Paul Roetzer’s Post

Here's a quick example of finding valuable AI use cases in real-time once you know what AI is capable of doing. Yesterday morning a journalist from AFP, a French news agency, reached out asking for my thoughts on the Google AI Overviews announcement and the impact that it might have on search. I wasn't familiar with AFP, or the journalist, so I put a quick prompt into ChatGPT to learn more. Prompt: Give me a summary of the type of stories that Thomas Urbain of AFP writes. Include links to some examples. It gave me a great summary of his writing, including links to recent articles. But, having owned a PR and marketing agency for 16 years, I know sometimes journalists have a certain angle they're going for when they conduct interviews, and it's good to be prepared. So, I followed up with ChatGPT and asked the following question. Prompt: What's the overall tone of his stories? Are they fact-based, or do they take a position that could be controversial? ChatGPT: "Thomas Urbain's stories for AFP exhibit a predominantly fact-based and objective tone. He focuses on delivering comprehensive coverage of events, presenting facts clearly and accurately without overtly taking sides or injecting personal opinion. His articles are detailed and thorough, often exploring the broader implications of the topics he covers." It continued on with a more in-depth summary of some specific articles to highlight his fact-based approach and then offered a conclusion about his overall tone and style.  I then performed the same sequence of prompts in Google Gemini and Perplexity to verify the information and compare the outputs. All told, in about 3 minutes I had the information I needed to reply to the inquiry and schedule a call with Thomas. As someone who has done PR his entire life I can tell you this process would have taken me 45-60 minutes to complete through traditional Google searches and note taking. And given how busy my day was already, I probably would have just passed on setting up the interview due to time constraints. When you understand what AI is capable of, you will continually find valuable use cases for it throughout your day. Keep learning and experimenting. And thanks to Thomas for a great conversation and writing a balanced, fact-based article. Just as the AI told me he would. https://lnkd.in/dg2zSHeQ

The Google I/O 2024 event was packed with hours of AI announcements and previews that have implications to developers, business leaders and knowledge workers across every industry. (More thoughts on all the details soon.) But, one thing they didn’t address on stage, which has the potential to be very disruptive to the Internet in the near term, is the impact of their new AI Overviews on brands and publishers who rely on Google search traffic. (Note: AI Overviews are generative AI summaries at the top of search results pages rather than blue links and ads. Google says users seem to like and engage with them, and personally I’ve found the search generative experience to be better than traditional results for many types of searches.) We’ve been under the assumption for the last year that organic search for many brands would decline as generative AI was infused into search, but Google appears willing to go all-in on AI Overviews without a clear understanding of how it will affect search traffic (or, at least they aren’t sharing data if they do know). This has the potential to rapidly deteriorate the value exchange between brands / publishers and Google. Publishers create helpful content, and Google rewards that content with traffic. That’s how it’s worked for 20+ years. But if the traffic doesn’t come, will there be enough incentive for publishers to create it? No one really seems to know what happens next here, and what it will mean to marketing and business. According to Google, “With AI Overviews, people are visiting a greater diversity of websites for help with more complex questions. And we see that the links included in AI Overviews get more clicks than if the page had appeared as a traditional web listing for that query. As we expand this experience, we’ll continue to focus on sending valuable traffic to publishers and creators. As always, ads will continue to appear in dedicated slots throughout the page, with clear labeling to distinguish between organic and sponsored results.” I’ll drop a couple insightful articles in the comments section that expand on the uncertainties and doubts. Definitely a topic to monitor moving forward. https://lnkd.in/d_4C6tdD

We’ve got a new, multi-modal model from OpenAI that is faster, better at reasoning, much cheaper for developers, and free to all ChatGPT users. It can see, hear and speak (with significantly less latency and way more expression), which will open up a wealth of new applications once it’s fully rolled out. The model name is GPT-4o (“o” stands for “omni”), and the text and image capabilities started rolling out yesterday. Now, none of that was known when we recorded this week’s podcast episode last Friday (but we did have a lot of rumors swirling), so if you’re looking for deep analysis of GPT-4o from me and Mike, you’ll have to wait until episode 98 next week! But, we had plenty to talk about on episode 97: Timestamps 00:03:15 — OpenAI is Planning Something Big (which we now know was GPT-4o) 00:11:35 — Google’s CEO Sundar Pichai Lays Out His AI Roadmap 00:18:00 — Microsoft Readies New AI Model 00:20:42 — Anthropic Founders Share Roadmap to Advance AI 00:25:36 — Zapier’s New AI Agent Central 00:28:39 — Microsoft’s Fourth Annual Work Trend Index: AI at Work Is Here 00:33:54 — Microsoft's New Prompting Features in Copilot 00:36:42 — Google DeepMind’s new AI can model DNA, RNA, and ‘all life’s molecules 00:40:57 — AI Companions https://lnkd.in/dEYYQv7y

We are heading into a big week of AI news. OpenAI has announced a virtual event on Monday at 1 pm ET, and the Google I/O developer conference is on Tuesday. According to Sam, regarding the Monday event: “not gpt-5, not a search engine, but we’ve been hard at work on some new stuff we think people will love! feels like magic to me.” You can get warmed up with the All-In podcast interview with Sam that just dropped. Here are a few takeaways: We may not get a GPT-5 launch. It may just be a continued improvement of the model. ”We may release it in a different way than we've released previous models. Also, I don't even know if we'll call it GPT-5. What I will say is, a lot of people have noticed how much better GPT-4 has gotten since we've released it, and particularly over the last few months. I think that's a better hint of what the world looks like, where it's not the 1, 2, 3, 4, 5, 6, 7, but you use an AI system, and the whole system just gets better and better fairly continuously. I think that's both a better technological direction, I think that's easier for society to adapt to, but I assume that's where we'll head.” Sam LOVES the iPhone, which may give credence to the growing rumors of them collaborating with Apple on next-gen Siri (or some larger collaboration with OpenAI apps). Voice will be a key either way. ”I love great new form factors of computing. And it feels like with every major technological advance, a new thing becomes possible. Phones are unbelievably good. So I think the threshold is very high here. I personally think an iPhone is the greatest piece of technology humanity has ever made.” More NVIDIA chips, more data centers and more energy matter to making smarter models, but there are unlocks in compute efficiency likely coming that will accelerate AI development. “we'll make huge algorithmic gains for sure, and I don't want to discount that. I'm very interested in chips and energy, but if we can make a same quality model twice as efficient, that's like we had twice as much compute.” He wants his personal AI to function like a great senior employee. This is a hint at the direction they are heading. “I want a great senior employee. It may get to know me very well. I may delegate it. You can have access to my email and I'll tell you the constraints, but I think of it as this separate entity. I personally like the separate entity approach better and think that's where we're going to head.“ He talks as though training data will be commoditized, and the real value will be access to unique data sources at inference (when the user interacts with the AI). This is likely why they are aggressively pursuing licensing deals. It’s the inference that matters more in the future. “As training data becomes less valuable, and what the system does accessing information in context in real time, or taking something like that, what happens at inference time will become more debated and what the new economic model is there.”

“ChatGPT is mildly embarrassing at best. GPT-4 is the dumbest model any of you will ever have to use again, by a lot. But it's important to ship early and often, and we believe in iterative deployment. If we go build AGI in a basement, and then the world is kind of blissfully walking blindfolded along, I don't think that makes us very good neighbors.” That was Sam Altman, co-founder and CEO of OpenAI, during a recent appearance at Stanford. He went on to offer some vague but meaningful insights into what we can expect from GPT-5 (or whatever they decide to name it): “The most important thing about GPT-5, or whatever we call it, is just that it's going to be smarter. . . .  I think that's among the most remarkable facts in human history, that we can just do something, and we can say right now with a high degree of scientific certainty, GPT-5 is going to be a lot smarter than GPT-4. GPT-6 is going to be a lot smarter than GPT-5. And we are not near the top of this curve. “And we kind of know what to do. This is not like it's going to get better in one area. This is not like it’s only going to get better at this eval or this subject or this modality. It's just going to be smarter in the general sense and I think the gravity of that statement is still underrated.” Episode 96 of The Artificial Intelligence Show kicks off with a conversation about Altman’s interview. Timestamps 00:04:22 — Sam Altman’s Possibilities of AI Talk at Stanford 00:15:27 — Microsoft’s Rushed OpenAI Decision (and the Email They Didn’t Want Released) 00:29:57 — Amazon Q is Now Available 00:36:24 — Why Perplexity Might Not Last 00:41:17 — Major U.S. Newspapers Sue OpenAI and Microsoft 00:43:45 — Elon Musk’s Plan For AI News 00:50:29 — First Commissioned Music Video Made 100% with Sora 00:52:47 — New Bill Would Require a Kill Switch for AI Models 00:56:32 — More Bad News for AI Hardware 01:01:29 — Quickfire AI Product Updates (ChatGPT Memory, Claude Team Plan and App, HeyGen Instant Highlights, Chat with Gemini in Chrome, Yelp’s AI Assistant) https://lnkd.in/gcyTQwsb

Locked in my MAICON opening keynote abstract this week. It’s going to be a continuation of the possible AI timeline I presented on episode 87 of The Artificial Intelligence Show podcast on March 12, 2024 (our most downloaded episode ever). That v1 timeline (which you can see in the attached PDF) was a very rough initial concept (literally drafted on a 3-hour flight that morning), but I haven’t stopped thinking about it. The plan is to continue researching, developing and visualizing the model in the months ahead. Assuming the timeline is at least directionally true, the questions I’m interested in exploring include: > What are the probabilities of the different phases and milestones occurring? > What can accelerate AI progress? > What can slow AI progress? > What does it mean to marketing? > What does it mean to business? > What does it mean to investing / startups? > What does it mean to education? > What does it mean to society? > What does it mean to the economy and workforce? > What remains uniquely human? > What can we do to thrive in the transformation? Here’s the MAICON 2024 keynote abstract: The (Incomplete) AI Timeline: What Happens Next, What It Means, and What We Can Do About it For more than 70 years, researchers have pursued artificial intelligence. They were driven by a belief that we could give machines the ability to think, reason, understand, create, and take actions in the digital and physical worlds. But, progress was often slow, and the impact on our professional lives was minimal. Then, everything changed—and accelerated—with the release of ChatGPT in November 2022, and the rise of Generative AI. Continuing AI advancements in language, vision, prediction, persuasion, reasoning, decisioning, and action will augment human capabilities and redefine knowledge work at a rate and scale that the economy has never seen. Millions of jobs will be impacted as companies realize the power and potential of AI to drive productivity, efficiency, and profits. Every business in every industry faces the opportunity to disrupt, and the risk of being disrupted. And this is the least capable AI we will ever have. So, what happens next? Join us for Paul Roetzer’s opening keynote in which he explores a possible near-term future in which smarter and more powerful AI models lead to an explosion in intelligence, embodied within multi-modal language models, AI agents and robotics. The future is coming faster than you may think. Be prepared to lead the transformation.

As a podcast host, it’s hard to really know who your listeners are (podcast analytics are notoriously opaque), what matters to them, and how (if at all) they’re getting value from your show. So reviews play a huge role, not only in helping to get the show found by other potential listeners, but to help the host(s) know what they’re doing well, and maybe how they can improve. Mike Kaput and I have been doing the weekly format for The Artificial Intelligence Show since October 2022, about one month before ChatGPT came into the world. Since that time we’ve gone from a couple hundred listeners to ~40,000 downloads per month. A few episodes ago we started adding a call-to-action at the end to leave a review if you’re enjoying the show. I looked on Apple Podcasts today and there was some incredible feedback from our audience (see screenshots below). I just want to share those reviews here to say how much they mean to us, and to encourage people to take the time to rate and write reviews for the podcasts (and books) that help you on your journey. Hosting podcasts and writing books can be a bit lonely. You take the leap to put your ideas out into the world, and you don’t really know the impact they’re having until someone tells you. So maybe pay it forward today and let someone know you appreciate their work. It matters to them. And thanks to everyone who listens to our podcast, especially if you’re one of those people who has taken the time to share a review. We appreciate all of you!

Where are all the AI-Emergent companies? In a May 2022 blog post titled The Future of Business is AI, or Obsolete, I wrote that, “With each day that passes, and each advancement in artificial intelligence language and vision technology, it is becoming more apparent that there will be three types of businesses in every industry: AI Native, AI Emergent and Obsolete.” In that post, we defined AI Emergent companies as, “established organizations that move quickly to adopt and scale AI across all areas of the organization. “They have visionary leaders who see the rapid advancement in AI capabilities, and invest the resources needed to build a smarter business. “These emergent companies have expanding AI and machine learning (ML) talent pools, they innovate faster than the competition, and they excel at personalization across marketing, sales, and service. “They have the data, customer bases and infrastructures to withstand the AI Native companies, if they move fast enough to transform. Plus, they have the money to acquire the AI Native companies before they grow to dominate.” On this week’s podcast, we start off with a case study on Moderna’s evolution into an AI Emergent company through its partnership with OpenAI. It’s a great example of what is possible when AI is approached by leadership as a business transformation opportunity vs a collection of tech tools. Moderna set an aggressive growth goal of, “15 new products in the next 5 years.” Moderna CEO, Stéphane Bancel, said, “If we had to do it the old biopharmaceutical ways, we might need a hundred thousand people today. We really believe we can maximize our impact on patients with a few thousand people, using technology and AI to scale the company.” Moderna is investing heavily in AI education, change management initiatives, and 100% adoption and proficiency of Gen AI by all its people with access to the tech. “Collective change management initiatives included an AI prompt contest to identify the top 100 AI power users who were then structured as a cohort of internal Generative AI Champions. Moderna’s culture of learning led to local office hours in every business line and geography, and scaled through an internal forum on AI, which now has 2,000 active weekly participants.” This is the way. 00:04:24 — AI Emergent Case Studies 00:20:20 — Microsoft Phi 3 00:29:19 — OpenAI’s Sam Altman and Other Tech Leaders to Serve on AI Safety Board 00:30:34 — OpenAI CEO Sam Altman talks AI development and society 00:35:04 — TSMC’s Debacle in the American Desert 00:39:53 — Expansion of Meta Ray-Ban Smart Glasses Collection 00:43:13 — Eric Schmidt-backed Augment launches out of stealth with $252M 00:45:59 — Musk’s xAI Is Close to Raising $6 Billion from Sequoia, Others 00:48:16 — Apple Intensifies Talks With OpenAI for iPhone Generative AI Features 00:50:33 — HubSpot AI Updates https://lnkd.in/gwgV6UUp

I’m working on a new Generative AI Policies course for our upcoming Scaling AI series, and looking for some solid examples. I’m planning to feature a collection of examples in the course to show how different organizations are guiding their employees’ use of generative AI text, image, video, audio and code technologies. For example, a generative AI policy would answer questions such as: * Are employees allowed to use AI tools to generate external content? * Are employees allowed to use AI tools to generate internal content? * Are employees allowed to use AI tools to outline, ideate, plan, etc? * Do employees have to disclose their use of AI in final outputs for internal audiences? * Do employees have to disclose their use of AI in final outputs for external audiences? * What data are employees permitted to use when working with generative AI tools? If you have one you’re willing (and able) to share, or have seen a great public-facing document from another company, please comment below. I’ll also add some examples in the comments for people to reference.

Dario Amodei, the cofounder and CEO of Anthropic, recently gave one of the most enlightening and unnerving interviews about the future of AI that I have heard in a long time. Amodei led the team building GPT-3 at OpenAI before leaving (and taking ~10% of the OpenAI staff with him) to start Anthropic in 2021. Anthropic has gone on to raise $7B+, and build one of the world’s most powerful LLMs, Claude Opus. The interview with Ezra Klein covers Amodei’s thoughts on the importance of scaling laws (more data + more compute + more training time = smarter and more powerful models) to forecasting the future abilities of AI models, AGI, AI’s emerging persuasion capabilities, his team’s efforts to understand how the models “think” (i.e. mechanistic interpretability), grounding models in truth (although society can’t seem to agree what truth is), copyright law, and so much more. But, the part of the interview that had me losing sleep was Amodei’s take on the near-term threats we face as the models get more powerful. The Anthropic team’s Responsible Scaling Policy has four AI Safety Levels (ASLs). Levels 1 and 2 (where we are now) are fine, in their opinion. Level 3 is a problem. Level 4 is potentially catastrophic. EZRA KLEIN: When you imagine how many years away, just roughly, A.S.L. 3 is and how many years away A.S.L. 4 is, right, you’ve thought a lot about this exponential scaling curve. If you just had to guess, what are we talking about? DARIO AMODEI: Yeah, I think A.S.L. 3 could easily happen this year or next year. I think A.S.L. 4 — EZRA KLEIN: Oh, Jesus Christ. DARIO AMODEI: No, no, I told you. I’m a believer in exponentials. I think A.S.L. 4 could happen anywhere from 2025 to 2028. Mike and I spent the first 20 minutes of Episode 94 discussing the Amodei interview. I feel like we could have done a full 2-hour show on it. Timestamps 00:04:09 — Anthropic CEO on the Trajectory of AI 00:23:07 — Stanford Releases 2024 AI Index Report 00:29:19 — Microsoft, Google, OpenAI Are Racing Towards Agents 00:32:40 — Rise in AI Job Roles 00:35:30 — AI is Already Reshaping Newsrooms 00:37:43 — How to Opt Out of AI 00:41:13 — AI Mind Reading 00:45:47 — Google Makes Structural Changes to Support AI Efforts 00:48:33 — The Ethics of Advanced AI Assistants 00:54:13 — On the Economics of Frontier Models 00:58:59 — Perplexity Raises $62.7M, Announces Enterprise Pro https://lnkd.in/gGtWHe3C