on the one hand, legal and will be determined in courts when it comes to questions of fair use and copyright when it comes to training, but it's also a societal question and a market question. It's inevitable that there are going to be a huge amount of AI generated music in the future. The impact that that will have on existing musicians and the way in which that comes to bear remains to be seen. Now, heading into a different dimension of AI risk and safety, the US and the UK have signed a new agreement, the first big bilateral treaty, on information sharing around AI safety testing.

4:36The deal was signed by the UK's science minister, Michelle Donilon, and the US Commerce Secretary, Gina Raimondo. The Financial Times writes, The agreement will specifically enable the UK's new AI safety institute, set up in November, and its U.S. equivalent, which is yet to begin its work, to exchange expertise through secondments of researchers from both countries. The institutes will also work together on how to independently evaluate private AI models built by the likes of OpenAI and Google. To me, the substance of the deal is less important than the fact that it exists. Commerce Secretary Raimondo said that AI was the, quote, defining technology of our generation.

5:08She continued, Our partnership makes clear that we aren't running away from these concerns, we're running at them. Speaking of running at AI full throttle, the Wall Street Journal today reports on how business schools in America are, as they put it, going all in on AI. They point to examples like American University's Kogod School of Business, which has created or adapted 20 classes from forensic accounting to marketing all around AI. And interestingly, one of the things that you're starting to see here is something that I wish would happen more broadly in education, which is educators viewing AI not as a cheating tool that they can somehow beat, but as a new reality and a new tool that students are going to use and that they thus have to adapt to.

5:47They cover, for example, the story of Robert Bray, writing, When Robert Bray, who teaches operations management at Northwestern's Kellogg School of Management, realized that ChatGPT could answer nearly every question in the textbook he uses for his data analytics course, he updated the syllabus. Last year, he started to focus on teaching coding using large language models, and enrollment jumped from 21 MBA students to 55. Before, engineers had an edge against business graduates because of their technical expertise, but now MBAs can use AI to compete in that zone. Bray encourages his students to offload as much work as possible to AI, treating it like a really proficient intern.

6:20Like I said, I think that this is the right attitude in general for educators to treat this like the force that it is, something that is not going to be put back in the bottle, and something that enables us and requires us to design completely different types of educational experiences. Now, one area where AI is not being as welcomed is in the courtroom. A Washington state trial judge has banned the use of AI-enhanced video as evidence in a trial of a man accused of killing three people. Wrote Judge Leroy McCullough, This court finds that admission of this AI-enhanced evidence would lead to a confusion of the issues and a muddling of eyewitness testimony and could lead to a time-consuming trial within a trial about the non-peer-reviewable process used by the AI mode.

7:00Basically, in this situation, a shooting that killed three people in Des Moines was caught on a cell phone. Obviously, that means that the video quality is somewhat low. Because of that, lawyers for the defendant wanted to use AI to enhance the video, but the prosecutors argued that AI enhancement software is not necessarily meant to reflect the truth, but to just make video more visually appealing. This might be the first ruling of its kind, but it would certainly not be the last time this issue comes up. Speaking of issues and resolution, you may remember recently when a set of podcast hosts trained an AI algorithm on George Carlin's work to produce a new George Carlin special that they called, I'm Glad I'm Dead.

7:38This arose the ire of the Carlin estate, and now the parties have apparently settled, with effectively the podcast hosts agreeing to take the content down and never do anything with George Carlin again. Now one thing to note is that these folks weren't trying to trick anyone into thinking it was actually George Carlin. Indeed, the presentation began, Hello, my name is Dudzi, and I'm a comedy AI. I just want to let you know very clearly that what you're about to hear is not George Carlin. It's my impression of George Carlin that I developed in the exact same way a human impressionist would. Yet that was not enough for Carlin's family, and so they filed suit in January.

8:11Lastly today, the talent battle in AI continues to be intense. Fortune covered recent departures from Meta AI, with three high-profile engineers leaving not for other AI companies but to start their own things. And the former head of developer relations for OpenAI, Logan Kilpatrick, has just announced that he's joined Google to lead product for their AI studio. I think people naturally feel that the flow from big companies like Google to smaller startups like OpenAI, although of course OpenAI is no longer a small company, is the more natural state. So when things go the other direction, people take notice.

8:43I think probably more than any other big trend, the reality is just that well-positioned people in this AI space are going to have lots and lots of job opportunities. And so hey, not a bad thing to learn about. But that will do it for this extended brief. Next up, the main AI breakdown. Hello friends, quick note before we get into the main part of the episode. If you've been listening to the show for the last few months, you know we have been running an education beta. This is a new approach to AI learning that is hyper practical, focused on getting you actually using AI tools in minutes, not hours, and certainly not days, based around video tutorials and companion challenges and projects that have step-by-step instructions that make it really easy to try out new AI platforms.

9:25And all of this is now culminating in the launch of a new platform, which we're hoping is the most practical and useful way to learn AI that anyone has yet created. If you want to be the first to know when that launches, go to bsuper.ai, that's bsuper.ai, and sign up for the waitlist. I cannot wait to tell you more about it. Welcome back to the AI Breakdown. Today we are looking at the latest research around American usage of chat GPT. And fascinatingly, I have some very different conclusions or at least very different interpretations in the way the Pew Research Center seems to be interpreting some of their results.

10:05There are two parts of this survey. Broadly speaking, it's just about chat CPT usage. But then specifically, it's about usage at work. That's one part. And trust when it comes to election-related information. That's another part. The first part I want to talk about is the work piece. So, first of all, by way of a general update, last July, 18 % of American adults said that they had used ChatGPT. In February, that was up to 23 % of U.S. adults. However, those numbers are significantly changed when you look generation by generation. When it comes to adults under 30 years old, so between 18 and 29, 43 % of them have now used ChatGPT.

10:42That's up 10 percentage points since last summer's 33%, or one-third. The percentage of adults 30 to 49 that have used ChatGPT has gone from 21 to 27 percent. Between 50 and 64, the number goes from 13 to 17 percent. And for the 65 plus crowd, it moved slightly from 4 percent last July to 6 percent in February. Perhaps unsurprisingly, the more educated people are, the more likely to have used ChatGPT they are. Only 12 percent of those with a high school diploma or less have tried ChatGPT. 23 percent of those who have some college experience have tried it. 29 percent of those with a bachelor's degree have tried it.

11:14and 37 % of those who have a postgraduate or other advanced degree have tried it, which is up 8 points since last July as well. Now what's most notable to me are the statistics around work. Pew Research asked three different reasons that someone might have tried ChatGPT, for entertainment, to learn something new, or for tasks at work. The percentage of Americans who say they've used ChatGPT for entertainment was 11 % last March, 15 % last July, and 17 % in February. So an increase but a small one. To learn something new was similar. The percentage of American adults who had used ChatGPT to learn something new was at 8 % last March, 14 % in July, and again just up a little bit to 17 % in February.

11:54However, when it comes to Americans who have used ChatGPT for work, the number has risen by double digits. Back last March, only 8 % of Americans had used ChatGPT for work. In February's survey, that was up 20%. The numbers are even more stark when again you break it generationally. For those ages 18 to 29, the percentage of those who said they had used ChatGPT for work jumped from 12 % last March to 31 % this February, nearly a tripling. For the group that's 30 to 49, it was up from 8 % last March to 21%. And even the 50-plus demographic saw a significant increase from just 4 % last March to 10 % last February.

12:31Now, this clearly indicates that work is the breakout use case for American adults when it comes to ChatGPT. which if you've ever used ChatGPT for anything at work, probably won't be that surprising. Now, the numbers also do look a little bit different generationally when it comes to learning something new and for entertainment. For example, if you break down the to learn something new numbers by generation, that 18 to 29 set jumps from 16 % who had used ChatGPT to learn something new last March to almost double at 31 % this February. When it comes to use at work, the numbers also follow educational lines.

13:03For those with a high school diploma or less, only 8 % had used ChatGPT at work. For those with some college, only 19 % had used it. For those with a bachelor's degree only, 25 % or a quarter had used it. And for those with a postgraduate degree, that was up to a third who had used ChatGPT for work. Now, interestingly, the percentage is statistically unchanged for workers with a high school diploma or less from a year ago. But the shares for each of the some college, bachelor's degree only, and postgraduate had tripled since March of last year. Professor Ethan Mollick, who just released his book about generative AI, had two interesting things to say about these numbers.

13:35The first, he writes, This is a really fast adoption curve for a technology that is, for most people, 16 months old. Just not a lot of support for the idea that AI use is topping out. The other thing that he points out, which is something that certainly we've seen over here at the AI Breakdown, he writes, The biggest AI opportunity companies are missing is that their employees are rapidly adopting AI and figuring out how to use it for work and not telling leadership. Organizations that have unclear AI policies are not stopping people from using AI, they stop them from sharing uses. for as much as corporate leaders and enterprise leaders and division leaders are being pressured to have AI strategies.

14:12The reality is that this adoption is happening from the ground up in a bottoms up way. And that is what I think makes it so clearly important and sustainable. Now, let's briefly touch on the one other part of this survey. They also asked survey respondents whether they trusted information about the election that comes from ChatGPT. Just under 40 % said that they don't have too much trust or none at all. Only 2 % have a great deal of trust, and only 10 % have some trust, 15 % were not sure, and 34 % have not heard of ChatGPT. In some ways, it felt like Pew was presenting this as a negative thing, that ChatGPT wasn't trustworthy.

14:45I think, on the other hand, that it's actually a good thing. My base case for why misinformation is going to be less disruptive in the short term, at least, than some think, is that I think we're going to adapt much more rapidly than many assume to a world in which LLMs are pumping out information. I think our trust hackles are going to be raised and we're going to be more mistrustful overall, which might bring its very own consequences, but I don't think being tricked by AI is necessarily going to be the thing that ends up being the biggest challenge. The fact that the vast majority of people don't trust what's coming out of ChatGPT when it comes to the election suggests that those walls are up right now.

15:19And while of course ChatGPT wants to build trust as a reputable source of information, I think net net for society, it's better that people go in more skeptical than blindly trusting when it comes to these technologies. Anyway, really fascinating stuff. Always great to get a report card on the state of adoption and usage. And I'll be interested to see whether these numbers surprise you either to the high end or the low end. For now, though, that is going to do it for this episode of the AI Breakdown. Until next time, peace.