Jordan Plawner’s Post

AI for Social Impact and Just Giving  When Debbie and I were on a wooden 2 mast schooner in West Penobscot Bay, Maine we got into an animated conversation with another couple. Michael was born into a large Bostonian working poor Catholic family and any time his parents were asked for help, they gave what was is in their pocket. “I dislike the expression ‘time to give back’. Why wait to be successful to justify helping others. It is always time to just give.” This (re)framing was humbling as I just finalized my role as The AI Strategy Executive Fellow at The Contingent, sponsored by SVP (Social Venture Partners). This opportunity aligns with my passion for leveraging AI tech and org transformation to address social challenges. The Contingent's approach resonates deeply: "We use smart digital strategies combined with empathetic relationships formed through a commitment to equity." Their belief that those experiencing challenges are best suited to design solutions reflects my own AI philosophy. My focus will be developing an AI strategy and product proposal for the Every Child program to accelerate recruitment and retention of families willing to open their homes to children in crisis and improving successful child-foster family matching. While the work is like any analytics project, it is palpable that the bits of information on a dashboard represent a vulnerable child. Two Weeks in and Looking Ahead I'm impressed by the org's mission-driven culture and visionary leadership. The potential to drive AI innovation to improve social outcomes is immense. I will share insights on AI in mission-driven contexts. Some may receive a call-a-friend when I need your expertise. “The opinions expressed are my own and do not necessarily represent the views of The Contingent.” #AIForGood #SocialImpact #AIStrategy #TechForChange The Contingent Social Venture Partners Portland

Sam Altman blogged that superintelligence is coming. He is not wrong since he gives himself "the next couple of decades". Yet, when I see results today like the below from CHATGPT and Perplexity, I wonder what are the architecture assumptions and approaches of the engineering mind that result in delivering fabrications 2 years after CHATGPT? How can that same approach to knowledge deliver super-intelligence? What is inherent in the design that requires one to specify that context of truthfulness for "help find(ing) actual" facts. Sam says AI gets "better with scale". We have another example of a service at scale; social networks. Are they better or smarter the larger they grow? Will super-intelligence emerge from an approach that confuses quantity of data bits with knowledge? #AI #ArtificalIntelligence #Machinelearning #chatgpt #perplexity #knowledge #responsibleAI #AGI

AI Privacy- where to even start? Thrilled to have participated in another local Oregon event, bringing greater AI conversation to our community. This time, we tackled the critical topics of AI privacy and security on a thought-provoking panel. Key takeaways: - AI to enable augmentation of people: I emphasized AI as a predictive tool augmenting human work, urging us to focus on being better humans— improving our innovation, relationships, and critical thinking skills. I argued that we shouldn't expect privacy on social networks or commercial AI systems, regardless of service terms or legislation. - Importance of legislation: Olga Kislinska, Cybersecurity expert challenged my view, advocating for collective efforts to pass and enforce privacy rules no matter how difficult it is to get it right. - Transparency or consent: This debate reminded me of my conversation with Brittan Heller, Stanford University Teaching Fellow, at the ReGenAI Summit. She shared with me the impossibility of AI transparency and doubted if privacy advocates understand what they mean by transparency and how they would connect that to privacy. She advocates for real and enforceable consent in which we know and agree to what will be done with our data. I have doubts legislation can keep up with the capitalist profit impulse and AI tech. Yet, Olga and Brittany convinced me we can’t give up and the more difficult the challenge the more we must move smartly on AI governance while maintaining balance with innovation. #AIEthics #PrivacyInTech #ResponsibleAI #OregonTech Sarah Goad #AI #ArtificialIntelligence Mike Smith Christian Hammer

Of the numerous industry Llama 3.1 announcements today this is the only one I find interesting. It would be ideal if it was open sourced, but the open sourcing incentives today are for incremental accuracy over the additional efforts to deliver efficiency. What makes this announcement interesting? - Real world vs the labs: Examples off a reference model which allow levers for accuracy, speed and efficiency; use cases require difference balances - Focus on inference cost: there is too much focus on training. Many projects fail to get to production due to long term production inference cost. - Focus on the AI infrastructure surrounding the model: containers, Kubernetes, microservices. Yea, it's the boring stuff, but this is why many models handed off from data scientists don't get to production. Congratulation Yonatan G. in continuing your mission inside Nvidia who benefits from models remaining 'fat'

How do we evolve if sourcing and synthesizing knowledge and coding can be automated. Cassie Kozyrkov's post addresses the answer which is in the domain of philosophy and the humanities. Fully agree with her conclusion; "The real work is not in getting the answer but in framing the problem, in knowing which questions to ask, and in critically evaluating the outputs."

Models Work and Don't Have Value, Part VI: Reading between the lines I am a recovering PowerPoint addict. Sean Varley and I had reputations as the best market'tecture slide creators. We could never have enough lines on our slides to capture how data is moved, processed, and stored. Once, Rob Hays, our manager, said, "I don't want to see any more lines. There are no arrows in a server and no lines in the DC. Lines hide the complexity and the gotchas." Which leads me to this ubiquitous image: with arrows and lines that assumes -1. The necessary data is readily available -2. Data inherently contains actionable insights and all you need to do is plug it in (draw a line) The first issue is a showstopper for any AI project. For the second, why do we assume our data has actionable insights? Analysts, researchers, and business specialists iterate endlessly to ask the right questions and then, when reviewing data, dispense with the obvious 'we already knew that' correlated insights, the one-time anomalies, and the noise outside the 95% confidence interval. The analytics leads to a hypothesis to be tested, not an insight to be implemented. Why do we expect actionable insights to be buried in rows and columns, and that there will be 4x insight because the dataset was doubled? Now with AI, claims made with every previous generation of analytics tech are being repeated. Here's one: "Watson AI can help you discover new insights from your data that you may not have known existed." AI does not have context, know about your business, have original thoughts, or understand the difference between 'oh interesting' insights and 'actionable insights' or if actionable insight is even in the data. In many cases, it is just identifying statistical anomalies. So how can AI help uncover 'actionable insight'? The answer lies in a nuanced understanding of AI's capabilities and limitations, coupled with human expertise and domain knowledge. I'll write more later on how organizations can realistically approach AI integration in their data analysis processes. #AI #ArtificalIntelligence #data #ML

Thank you Larry Magid for composing a deeply thoughtful and entertaining Podcast which captured the essence of the RegenAI Summit. Anyone interested in what people are discussing about AI in regions outside the core VC centers, this is a good survey. #AI #artificialintelligence #responsibleAI #futureofwork

Models Work and Don't Have Value, Part V: How to prepare to learn along side a knowledge generating machine? I had the privilege of presenting at RegenAI Summit in Ashland, Oregon— organized by a team dedicated to creating “a future where we grow people and systems that are in alignment.” Immersing myself with educators, farmers, environmentalists, non-profit execs, doctors, storytellers, inventors, lawyers, land managers, creatives, policymakers, and data visualization experts was the most significant de’tech experience I’ve had since leaving big tech. This gathering highlighted that technology is the WHAT, not the WHY. What struck me was the agency exhibited by those who understood little about AI but recognized the need to learn within a community. This reflects a first step in AI learning: reinforcing connectivity with others. While I presented a human-centric AI keynote that resonated with the audience, non-tech civilians are focusing less on what AI is and more on how we model thinking. A conference organizer, Ian Ingram, shared “If historically writing was the externalization of our thought, AI is an externalization of our thought processes.” If writing advanced cognition over centuries, we must prepare for how AI will enhance learning in years. My takeaway is that AI democratization through open sourcing models is a limited idea. The goal should not be to make everyone an AI consumer. Once everyone is using AI, disparities will grow; only some will effectively leverage sophisticated theories of mind to use AI as a co-pilot for advancing their knowledge and developing transformative plans, while others may remain stuck at simplistic prompting that reinforces their worldviews. This raises several important questions: - How do we augment ability to deepen learning? - How can we raise awareness about how AI generates knowledge so that people can engage with it critically rather than just consuming it? - How do we reinforce relationship skills as AI automates some job functions? I’m still exploring where this conversation leads and invite others to share your thoughts below on how we shape a future where technology aligns with our values. Ian Ingram Jay Golden Thor Muller Deepthi Madhava Raiya Kind #AI #ArtificialIntelligence #futureofwork #responsibleAI

This insightful post reminds me of something I read— “for AI to change business operations, an organization must change business operations.” That does not happen because someone open sourced a model, added AI FLOPs to a laptop or used an API to generate a marketing document.

Models Work and Don't Have Value, Part IV: 'Because the Way Life Should Be Needs to Connect with the Way Life Is' After a few weeks disconnecting, I found insight about AI and humanity during a recent personal milestone: dropping our daughter off at university. At her matriculation ceremony, Dr. Wesley Wildman, a professor of Theology and Data Sciences at Boston University, spoke about two ongoing global social experiments: a world without belief and a world with AI. He emphasized AI will disrupt social, civic, and political bonds, and expressed hope that the next generation will work to repair, refine, and renew these connections. Following this event, my wife Debbie and I enjoyed a vacation in Maine, a period she has dubbed our "open nester" phase. At the Blue Hill Library, I saw the below sign (from a Christina Baker Kline novel). This quote resonated with me in relation to our approach to AI. We seek guidance from AI experts who offer either utopian or dystopian future. However, if AI is reshaping our understanding of knowledge and relationships, we cannot rely on the preachers of radical tech change to navigate the future path rooted in understanding society. We wouldn't consult social networking app CEOs on how to raise our device-addicted children. Instead of diving deeper into technology's rabbit hole, we should lead 'beyond the algorithm' conversations—ones that focus on evolving from mere consumers of AI to people with greater consciousness of how we can adapt to this complex technology. Let's replace the model of inevitable robotic super-intelligence with one that presupposes agency to enhance human intelligence alongside AI. This moment is an opportunity to deepen our understanding of knowledge & learning. By doing so, we can successfully integrate intelligent machines into our lives while actively working to balance their adverse impacts. It's time to connect the way life should be with the way life is in our AI-driven world #ArtifcialIntelligence #transformation #genai #futureofwork #machinelearning #innovation Wesley Wildman