The Next Phase of Generative AI Adoption

0:00Today on the AI Daily Brief, we are discussing the next phase in generative AI adoption. And before that, on the brief, OpenAI shuts down a set of Iranian chat GBT accounts. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes.

0:24Welcome back to the AI Daily Brief Headlines Edition, all the daily AI news you need in around five minutes. In our coverage of SB 1047, the controversial California AI regulatory bill that's been in the discussion recently, one of the big complaints from people who are opposed to the bill is that it prioritizes existential risk factors versus those that are more here and now. This is the basis, for example, of a lot of the current congressional Democrats from California saying that SB 1047 lacks an evidentiary basis. Whether you agree with that or not in terms of SB 1047, if you watch the news, it's definitely clear that there are challenges here and now with AI that we're starting to have to deal with.

1:04Case in point, OpenAI recently deactivated a ChatGPT account that was linked to an Iranian disinformation campaign. One of the big concerns that people have had is AI impacting the current presidential election cycle. And while this wasn't the first time that OpenAI has had to shut down an account from a foreign adversary, it does appear to be the first time that OpenAI had removed an account that was focused on disinformation around the U.S. election. Axios writes that OpenAI actually identified, removed, and banned an unspecified number of ChatGPT accounts this week that were being used to create content about the U.S.

1:39presidential elections. OpenAI writes, This week, we identified and took down a cluster of ChatGPT accounts that were generating content for a covert Iranian influence operation identified as Storm2035. We have banned these accounts from our services, and we continue to monitor any further attempts to violate our policies. The operation used ChatGPT to generate content focused on a number of topics, including commentary on candidates on both sides of the U.S. presidential election, which it then shared via social media accounts and website. That said, OpenAI said that the operation didn't appear to have achieved meaningful audience engagement, and that the majority of social media posts that they identified received few or no likes, shares, or comments.

2:14They referenced something called the Brookings Breakout Scale, which assesses the impact of covert intelligence operations on a scale from 1, which is the lowest, to 6, which is the highest, and considered this to be the low end of Category 2, which meant activity on multiple platforms but no evidence that real people picked up or widely shared their content. Storm2035 was using ChatGPT both to generate long-form articles as well as to generate shorter social media comments. The content that was created was published to both progressive and conservative news outlets, and the social media accounts that were created posed as both progressives and conservatives.

2:45And interestingly, OpenAI says, they intersperse their political content with comments about fashion and beauty, possibly to appear more authentic or in an attempt to build a following. Said Ben Nemo, principal investigator from OpenAI's intelligence and investigations team, we all need to stay alert but stay calm. There's a big difference between an influence operation posting online and actually becoming influential by reaching an audience. Meanwhile, that wasn't the story that was getting a lot of attention when it comes to AI and the elections. Former President Donald Trump took to his Truth social network and shared several screenshots of ex-posts showing women wearing Swifties for Trump t-shirts.

3:18Another screenshot was Taylor Swift, created to look like Uncle Sam, with the message, Taylor wants you to vote for Donald Trump. Trump himself captioned the post, I accept. And there have been approximately 100 ,000 different media outlets writing about this. Now to me, this actually brings up a question from Professor Ethan Malek, who wrote, The first predictable bad effect of AI, easy deepfakes, is already here thanks to open flux, and there has been remarkably little wide outcry, either in the press or among policymakers. I wonder if it hasn't sunk in yet or if it ends up being less disruptive than anticipated.

3:49He continues, It looks like we're all going to just live through the end of reliable images and see how it works out. The interesting question to me, which we'll just have to wait and see, is what actual impact things like Trump sharing those AI photos actually has. In that case, it strikes me that they're pretty obviously not real, and that it doesn't seem like there was an actual attempt here to convince anyone that they were? Does that have the impact of us just sort of assuming things are AI, so to not take anything too seriously? Or does it lull us into a false sense of security, where when a politician shares an image that's actually been created by AI, and attempts to convince us it's real, we actually still take it seriously?

4:26It's a pretty significant and fascinating sociological question that I'm not sure anyone's going to know until it actually happens. Meanwhile, over in AI on Wall Street, AMD is making a$4.9 billion AI acquisition in its attempt to catch up to NVIDIA. The company is planning to buy ZT Systems to bolster its ability to provide AI infrastructure. According to a release, AMD says the deal will bring about, quote, the next major step in AMD's AI strategy to deliver leadership AI training and inferencing solutions based on innovating across silicon software and systems. Writes MarketWatch, the company will be tapping into ZT Systems' expertise in both the design and optimization of cloud computing offerings.

5:04Said Mizuho analyst, the deal validates NVIDIA's integrated strategy of self-contained full-rack system hardware designs, and the deal could make investors quote a wee bit more confident about AMD's ability to hit Wall Street's target for$9-10 billion in AI GPU revenue next year. Meanwhile, the information is arguing that another AI-related stock in ARM is priced too highly. They write, since chip design firm ARM Holdings went public last September, excitement around generative AI has sent its stock soaring, to a high last month of$188, more than triple its IPO price. While the stock has since fallen back a bit, investors are still valuing Arm at a big premium to NVIDIA both on a multiple of future revenue and profits.

5:41The valuation gulf makes little sense. By way of example, they say that AI is a much smaller part of Arm's business than of NVIDIA's. They compare NVIDIA's 126 % revenue increase as composed to ARMS 21%. Ultimately, this author suggests ARMS stock looks severely overvalued trading at 33 times estimated next 12-month sales compared to 23 times for NVIDIA. They also point out that AMD is trading at just 8.5x. I share this only because I continue to think that Wall Street is in a potential repricing moment when it comes to AI stocks, which, as we will discuss again a little bit in the main part of the episode, I believe has exactly nothing to do with what AI means as a technology.

6:18Lastly today, yet another lawsuit around copyright infringement, with a group of authors suing Anthropic this time, alleging that it committed quote large-scale theft in training Claude on pirated copies of copyright books. At some point soon here, we're going to have to start declaring a moratorium on covering news of new suits in favor only of actual updates as these things work their way through the legal system. For now though, that's going to do it for today's AI Daily Brief Headlines edition. Up next, the main episode. Today's episode is brought to you by Venice. The leading AI companies store your entire conversation history and attach it to your identity forever.

6:52That's every question you ask, every answer you receive, every image you generate, every thought you share with the machine, it's all being spied on. If you trust all the company's hackers and NSA board members that will ever have access to your AI conversations, then rejoice, for you are well served. For the rest of us, Venice is an alternative. Venice is a powerful AI app for text, image, and code generation that respects you as a sovereign individual. and believes privacy and free speech are not only human rights, but necessary for civilizational advancement. Private, permissionless, and uncensored, you can try it for free without an account.

7:23AI Daily Brief listeners receive a 20 % discount on Venice Pro. Visit venice.ai slash nlw and enter the discount code nlwdailybrief. That's nlwdailybrief. All one word. Today's episode is brought to you by Superintelligent, which is, of course, our platform that helps you learn how to use AI tools, and perhaps even more importantly, gives you ideas on the best use cases that are actually going to help you achieve whatever it is you want to achieve. To recognize the end of summer and back to school slash back to work, we are running our best promotion ever. When you sign up for Superintelligent, between now and the end of August, using code SOBAC, your first month will be 100 % free.

8:06The platform features over 600 fun, highly practical AI tutorials that get you using AI fast and with an eye to actually transforming how you get things done. We've just launched Super for Teams, so if you have a group of people at your company that want to figure out how to use AI together, I highly suggest you check it out. But for those of you who are using Superintelligent as an individual, once again, if you sign up for Superintelligent between now and the end of the month using code SOBAC, you will get your first month 100 % free. go to besuper.ai and check it out today. You have heard it endlessly throughout this summer.

8:43AI is losing its hype. It's losing its luster. I have also hopefully explained a number of times why I think there is a massive, massive difference between the way that Wall Street prices AI companies and artificial intelligence in general. Seemingly no journalists are able to make this distinction, but here we are. Today though, instead of ragging on all of that, I'm going to focus on what might be coming next, because there's no doubt that we are in a transitional moment in terms of how people think about AI adoption. Particularly in firms who have quietly been experimenting this whole time, there is, I think, a sense of a shift or at least a desire for a shift.

9:20And interestingly, McKinsey really captured this in a recent survey. This came out on August 7th, and the blog post summing it up, they called Gen AI's next inflection point, from employee experimentation to organizational transformation. This informs a huge amount about how we're thinking about AI at superintelligent, and so today we're going to dig into some of these numbers and what I think the implications are. To set this up, I want to come back to a study which I frequently reference here on the show, which is the 2024 Work Trend Index from Microsoft and LinkedIn. This came from a survey of around 31 ,000 people across 31 countries.

9:53One of the really notable parts of this survey was the speed at which employees are adopting AI even though their bosses are getting hung up. The study found that 75 % of knowledge workers were using AI, and among those, 46 % had started in the last six months, while meanwhile, 79 % of leaders thought that AI was critical to remain competitive, but 59 % were worried about quantifying productivity, and 60 % said that their company lacked a vision and a plan to implement AI. The net result was that of those 75 % who were using AI, 78 % were just doing it on their own, not telling their bosses about it, bringing their own personal uses of AI to the office.

10:28The net impact of this is that right now, all of the benefits of AI are accruing to individual employees, not to companies. In other words, when AI is invisible, it's great if employees are doing things much faster than they might otherwise, but unless they are proactively deciding to use that extra time to move farther or faster with their work for the company, the company isn't necessarily realizing any of that benefit. This has created a prerogative for enterprises to figure out how to scale individual employee benefit from AI to broad organizational benefit. And so this sets us up for this McKinsey survey.

11:03The kickoff line is, as many employees adopt generative AI at work, companies struggle to follow suit. To capture value from current momentum, businesses must transform their processes, structures, and approach to talent. So as you can see, sounds familiar. Now again, especially because we've got all these sort of articles like this one from The Economist, artificial intelligence is losing hype, it's important to counteract this with the results of these surveys. After nearly two years of debate, McKinsey writes the verdict is in. Gen.ai is here to stay and its business potential is massive. But as I said, employees are far ahead of their organizations in using Gen.ai, and companies have been slow to adopt in ways that could realize Gen.ai's trillion-dollar opportunity.

11:40So let's talk about some of the study's findings. First of all, this study found even more usage than that Microsoft and LinkedIn survey. This had 91 % of employees using generative AI split between 21 % being heavy users and 70 % being light users. All of those folks, as well as the non-users, anticipate that generative AI will positively impact their work experience. Of those who are using it the most, 98 % say they believe it will positively impact their work experience. Among light users, 91%. And even among non-users, 80 % believe that Gen.AI will positively affect their work experience. In terms of what they think will be improved, communication, creativity, critical thinking, and decision-making and ability to collaborate all score highly.

12:20To take the light users as a benchmark, 81 % of light users think that Gen.AI will improve their communication, and 75 % think it will improve their creativity. The biggest gap between light users and heavy users was in attitudes around critical thinking and decision-making and the ability to collaborate. 65 % of light users thought that Gen.AI would positively impact their ability to collaborate, as opposed to 84 % of heavy users. 84 % of heavy users also thought that it would impact their ability to think critically and make decisions, while 67 % of light users thought the same. In terms of this idea that organizations are lagging behind, McKinsey points out that only 13 % of respondents' companies have implemented multiple use cases.

13:00Not surprisingly, the organizations that had implemented two or more use cases tended to have a higher concentration of those heavy users. Organizations that had multiple use cases, a group that McKinsey called early adopters, saw 49 % of their employees as light users and 43 % of their employees as heavy users. They write, the CIO of a global heavy industry company sees these trends at his own organization. Employees are experimenting with Gen.AI through publicly available and embedded tools, which is increasing curiosity and encouraging greater openness to experimentation. Yet he notes that there's no easy-to-prove business case for employee-driven adoption and the piecemeal implementation of use cases.

13:34And that leads McKinsey to what they believe is the next inflection point, moving from individual experimentation to strategic value capture. So what are their suggestions for how organizations can make this transition? They argue that there are three key steps. The first is reinvent domains by translating vision into value. The second is reimagine talent and skilling by putting people at the center. And the third is reinforce changes through formal and informal mechanisms that ensure continuous adaptation. So basically, the word salad of reinvent domains by translating vision into value is that companies should take a domain-based approach.

14:09Basically, let units like product development, marketing, and customer service think holistically about solutions and implementations and new workflows that work for their domain. Effectively, thinking about Gen.AI adoption at only the individual level is too small, but at the cross-organization level might be too big, whereas within a specific domain or department, that might be a better starting point. Now, when it comes to the second bucket, putting people at the center, The companies that are more adept at using AI right now, what McKinsey calls early adopters, also, quote, prioritize talent and the human side of gen AI more than other companies.

14:40Two-thirds of them have a clear view of their talent gaps and strategy to close them, compared to just 25 % of the organizations that are just experimenting with AI. McKinsey also writes that these early adopter firms focus heavily on upskilling and reskilling as a critical part of their talent strategies, as, quote, hiring alone isn't enough to close gaps and outsourcing can hinder strategic skills development. This is really, really important. Yes, new hiring processes are absolutely going to prioritize AI skills as part of it, but you're going to have to work with your existing workforce as well.

15:09Lastly, and this one is really interesting, 40 % of early adopter organizations provide extensive support to encourage employee adoption. In other words, these firms are encouraging people to not keep their AI usage secret and instead share what they're learning. This is pretty much central to the way that superintelligent approaches unlocking organizational AI value. While nominally, we are a platform where people can learn how to use AI tools, in point of fact, the much more valuable aspect of it is getting people to share what they're using AI for. And that happens both between and within organizations.

15:41Our Super for Teams product, for example, is basically entirely focused on getting people to share with their colleagues the high-value AI use cases they're finding that could actually be driving value inside their organizations. Now, this last idea that you have to reinforce the changes to continue transforming is on the one hand obvious, but also even more important in AI than basically any other domain. The speed with which the technology is changing effectively demands that organizations need to build infrastructure that assumes change. McKinsey suggests that governance is a right part of this, and that, quote, a centralized model with a Gen AI dedicated center of excellence helps align AI vision with execution.

16:16A second part, they say, is treating the changes like a true transformation. That means defining its infrastructural roles and measurement criteria and ensuring accountability within business units. Ultimately, though, they say the big thing is mindset. For Gen AI, that means that leaders should visibly adopt generative AI in their own ways of working, that organizations should communicate the reasons behind implementing Gen AI, that there need to be comprehensive and ongoing training programs, and that companies should start to integrate AI goals into performance metrics and evaluation processes.

16:44So that is this study from the front lines. To reiterate why I think this is so important is that as the headlines and the news outlets debate AI hype or not, every organization in the world, every enterprise, every company, every small business, is basically going through this process that McKinsey is describing. Employees are experimenting and iterating without being told to do so, in fact, sometimes in defiance of what they've been told. And organizations are now finally racing to catch up with them in order to actually translate the benefits that they see their employees getting to organization-level and business-level benefit as well.

17:17It is actually, despite these headlines, an incredibly exciting time when it comes to how this technology is finding its way into the workplace. and I hope you now have a better sense of that. That, however, is going to do it for today's AI Daily Brief. Until next time, peace.