Recent Developments in Applied AI and Machine Learning

𝗕𝗲𝘆𝗼𝗻𝗱 𝘁𝗵𝗲 𝗔𝗜 𝗕𝘂𝗯𝗯𝗹𝗲: 𝗥𝗲𝗮𝗹 𝗘𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗩𝗮𝗹𝘂𝗲 𝗶𝗻 𝟮𝟬𝟮𝟲 I've researched the Value Creation story behind Applied-AI progress. It's far more substantive than the myopic AI Bubble headlines suggest. As someone who's tracked digital transformation for years, I've been eager to fully evaluate Applied-AI Initiatives impact in the Global Networked Economy. So, let's break it down: 💡 Applied-AI has evolved from experimental pilots to large-scale deployments that deliver measurable value and ROI. 💡 Between 2023 and 2025, it led in innovation scores among emerging tech and ranked in the top five for global technology investments. 💡 We're seeing gigawatt-scale "AI Factories" powering Agentic AI systems that autonomously reason, plan, and execute complex tasks. These are backed by strategic energy alliances, like the U.S. DOE's "Speed to Power" initiative, which accelerates multi-gigawatt projects to handle 25% domestic load growth from data centers by 2030. On the corporate front, giants like JPMorgan Chase are investing $2 billion annually in AI, with over 200,000 employees using their LLM Suite daily. Nationally, the U.S. is pushing an "AI-first" defense strategy via Project Replicator, deploying thousands of autonomous systems. China’s "AI+ Initiative" integrates AI into industries, with models like DeepSeek-R1 achieving top results using fewer resources. Singapore’s "NAIS 2.0" is tripling AI practitioners to 15,000, while India’s "IndiaAI Mission" deploys 38,000 GPUs and multilingual tools like Bhashini for public services. Across sectors, the ROI impact is real: In healthcare, Carle Health’s AI reminders hit 87% response rates, and Insilico Medicine advances drug candidates to trials in just 30 months. Finance sees BloombergGPT outperforming by 25-30 points on sector tasks. Manufacturing leaders like BMW use digital twins to cut maintenance by 25%, and agriculture benefits from John Deere’s autonomous tractors with 95% seed accuracy. Even humanitarian efforts, like the UN’s PulseSatellite for disasters, show AI protecting ecosystems. The AI Bubble talk? It misses the use case maturity: "Pacesetter" organizations report 67% gross margin boosts, and governance frameworks like the "EU AI Act and ISO standards" are enabling responsible scaling — with the AI governance market reaching $1.3 billion by 2026 at 47% CAGR. This isn't speculation; it's Applied-AI becoming the backbone of economies. My latest post on the Business Technology Roundtable includes the full story. https://lnkd.in/gmrsHJJy #AppliedAI #DigitalTransformation #Infrastructure #Investment

2024 was an important year for AI. Over the past year, I’ve followed the trends closely—reading hundreds of research papers, engaging in conversations with industry leaders across sectors, and writing extensively about the advancements in AI. As the year comes to an end, I want to highlight the most significant developments and share my views on what they mean for the future of AI. Generative AI continued to lead the field. Tools like OpenAI’s ChatGPT and Google’s Gemini introduced improvements like memory and multimodal capabilities. These features extended their usefulness, but they also revealed limitations. While impactful, generative AI remains just one piece of a larger shift toward more specialized and context-aware AI systems. Apple Intelligence stood out as one of the most impactful moves in this space. By embedding generative AI into devices like iPhones and MacBooks, Apple showed how AI can blend seamlessly into everyday life. Instead of relying on standalone tools, millions of users could now access AI as part of the systems they already use. This wasn’t the most advanced AI, but it was a great example of making AI practical and accessible. Scientific AI delivered some of the most meaningful progress this year. DeepMind’s AlphaFold 3 predicted interactions between proteins, DNA, and RNA, advancing biology and medicine. Similarly, BrainGPT, published in Nature, outperformed human researchers in neuroscience predictions, accelerating complex discoveries. AI models using graph-based representations of molecular structures revolutionized the exploration of proteins and materials, enabling faster breakthroughs. Another notable development was AlphaMissense, which classified mutations, helping with genetic diseases. These achievements highlighted AI’s effectiveness in solving critical scientific challenges. Hardware advancements quietly drove much of AI’s progress. NVIDIA’s DGX H200 supercomputer reduced training times for large-scale models. Meanwhile, innovations like Groq’s ultra-low-latency hardware supported real-time applications such as autonomous vehicles. Collectively, these advancements formed the backbone of this year’s AI breakthroughs. In my view, here is what we should expect in 2025: 1. Specialized AI models: I expect more tools tailored to specific industries like healthcare, climate science, and engineering, solving problems with greater precision. 2. Human-AI collaboration: AI will evolve from being just a tool to becoming a partner in decision-making and creative processes. 3. Quantum-AI integration: Maybe not in 2025, but combining quantum computing and AI could unlock entirely new possibilities. 2024 showcased AI’s immense potential alongside its limitations.But perhaps most importantly, AI entered everyday conversations—from TikTok videos to debates on ethics—bringing public attention to its possibilities and risks. As we move into 2025, the focus must shift to real-world impact—where AI’s true power lies.

AI is changing faster than ever. Every few months, new frameworks, models, and standards redefine how we build, scale, and reason with intelligence. In 2025, understanding the language of AI is no longer optional — it’s how you stay relevant. Here’s a structured breakdown of the terms shaping the next phase of AI systems, products, and research. 𝗖𝗼𝗿𝗲 𝗔𝗜 𝗖𝗼𝗻𝗰𝗲𝗽𝘁𝘀 AI still begins with its fundamentals. 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝘁𝗲𝗮𝗰𝗵𝗲𝘀 systems to learn from data. Deep Learning enables that learning through neural networks. Supervised and Unsupervised Learning determine whether AI learns with or without labeled data, while Reinforcement Learning adds feedback through rewards and penalties. And at the edge of ambition sits AGI — Artificial General Intelligence — where machines start reasoning like humans. These are not just definitions. They form the mental model for how all intelligence is built. 𝗔𝗜 𝗠𝗼𝗱𝗲𝗹 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 Once the foundation is set, development begins. Fine-tuning reshapes pre-trained models for specific domains. Prompt Engineering optimizes inputs for better outcomes. Concepts like Tokenization, Parameters, Weights, and Embeddings describe how models represent and adjust information. Quantization makes them smaller and faster, while high-quality Training Data makes them useful and trustworthy. 𝗔𝗜 𝗧𝗼𝗼𝗹𝘀 𝗮𝗻𝗱 𝗜𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 Modern AI depends on a specialized computing stack. GPUs and TPUs provide the horsepower. Transformers remain the dominant architecture. New standards like MCP — the Model Context Protocol — are emerging to help models, agents, and data talk to each other seamlessly. And APIs continue to make AI accessible from anywhere, turning isolated intelligence into connected ecosystems. 𝗔𝗜 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗲𝘀 𝗮𝗻𝗱 𝗙𝘂𝗻𝗰𝘁𝗶𝗼𝗻𝘀 How does AI actually think and respond? Concepts like RAG (Retrieval-Augmented Generation) merge search and reasoning. CoT (Chain of Thought) simulates human-like logical steps. Inference defines how models generate responses, while Context Window sets the limits of what AI can remember. 𝗔𝗜 𝗘𝘁𝗵𝗶𝗰𝘀 𝗮𝗻𝗱 𝗦𝗮𝗳𝗲𝘁𝘆 As capabilities grow, so does the need for alignment. AI Alignment ensures systems reflect human intent. Bias and Privacy protection build trust. Regulation and governance ensure responsible adoption across industries. And behind it all, the quality and transparency of Training Data continue to define fairness. 𝗦𝗽𝗲𝗰𝗶𝗮𝗹𝗶𝘇𝗲𝗱 𝗔𝗜 𝗔𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 The boundaries between science fiction and software continue to blur. Computer Vision and NLP are powering new interfaces. Chatbots and Generative AI have redefined how we interact and create. And newer ideas like Vibe Coding and AI Agents hint at a future where AI doesn’t just assist — it autonomously builds, executes, and learns. Understanding them deeply will shape how we design, deploy, and scale the intelligence of tomorrow.

NVIDIA had an interesting announcement this month regarding its new “Cosmo” AI model, which builds an internal representation of the physical world in order to optimize outputs. This shift from analysis to simulation is powerful, as we have also seen with Physics-Informed Neural Networks (PINNs), which our portfolio company Basetwo AI uses to revolutionize process manufacturing. Why does this matter? Because world models, PINNs, and similar AI architectures represent a paradigm shift in AI, moving from passive pattern recognition to active reasoning and decision-making. While challenges remain, these models could be a huge step forward in making AI more adaptable, explainable, and effective in applications that impact the real world. For business leaders, understanding this shift is crucial, not just to leverage AI’s potential, but to prepare for a future where machines do more than just analyze data. You can read more about it in my latest AI Atlas:

“Right now, it feels as if Big Tech is throwing general-purpose A.I. spaghetti at the wall and hoping that nothing truly terrible sticks.” — 𝘛𝘩𝘦 𝘕𝘦𝘸 𝘠𝘰𝘳𝘬 𝘛𝘪𝘮𝘦𝘴, 𝘖𝘤𝘵𝘰𝘣𝘦𝘳 2025 The quote captures a growing unease in the industry. The obsession with Artificial General Intelligence (A.G.I.) has turned into a winner-takes-all race — driven by a handful of firms, enormous compute budgets, and vague promises of human-level reasoning. But while that race dominates headlines, the real opportunity may lie elsewhere: in 𝗮𝗽𝗽𝗹𝗶𝗲𝗱 𝗔𝗜 𝘀𝗼𝗹𝘃𝗶𝗻𝗴 𝘀𝗽𝗲𝗰𝗶𝗳𝗶𝗰, 𝗯𝗼𝘂𝗻𝗱𝗲𝗱 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 — from industrial optimization to grid management to predictive maintenance. The NYT offered a striking example: a simple chess program built for the Atari 2600 in the 1970s recently outperformed a modern large language model when both were asked to play chess. A reminder that intelligence is not a function of model size, but of purpose and design. A.G.I. may one day arrive. 𝗕𝘂𝘁 𝗳𝗼𝗿 𝗻𝗼𝘄, 𝗮𝗽𝗽𝗹𝗶𝗲𝗱 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗯𝗲𝗮𝘁𝘀 𝗮𝗿𝘁𝗶𝗳𝗶𝗰𝗶𝗮𝗹 𝗴𝗲𝗻𝗲𝗿𝗮𝗹𝗶𝘁𝘆 — especially in Europe’s industrial core, where the real value of AI will be measured not in parameters, but in megawatts saved, downtime reduced, and systems made more resilient. https://lnkd.in/dRhBhc5e

I recently joined the Mixture of Experts podcast to explore four recent major developments shaping the future of AI—from neuroscience and autonomous driving to generative media and legal precedent. It was a great and engaging discussion with Gabe Goodhart, Ann Funai, and our amazing host Tim Hwang! 1. ChatGPT and the Brain A new neuroscience study shows that using ChatGPT reduces activation in the brain’s problem-solving network while increasing passive reflection. Users write faster, but retain less and think less deeply. https://lnkd.in/ef3R4c4W ⁉️  Are we outsourcing our thinking, weakening our brain musle—or augmenting it? 2. Waymo’s Shift Toward Human-Like Driving Waymo’s autonomous vehicles are now more assertive on the road, mimicking human behavior to better blend into traffic. This raises new design challenges for AI operating in social systems. https://lnkd.in/eJtH_K2B ⁉️ Should we make AI behave more like us—even when we're flawed? 3. AI-Generated Advertising at Scale Kalshi’s fully AI-generated NBA Finals ad reached 18 million viewers in 48 hours—produced in days, not months, and at a fraction of the cost. This could reshape how we tell stories and capture attention. https://lnkd.in/esDNU-h2 ⁉️ If anyone can generate media instantly, what becomes of originality? 4. Legal Rulings on AI Training Data A federal judge ruled that training AI on lawfully digitized books is fair use, but relying on pirated data is not. The line between innovation and infringement is now clearer—and more consequential. ⁉️ In the race to scale AI, are we building on ethical foundations—or shortcuts? Thanks to the #MoE team for the opportunity to unpack these big questions. The future of AI isn’t just technical—it’s deeply human. https://lnkd.in/eAQhGsKV Tune in on your preferred platform to listen or watch the latest episode, episode 61: Your brain on ChatGPT, human-like AI for safer AVs, and AI-generated ads  (video) → YouTube https://ibm.biz/Bdn5kM  (audio) → Casted  https://ibm.biz/Bdn5kS  (audio) →  Apple Podcasts https://ibm.biz/Bdn5kv  (audio) → Spotify https://ibm.biz/Bdn5km #AI #LLMs #ChatGPT #Waymo #GenerativeAI #AIethics #DataGovernance #CognitiveTech #MoEpodcast

Our new article, “Computation and Machine Learning for Materials: Past, Present, and Future Perspectives,” is now published in the MRS Bulletin 50th Anniversary collection (https://lnkd.in/g3dJ-gSZ). In this invited perspective, Sobin Alosious, Ph.D., Meng Jiang, and I discuss how computational methods and machine learning (ML) are transforming the way materials are discovered, designed, and optimized. The integration of computation, AI, and automation is reshaping materials science—paving the way for faster, smarter, and more interpretable discoveries that bridge simulations and experiments. 🔍 Highlights: - Tracing the evolution from density functional theory (DFT) and molecular dynamics (MD) to data-driven discovery. - Reviewing how machine learning, active learning, and generative models accelerate materials screening and inverse design. - Exploring emerging trends such as multimodal foundation models, self-driving laboratories, and quantum computing for next-generation materials innovation. #MachineLearning #MaterialsScience #ComputationalMaterials #ArtificialIntelligence #DataDrivenDiscovery #MRSBulletin

One of the most important developments in applied AI is the acceleration of AI-driven simulation in high-stakes, real-time decision environments. Across F1, markets and enterprise operations, the opportunity is no longer prediction alone, but how systems simulate scenarios, manage uncertainty and support decisions under extreme constraints and time pressure. What’s increasingly clear is how mature these systems have become: real-time, decision-centric, and governed to inform consequential outcomes. The shift is unmistakable: from analytics that explain the past to simulation systems that help navigate uncertainty in the present.

I'm excited to share my latest research paper on the potential of AI-driven debates as a tool for advancing understanding and decision-making across a wide range of domains, from philosophy and theology to policy and governance. By leveraging recent advances in machine learning, natural language processing, and multi-agent systems, we propose a novel approach to idea exploration and argumentation that could transform the way we engage with complex issues and make important decisions as a society. Our methodology involves training AI models to represent different stakeholder perspectives, engaging them in structured, iterative debates, and using the outputs as "synthetic data" to train higher-level models for decision-making and policy analysis. The potential applications are vast, from stress-testing policy proposals and building consensus on contentious issues, to exploring scientific hypotheses and advancing philosophical and ethical understanding. While there are important challenges and considerations to keep in mind, such as ensuring fairness, transparency, and accountability, we believe this approach has the potential to create a future in which our most important decisions are informed by rigorous, data-driven debates that draw on the best available evidence and reasoning. By embracing this bold and ambitious vision, we can harness the power of AI to help us navigate the complex challenges we face as a society and unlock new frontiers of human knowledge and understanding. Read the full paper to learn more about this exciting new direction in AI research and its implications for the future of decision-making and problem-solving. #AIDebates #MachineLearning #DecisionMaking #MultiFaith #Argumentation #Reasoning #SyntheticData #MultiAgentSystems #FutureOfAI #KnowledgeAdvancement #ComplexProblemSolving

The future of AI is shifting from resource-intensive data centers to potentially running anywhere, thanks to breakthroughs in inference optimization - the process where AI models apply their training to solve problems. Just as a professor doesn't need to repeat years of study to answer each student question, AI systems can be streamlined to run efficiently on existing hardware infrastructure. Through techniques like quantization and knowledge distillation, powerful AI capabilities could become accessible to small businesses, schools, and individual developers without requiring massive investments. This democratization of AI isn't just a technical achievement; it represents a fundamental shift in how we can deploy and use AI technology responsibly.