Is GPT-5 On the Way? OpenAI Files Trademark

0:00Today on the AI Breakdown, we're doing a deep dive on McKinsey's State of AI 2023 report. Before that on the brief, Meta is preparing to launch AI Personas as OpenAI trademarks GPT-5. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to breakdown.network for more information about our YouTube, our newsletter, and our Discord. Hello friends, a quick request for help before we get into today's brief. I really appreciate each and every one of you who has joined the AI Breakdown journey. The show has pretty consistently been the top AI show on the Apple podcast charts, and that's all because you guys are sharing it, learning from it, giving me ideas for content, and expanding the community to your friends, family, and colleagues.

0:44Well, now we have an exciting job opportunity for someone in this AI breakdown community. We're looking for a researcher and writer who will be working across the podcast, YouTube, and the newsletter. Now, this should be someone who's in the APAC region or Australia or New Zealand, or if they are in Europe, who works very early, or if they are in North or South America, works very late. The point being, this is an overnight roll relative to where I'm sitting on the East Coast of the U.S. If this is you or someone you know, reach out to jobs at breakdown.network. And thanks in advance for your help.

1:17Welcome back to the AI Breakdown Brief, all the AI headline news you need in five or so minutes. Today, we're going to try something a little bit new for clarity. I'm going to actually give you what number topic we're on so you can more easily follow along. And we're actually going to perfectly follow the AI Breakdown First 5, which came out as an email this morning. And we kick it off with topic number one, Dell teaming with NVIDIA on AI. As The Verge puts it, Dell, the PC maker, is going all in on generative AI and offering hardware to run powerful models and a new platform to help organizations get started.

1:49So there are a few parts of this initiative. Dell Generative AI Solutions will help clients with new hardware setups, will offer a managed service platform, and will basically be designed to help big enterprises get access to LLMs, as well as to create unique, customized generative AI projects. So in this way, they're sort of competing in the same space as Amazon, which has been offering managed service platforms for enterprise clients, as well as consultancies like Accenture that have spun up generative AI services for clients as well. Now, they've also partnered in this initiative with NVIDIA for their infrastructure.

2:22They say they're combining NVIDIA's Tensor Core GPU and Dell's Enterprise AI software so that companies can run AI models more efficiently and quicker. Dell also said that it's going to be releasing new computers with built-in AI capabilities, which is something we've heard from other companies as well. TLDR, if you take away one thing from this, is that the AI boom is not just coming for software companies, but for hardware makers as well. Next up on the AI Breakdown Brief, topic number two, it appears that Instagram is working on an AI content label. This report comes from Alessandro Paluzzi, who is best known for leaking future features that come from the meta empire of companies.

3:00A couple days ago, he tweeted, Instagram is working to label the contents created or modified by AI in order to be identified more easily. He then showed a screenshot that appeared to be a notification from Instagram, which reads, The creator or meta said that this content was created or edited with AI. What is generative AI? People use AI tools to create text, images, and video from single descriptions. How to know when posts use AI? Content created with AI is typically labeled so that it can be easily identified. Now, in addition to that, it also appears that Instagram is working on a message summarization feature that will live inside the direct messages tab, and Alessandro assumes, and I agree, that it will be powered by meta AI.

3:40Now, of course, the question with this is, is this meta being responsible and trying to help avoid confusion and deepfakes and things like that? Or is it them simply trying to get ahead of regulations, which will likely require some sort of content labeling like this? Of course, the best answer is likely that it's both. Topic number three, and frankly, the one that has the biggest potential for buzz, is that OpenAI has apparently filed a trademark application for GPT-5. YK, aka CSDojo, posted this on Twitter. He writes, OpenAI has filed a trademark application for GPT-5, which includes software for the artificial production of human speech and text, conversion of audio data files into text, voice and speech recognition, machine learning-based language and speech processing.

4:25Now, this filing was made on July 18th, 2023, so just a couple weeks ago. From a feature standpoint, the thing that most people have taken note of is the artificial production of human speech part, wondering if OpenAI is maybe getting into the space currently inhabited by companies like 11 Labs. However, others did point out that the actual language appears pretty identical to GPT-4. Now, the last we heard about GPT-5 came in June. At a conference hosted by the Indian newspaper Economic Times, Altman said that they were still not training GPT-5 yet. This repeated statements that he had made in April, and Altman said, We have a lot of work to do before we start that model.

5:01We're working on the new ideas that we think we need for it, but we are certainly not close to it to start. Altman has also continued to point out that even if they started training GPT-5, there would be a whole additional process before it was released to the public. Altman said, when we finished GPT-4, it took us more than six months until we were ready to release it. My strong guess is that OpenAI believes, and I wouldn't be surprised if there are behind-the-scenes conversation to this effect in Washington, that any more advanced models than we have right now, than GPT-4, are going to require some additional level of scrutiny before being released to the public.

5:36In fact, at this stage, I would be shocked if that wasn't the case. So is the timing of the GPT-5 trademark indicative of a training process maybe coming closer to starting? Or is it just a company dotting its I's and crossing its T's? For now, it's just going to have to stay in the realm of speculation. Something that isn't in the realm of speculation and is frankly just about the most expected news that we could possibly have here in AI world is that Google's Assistant app is getting an AI makeover. First reported by Axios after getting access to an internal email that was sent to employees working on Google Assistant, the app is going to be totally overhauled focused on using generative AI such as the Barred chatbot.

6:15Now as part of this move, Google is reorganizing the teams that work on the product and that a small number of layoffs are happening as part of the shift. In follow-up, Google told Axios that the number of jobs being eliminated was in the dozens, while the number of people who work on Assistant numbers in the thousands. The email begins, Since we launched Assistant seven years ago, we've built great experiences for the hundreds of millions of people who use it every month. And we've heard people's strong desire for assertive conversational technology that can improve their lives. As a team, we need to focus on delivering high-quality, critical product experiences for our users.

6:47We've also seen the profound potential of generative AI to transform people's lives and see a huge opportunity to explore what a supercharged assistant powered by the latest LLM technology would look like. A portion of the team has already started working on this, beginning with mobile. And from there, they go into the details around how things are being reorganized. Showing his prescience, Peter Yang, an AI content creator, had tweeted just about a day before this, when is someone going to port an LLM over to Alexa, Google Assistant, or Siri? These devices are dumb as hell. Well, Peter, your answer is right now, at least in the case of Google Assistant.

7:20Finally, Meta could launch AI Personas, a new human-like chatbot feature that lives inside their apps, as early as this month. The story was a scoop from the Financial Times and basically pins it on a need to keep users engaged. The FT spoke to three people with knowledge of the plans and said that the chatbots that are being worked on take the form of different characters. For example, one advises about travel in the style of a cool surfer dude, and another emulates Abraham Lincoln. FT writes, The chatbots could launch as soon as September. Their purpose will be to provide a new search function and offer recommendations, as well as being a fun product for people to play with.

7:56Now, they also point out that on top of just increasing engagement, in other words, the amount of time people spend on Instagram, Facebook, etc., the chat interface could also collect a ton of new information about what users are interested in and what they're looking for, which would help Meta with their main business model of serving up targeted ads. There are, of course, other personality-style chatbots in the space. Character AI has been a hot-button Silicon Valley darling, which generates conversation in the style of specific people, and Snap also launched its My AI feature in February, saying that 150 million users have interacted with it so far.

8:29Meta did not confirm the AI personas, but last week on their earnings call, CEO Mark Zuckerberg said that they would release more details about their roadmap for AI at their Connect Developer event, which is happening in September. So that is certainly an event to look forward to. But for now, that's going to do it for today's AI Breakdown Brief. If you enjoyed it, do me a favor, go check out the AI Breakdown newsletter. You can find it at the AIBreakdown.Beehive.com, that's B-E-E-H-I-I-V.com, or you can find a link on Breakdown.network. Thanks as always for listening and watching, and I'll be back soon with the main AI Breakdown.

9:04Before we get into the main AI breakdown, I want to tell you about today's sponsor, Supermanage. If you work in a professional setting, you probably have some version of a one-on-one meeting, either with the people that work for you or the people that you work with. Unfortunately, all too often, those one-on-one meetings become glorified catch-up calls. Don't you wish you could jump right to the stuff that really matters? That's where Supermanage comes in. Supermanage AI magically distills your team's public Slack channels into a real-time brief on any employee, any time. Catch up on contributions, work in progress, challenges they're facing, sentiment, everything you need to show up ready for a truly meaningful conversation.

9:43And it's completely free. Visit supermanage.ai forward slash breakdown today to start making the most of your one-on-ones. And thanks again to Supermanage for sponsoring the AI Breakdown. Welcome back to the AI Breakdown. Today we are looking at a just released report by Quantum Black, which is the AI division of consulting giant McKinsey. Now, some of you are rightly probably skeptical of McKinsey reports or really any reports from these big consulting houses. I think that's especially warranted when we're dealing with predictions, right? These firms have a tendency to overstate how big different industries are going to get or how significant big changes are because they get paid to help navigate change.

10:25At the same time, this isn't a set of predictions. This is a survey of actual behaviors, attitudes, and usage. And that, I think, makes it a little bit different. Now, to the extent that there's anything to be skeptical of or just hold with a grain of salt, it's that this survey was in the field in the middle of April. In other words, there is roughly the same amount of time between chat GPT launching and when this survey was conducted as there is between when this survey was conducted and now. I think that makes it fairly safe to assume that all of these numbers, when it comes to enterprise adoption, are likely higher than they were in this survey.

10:56So by way of highlights, they say the latest annual McKinsey Global Survey on the current state of AI confirms the explosive growth of generative AI tools, or as they shorthand, Gen AI. Less than a year after many of these tools debuted, they said, One-third of our survey respondents say their organizations are using Gen.AI regularly in at least one business function. AI has risen from a topic relegated to tech employees to a focus of company leaders. Nearly one-quarter of surveyed C-suite executives say they are personally using Gen.AI tools for work, and 40 % of respondents say their organizations will increase their investment in AI overall because of advances in Gen.AI.

11:31Their first section is all about general adoption. They sum up, it's early days still, but use of Gen.AI is already widespread. Highlights include 79 % of respondents saying they've had at least some exposure to Gen.AI, either for work or outside of work, and 22 % saying they're regularly using it in their own work. So let's look at who's already using these tools. Perhaps unsurprisingly, the industry that has the highest exposure and usage is the technology, media, and telecom space, as McKinsey defines it. Of that cohort, 14 % say they're regularly using it for work, 19 % say they're regularly using it for work and outside of work, 17 % say they're regularly using it outside of work, and 37 % said they've at least tried it once.

12:14Only 12 % have no exposure or don't know. The next highest category is financial services. 8 % regularly use it for work, 16 % regularly use it for work and outside of work, that makes almost a quarter of people using it regularly. Then another 18 % on top of that regularly use it outside of work but not at work, and 41 % have tried at least once. Next up include categories like business, legal, and professional services, consumer goods and retail, with healthcare, pharma, and medical products and advanced industries, whatever that means, taking up the rear. Reorganizing by region, North America saw the highest percentage of people using it regularly for work and outside of work at 28%, followed by Europe at 24%, Asia Pacific at 22%, developing markets at 20%, and perhaps surprisingly greater China at 19%.

13:01Now, I don't know how much we should read into this, given that these are all pretty similar numbers. But is it possible that the tension between, on the one hand, wanting AI to be a tool that allows China to leapfrog its Western competitors, and on the other, the concern that it might undermine social control, could that be affecting these numbers? I'm not sure. There are pretty similar levels of use by C-suite executives, senior managers, and mid-level managers. And not surprisingly at all, the younger you are, the more likely you are to have used these tools or be regularly using them for work.

13:31Now, what about where in the organization these tools are being used? The most common areas are in marketing and sales, product and service development, and service operations. Again, this strikes me as right what you might expect. The percentage of respondents that say their organizations are regularly using generative AI in a given function, 14 % of companies are using it regularly in marketing and sales, 13 % are using it for product and or service development, 10 % are using it for service operations, and from there, the numbers go down significantly. 4 % are using it for risk, 4 % are using it for strategy and corporate finance, 3 % for HR, 3 % for supply chain management, and only 2 % for manufacturing.

14:08Within those functions, a lot of it looks like basically what you can do with ChatGPT really well. Marketing and sales, for example, is using it to craft first drafts of text documents, summarize text documents, and support personalized marketing. Perhaps even more than how people are using it now, the survey also reflects where people anticipate it to go. McKinsey says that three-quarters of all respondents expect Gen.AI to cause significant or disruptive change in the nature of their industry's competition in the next three years. Industries that rely most heavily on knowledge work have a double-edged sword.

14:39They're likely to see the most disruption, but they're also potentially likely to reap the most value. Now, what are reasons an organization might not be adopting it yet? The risks that organizations think are most relevant include right at the top, inaccuracy at 56%, cybersecurity with 53 % of organizations being concerned, intellectual property infringement at 46%, and regulatory compliance at 45%. Now, interestingly, all of those are reasons why on-premise customized solutions might be a preferred option for many companies, as opposed to the business version of ChatGPT as a for example. Of those issues, the one that organizations are most working to mitigate risk around right now is cybersecurity, with 38 % of organizations working on that issue.

15:19Now, what are organizations' perceptions of the value they might get from AI? Well, across all organizations, when asked what their top objective for generative AI efforts were, 33 % said reduce costs in core businesses, 33 % said increase value of operating by integrating AI-based features or insights, 21 % said increase revenue from core business, and 12 % said create new business and or sources of revenue. So two-thirds of all respondents say that their top priority is either reducing their core costs or increasing the value of their existing offerings. And really, when it comes to increasing the value of existing offerings, you also had 21 % say they wanted to increase revenue from core businesses, which has to be seen as pretty similar.

16:01Interestingly, these priorities changed among what McKinsey called AI high performers, which are basically companies that had 20 % of their revenue or more from AI last year. The top priority for those companies was 30 % increasing the value of their offerings and 27 % increasing the revenue from core businesses, but then 23 % to creating new businesses and or sources of revenue up from 12 % of the general pool and only 19 % to reduce costs in core businesses down from 33 % among the general pool. Now when it comes to the biggest stumbling blocks or barriers, for those high performers, it's models and tools and talent that are the most important issues.

16:3524 % say that models and tools are their most common stumbling block, and 20 % say talent is their common stumbling block. When it comes to all the other respondents, it sort of appears like they haven't even had the chance to have models and tools become a problem yet. Only 6 % cite that as their common stumbling block. Among that group, 24 % say that strategy is the big issue. Talent was also an issue for the general group, with 21 % saying that it was the most important stumbling block. And that brings us to McKinsey's third category of insights. AI-related talent needs shift, and AI's workforce effects are expected to be substantial.

17:08Interestingly, while hiring for AI-related roles remains a challenge for companies, the reported difficulty has decreased since 2022 for many roles. That includes AI data scientists, translators, data architects, software engineers, data engineers, design specialists, and data visualization specialists. Perhaps this reflects just the greater number of people who are orienting their careers around AI, as it's having, as McKinsey calls it, its breakout year. But what about companies' assumptions about how AI will impact their actual workforces? There's definitely not a strong confidence interval, I don't think.

17:39In fact, 12 % said that they just don't know how to expect a change in the number of employees. 30 % said they believe there will be little or no change. 25 % said they thought there would be modest decreases of 3 to 10 percent, 10 percent identified a larger increase of between 11 and 20 percent, and only 8 percent anticipated a decrease of more than 20 percent. In total, only 15 percent thought that it would increase the number of employees by any amount. However, when it comes to expectations about the percentage of employees that will need to be re-skilled, that is much, much clearer. 38 percent of respondents said that they believe that over 20 percent of employees will need to be re-skilled, with another 18 percent saying they expect between 11 and 20 percent needing to be re-skilled.

18:19Now, McKinsey's fourth insight category argues that AI adoption and impact remain steady. They write, while the use of Gen AI tools is spreading rapidly, the survey data doesn't show that these newer tools are propelling organizations' overall AI adoption. The share of organizations that have adopted AI overall remains steady, at least for the moment, with 55 % of respondents reporting that their organizations have adopted AI. What's more, less than a third of respondents say their organizations use AI in more than one function. Now, when it comes to those two big goals, cost reduction and revenue increase, across all job functions, in 2022, 4 % of organizations said that they had greater than 20 % cost decreases from AI adoption, and another 10 % said that they had seen cost decreases from AI adoption of between 10 % and 19%.

19:04For revenue increases from AI adoption in 2022, 6 % of companies saw greater than 10 % revenue increases in 2022, with another 18 % seeing between 6 % and 10 % increases. Now, when it came to Twitter reactions to this research, my favorite weren't actually reactions per se, but people like Stan Gerard, who's the creator of Quiver, which I featured on the show a couple weeks ago, using AI tools to not have to read the report at all and simply ask questions to the report. As Stan said, it helped me understand it in 22 seconds. When it comes to my reflections, I think there are two things that stand out to me.

19:38One, when push comes to shove, I think that organizations are likely to bifurcate generative AI strategy into two areas. The first is the executive level and what people get to say they're doing, so long as that continues to make press headlines. And the second is how it actually functions and is useful within specific departments and organizations. I would expect by the end of this year, for example, that we will see easily over half, if not two-thirds or more of all marketing departments using these tools in some way. It's simply too powerful and frankly not really controversial to use it for that function.

20:08And I don't think it'll take any mandates from the top for marketers to natively adopt things that actually make their lives easier and make their outputs better. The second thing that stands out to me is that I do think that there's an adoption lag from moments of peak hype. Even though these survey results weren't that different from what McKinsey saw in 21 and 22, I think that if they did this again right now, and who knows, perhaps they are, they'd find very different results than even just three months ago. In other words, my suspicion is that April wasn't sufficiently far out from the launch of ChatGPT in the first place for this data to really capture the impact of GPT-4 level tools in the enterprise in 2023.

20:44But who knows? We'll have to wait and see. Anyways, guys, hopefully this was interesting and gave you a little bit better of a sense of how corporations and enterprises are looking at, as McKinsey calls it, Gen AI. I appreciate you listening or watching, and until next time, peace.

21:05It was so strong. It could have been a chance and it could have a sequel a season But don't you want to have a ticket under interessant? You want to give either of them you? you? Thank you.