Lisa Peyton

Looking forward to advance the conversation on the ethical use of AI in journalism!

15

1 Comment

The local AI movement continues to build momentum. Don't blink. (I'm serious!) Real flags are being planted, especially around #ResponsibleAI. Groups like #Cincy AI for Humans continue to pack the University of Cincinnati’s UC Digital Futures building, proving that the progress is not only real and meaningful but wonderfully authentic. Startups like my own BrandRank.AI are popping up. Another group, CAIC (Cincy AI Catalyst, which I co-founded with Carl Fraik, is pushing responsible AI, AI in education, startup creation, and putting the Midwest on the national map. On top of that yet another force, the Enterprise Technology Association (led by Summer Crenshaw) is spearheading with CAIC and other groups are pushing Cincy AI Week June 11-13 (yes right around the corner). Today we build a huge whiteboard of ideas, concepts, and recommendations for the event. We expect quite a few national leaders to make the trek out here. Again, go to cincyaiweek.com to learn more. Jon Salisbury Nate Sowder Kelly Cohen Jocile Ehrlich Zachary Huhn Christopher Brock Helen Todd (Human) Kendra Ramirez Rob Richardson Elisha Herrmann Michael Beck, MBA Neville G. Pinto JP Nauseef ryan hays Katie Trauth Taylor, PhD Maddie Bell ⚡️💬🗓️ Liz Engel Marco Pagani Mike Halloran Gregory Crawford

80

8 Comments

How can AI turn complex qualitative data into clear insights? Learn how on our un-gated webinar, "How to Prompt Your Way to Qualitative Insight”: https://lnkd.in/ezWAPQ-v And if you want to design AI prompts for your qual projects, book an AI workshop: https://lnkd.in/eKdpSeKU #voxpopme #ai #insights #qual #prompting

17

1 Comment

From Google SGE to Microsoft's integration with OpenAI, Generative AI is changing the #search game. Check out my, Andrew Miller, Francesco Maffei and CMI Media Group's latest POV on AI and Search:

47

We see a lot of media mix modeling RFPs here at Recast. Unfortunately, a lot of them put a ton of emphasis on things that don’t actually matter that much to rolling out a successful MMM practice. Every MMM RFP should focus on 3 things: • Verifiability of the model The first (and most important) thing that you should be thinking about is verifiability and how are you going to validate the results are actually correct. It is trivially easy to run “an MMM”, and it is very, very difficult to run a good one. You want to ask every vendor you meet how transparent are the results going to be and how –specifically– are you going to be able to verify and validate them. What you should look for here is a way to validate the vendor’s results outside of the model itself. • Model operations You also want to think about all the operational components of your MMM. Make sure you’re asking questions like: - What's the refresh cadence going to be? - How often are you actually going to be re-estimating the model and re-estimating all of the models' parameters, or are you just re-using stale parameters on fresh data? - What's the procedure for identifying when things have changed? Beware that a lot of MMM vendors use confusing terms here or have different types of refreshes that aren’t actually refreshers… put simply, you really want to be clear on exactly what's happening. • Usability of the results Here, you want to ask questions like: - How are we going to have access to the results of the model? Is it a PDF / powerpoint? What is available in the dashboard? - How will we actually use these results in our planning workflow? - What underlying data are you going to have access to? If you want to use MMM in a fast-paced, on-demand way, great usability is what unlocks it. If every time you want something you have to ask a support question and they have to ask their team in some foreign country to pull the data, it will slow you down a lot. Whether you use Recast or not, I highly recommend you ask all these questions in your MMM RFP. Now, question for you: would it be useful if we made a free MMM RFP template you could download? If so, we’ll do it. Just let me know in the comments.

32

8 Comments

More details on the deals that OpenAI is hitting up publishers with and potentially its "search" product. https://lnkd.in/gEJ6XmtE I decided to get an impact statement from Google Gemini on Google Search :) Overall, the PPP has the potential to disrupt the search landscape and challenge Google's dominance. The success of the program and its long-term impact will depend on various factors, including user adoption, publisher participation, and ongoing technological advancements. I particularly like these observations Increased Competition for Google: Direct Answers & Reduced Click Through Rates:  The PPP aims to provide direct answers within ChatGPT, potentially reducing the need for users to click through to publisher websites. This could directly compete with Google's featured snippets and knowledge panels, decreasing traffic to Google Search and publisher sites. Improved Search Experience within ChatGPT:  With richer brand expression and priority placement for participating publishers, ChatGPT could become a more attractive platform for information discovery, especially if it consistently provides high-quality and relevant results. This could draw users away from Google Search, particularly for complex or research-oriented queries. Data Advantage for OpenAI:  Access to publisher data allows OpenAI to train its models on a wider range of high-quality content, potentially leading to more accurate and comprehensive search results compared to Google. Changes in User Behavior: Shifting Search Habits:  As users become accustomed to receiving direct answers and engaging with content within ChatGPT, their search habits may evolve. They may rely less on traditional search engines and more on AI-powered platforms like ChatGPT for information discovery. Browsing vs. Searching:  The PPP's emphasis on "browsing" (queries resulting in responses with links) could change how users approach search. Instead of seeking specific answers, users might explore topics more broadly through the curated content and links provided by ChatGPT.

2

Hold up, folks. This AI stuff is getting out of hand! A high school teacher in Maryland just got busted for using AI to create a fake recording of their principal saying some pretty messed up things. The FBI traced it back to the teacher, who clearly didn't think this through. ️‍️ But hey, it's not all bad news in the world of AI! Check out these game-changing martech releases and updates: • Blings' MP5 Technology: Personalized videos at lightning speed • Streetbees' SBX: Real-time consumer insights with photos and videos • Adludio's AdSapiens: AI-powered campaign planning and optimization • Multilocal + Onetag: Access to premium publisher inventory worldwide • CoSchedule's Hire Mia: Free AI social media content creation tools 🆓 • TargetVideo's AI feature: Automatic video content categorization • WordLift's AI SEO Agent: Boosting your SEO and content game • MetaSoul's Azure API: Emotionally expressive AI voices for animations ️ • Geenuity's personalized AI tools: Emotionally resonant content and complex marketing concepts • MCI USA's OneSystem Plus: AI-enhanced event tech for better messaging, analytics, and customer service Want the full scoop? Check out this article: "AI-powered martech releases and news: April 25" https://lnkd.in/gzS6dxDB Happy Friday! Let me know your thoughts. Casey Jones #AI #Martech #Innovation #Marketing #Technology

1

Great write-up by Jo Roberts following my presentation on AI and content automation best practices. Keep it human people!

9

1 Comment

An interesting story from Pew Research: When Online Content Disappears "A quarter of all webpages that existed at one point between 2013 and 2023 are no longer accessible, as of October 2023. In most cases, this is because an individual page was deleted or removed on an otherwise functional website. For older content, this trend is even starker. Some 38% of webpages that existed in 2013 are not available today, compared with 8% of pages that existed in 2023. 38% of webpages that existed in 2013 are no longer accessible a decade later" https://lnkd.in/ejXT6BnG

6

4 Comments

Companies hiring CMOs often ask the finalist candidates to do presentations. These sessions are controversial, optional, and often very revealing and maddening in equal measure! I've structured and attended all manner of these sessions. Some tips from the recent piece I wrote on 'The Anatomy of a Successful CMO Candidate Finalist Presentation.' Pro tips for companies: 1) Structure the guidance to be relevant to your business challenges but also candidate-friendly 2) Make yourself accessible during preparation 3) Observe HOW the candidate educates you and tackles the project; your CMO is also a chief marketing education officer 4) Don't make up your mind too fast on who the 'winner' is: candidates can surprise you 5) Reserve ample room for dialogue in the session Pro tips for candidates: 1) Demonstrate both high-level strategic vision and the ability to translate that into execution and details and measurement 2) Engage in advance if that's an option to get input and provide a preview 3) Be overly-explicit about how your content connects to their challenges 3) Be confident: make yourself and the team comfortable 4) Engage the group; don't just talk AT them 5) Say that you want the job, if that's the case. Make the team feel special! 6) Allay any objections that have come up in the process What have you found works well in these sessions? https://lnkd.in/e-RwGm5d

113

32 Comments

Fellow media, news and publishing professionals… I'm curious to learn more about how other mediaorganizations plan and execute their tentpole initiatives. If you're involved in planning tentpoles for your organization and open to sharing insights via interview, please comment or DM me. All perspectives are welcome (newsroom, revenue, operations, marketing, etc.). Thanks in advance! #publishing #media #news #tentpoles

15

2 Comments

AI-generated dreck posing as local news, aka pink slime journalism, is proliferating. The cost of making it is practically zero. The incentives for bad actors to create it range from money -- by delivering programmatic advertising and affiliate links to unwitting readers -- to political manipulation via misinformation and disinformation. And the platforms that distribute it don't care. John Biggs at The Media Copilot recently lamented this situation. "AI isn’t just stealing our jobs," he wrote. "It’s also going to steal the truth." It doesn't have to, though. Here are two things you can do to protect the truth: 1. Give your attention to real publications. Taproot Edmonton employs actual people who you can meet in real life. We live here and we care about this place. And we put our names on what we publish. You'll never be fooled by pink slime if you intentionally read and listen to work made by humans you know instead of just consuming the news-eque junk an algorithm may serve you. 2. Help real publications pay for those people. A $20/month subscription to ChatGPT can generate an awful lot of words; a full-time employee doing the work of finding stuff out and telling you about it costs considerably more. We don't put a paywall around Taproot's work because we don't want to live in a world where only those who pay get real journalism and the rest get pink slime. That means we have to fund ourselves by other means. If you have a business, you can now buy a business membership to Taproot. That lets you easily sign up all of your employees to receive Taproot's work, which in itself will help your organization better understand the community it serves. But you'll also be helping us continue to inform and connect the people who live in this place that we love. You can sign up for a business membership here: https://lnkd.in/gRWAm_Ay. And you can learn more about what this is and why we're offering it here: https://lnkd.in/gNc3yvAW If a business membership isn't right for you, you can become an individual member for $100 per year or $10 per month here: https://lnkd.in/g6yRVG77 As Mack Male said at Taproot Publishing Inc.'s event during Upper Bound with Katrina Ingram, AI helps our responsible local news outlet scale our ability to pay attention to our community. We're excited about the technology's potential to help us do the work. But humans report and write what you read. "While human-led AI is a boon, non-human-led AI is an existential threat," Biggs writes (https://lnkd.in/gDj55fEm). "We all need to be ready to fight it." I'm inviting you to join us in the fight. #yegmedia #edmonton #yeg #edmontonmedia #futureofmedia

60

6 Comments

Special AI News - If you sell, resell or have users using AI in Colorado view the Colorado Law and AI Developer Requirements to Meet Compliance for Colorado Law and Guidance to help you comply https://lnkd.in/g-QAkj3y #colorado #AI #chatgpt

2

Google Update: New Generative AI in Search - You must learn at least this... Image generation is not quite there yet (notice anything strange in the image below) but Google's latest update: New Generative AI in Search is an SEO game changer, dare I say Shape shifter. I love this stuff but I appreciate most of you may not but these days you just want the high level. This is a tiny summary with micro points of all you need to now: https://lnkd.in/gmWFPjxQ

4

“How do I use AI for research?” Generative AI can be an incredible starting point for background knowledge and research. For example: * Curate industry trends: Identify emerging opportunities and threats within your industry. * Source expert opinions: Gather insights from experts on specific topics for informed decision-making. * Source training materials: Identify and source materials for employee development programs. * Generate technology reports: Provide updates on technological advancements relevant to the industry. * Compile best practices: Gather and compile best practices from industry leaders to enhance business operations. * Source training materials: Identify and source materials for employee development programs. * Identify relevant conferences: Highlight conferences and events for networking and learning opportunities. * Research leads: Get a summary of a potential customer’s firmographics and market positioning. … In other words, anytime you’re looking for a starting point to begin your research. ✏️ RESEARCH-RELATED STARTER PROMPTS ✏️ * What do you know about [general topic? Specifically, what can you tell me about [subtopic], [subtopic], [subtopic]?  * What are some of the most famous books on [topic, topic synonym, or related topic]?  * Explain [topic] at a 12th-grade level. * Now reduce the Lexile level by one and explain it again.  * Summarize [topic] for a [describe audience]. * True or false: [topic] is a [your current understanding of it]? Make your research prompts as specific as possible. Use broad keywords (“Go To Market strategies”) along with specific questions, subtopics, synonyms, or examples (“GTM,” “inbound marketing,” “product-led marketing”). Large Language Models are statistical predictors. LLMs are trained on vast amounts of text data, learning the relationships and probabilities between words. When you provide a broad prompt, the LLM has a wider range of statistically relevant words to choose from, which can lead to a more generic response. Specificity refines the search space: With a specific prompt, you're guiding the LLM toward a smaller and more relevant section of its knowledge base.  Adding subtopics and examples acts like keywords in a search engine, focusing the LLM on the data that’s most likely to be useful and relevant to your request. —— 📌 Want to use generative AI to save time? Just remember RADIC: * Research: To find or collect data on a topic. * Analyze: To interpret or find patterns in data.  * Decide: To make a choice or recommendation based on data.  * Innovate: To combine or reimagine ideas in a new way. * Create: To make something new — whether text, audio, video, image, or code. For a more in-depth explanation of the RADIC framework — and 20+ AI use cases that (nearly) any knowledge worker can apply — read our blog post. Link in the comments. ⬇️

10

1 Comment

A significant unmet need has existed in specialized communities for quite some time. We are used to receiving valuable information on the marketplace from listening platforms and we have data lakes filled with information. Plenty of info exists. But, what we are often missing is the ability to identify earlier patterns of innovation in the fields of technology and science. For example, in enterprise technology, these early indications of change are likely to be discussed in a computer language community or a tech forum well before mainstream media learns about their progress. So, by listening and learning to the world's software programmers, they can accelerate our ability to understand what's next. It is not a volume thing. It is not even a trend that we may see, at first. Rather, it is often a problem being solved or a spark of interest in a new area that shows us what is ahead. Take Discord for example. If we studied Discord's progress within the gaming industry, we would have learned alot, but we may have missed how the introduction of audio chat years ago turned out to be the breakthrough that propelled part of their success. Audio chat doesn't sound like a big deal. But it was and is. The mini-breakthroughs are right in front of us. Our new model, CLAIR, designed by our teams at 104 West Partners and Ringer Sciences, is built to help our clients stay ahead of the market they are helping to build. Looking forward to what's next. Best, Bob The Bliss Group #analytics #technology Boundless Life Sciences Group Next Practices Group #nextsolutionsgroup #innovation

31

2 Comments

Ethan Mollick has shared an interesting paper that claims to show LLMs can be more creative as humans. I think the conclusion may be true, but I don't think the tests use here are as applicable to LLM creativity as to human creativity. The findings primarily rest on the results of a test called the "Divergent Association Task" (DAT) in which participants generate 10 words that are as different from each other in meaning as possible. In humans, the researchers say, this test is associated with creativity. But in LLMs, I'm not sure this is the case, for two reasons: (1) LLMs use embeddings. Semantic distance between words can be calculated. While LLMs themselves don't explicitly calculate semantic distance when producing their DAT lists, I can envision creating a very dumb algorithm that produces maximally diverse DAT lists simply through embeddings—just calculate the word that's most distant from the average embedding of all existing words on the list. Given that I can maximize creative output in this way, it's not clear to me that for LLMs the DAT is a great test of creativity, even if it makes LLMs a great creative brainstorming partner. (2) LLMs are autoregressive. A human might write approach this task from multiple directions, literally. Like, maybe they'd write the first word in their list, then the second word, then think of a new word that might fit between them, or what have you. An LLM is going to always output the next word based on the sequence of the first two words. So, contrary to point (1) above, this might artificially constrain an LLM's creativity, in a way that wouldn't be applicable in other creative exercises. What this all might drive home more than anything is that LLMs think differently, at least in some ways, from us, and just as the jagged frontier exists in other areas, it probably exists in creativity as well. Just because LLMs perform well, or poorly, on the DAT, doesn't mean that the creativity correlations that hold for the DAT in humans hold for them as well.

1

There is no "one size fits all" source for addressability in the post cookie world. Why limit your solution to just one, or even a few identity providers? Check out Narrative I/O's ID-agnostic (graph of graphs) approach, providing customization, scalability, across multiple suppliers. It's also automated and interoperable with your existing stack, and partners. Read more here: #DataCollaboration #IdentityGraph #Narrative #GraphofGraphs

21

1 Comment