by Csongor Fekete | Sep 25, 2025 | AI, Business, Machine Learning
A new development from MIT researchers highlights how custom tools can significantly improve the performance of generative AI models in creating groundbreaking new materials. The tool, called Memory-Assisted Reinforcement Learning (MARL), provides a novel methodology that combines generative design with machine learning to propose only high-potential material candidates, bypassing less promising ones.
Traditional generative AI models typically suffer from a repetitive loop — generating material structures that resemble past successes but rarely achieve true innovation. MARL addresses this by learning from both successes and failures, enabling AI systems to avoid previously unproductive paths. This leads to a more efficient discovery process, drastically increasing the probability of finding high-performance, novel materials.
For businesses leveraging AI, this is a powerful case study in the value of custom AI models and domain-specific ML integrations. While MIT applies the concept to advanced materials science, the underlying approach can be reframed for martech and CRM systems. For instance, a similar strategy can enhance campaign creation in marketing operations by guiding models away from ineffective approaches based on feedback histories — improving customer satisfaction, conversion rates, and campaign efficiency.
A Holistic martech strategy embedding such a Machine Learning model could reduce wasted ad spend, accelerate personalization, and deliver higher ROI. AI consultancy teams and AI experts in forward-thinking CRM and AI agency ecosystems should explore how reinforcement learning can transform not only material innovation but also customer behavior modeling, content generation, and decision-making in digital marketing.
Original article: https://news.google.com/rss/articles/CBMipgFBVV95cUxQb1htMWJMSC12RXYzTm9fWnVHZmN3Zy1VbmRiekFoQm80Rl9SSDdxT1ZDenVBc1NaYmxwenBWMXFRVXJwTFBKYXhlbW11MFBoalgyVHpUMmUtem95TFRvUlJVdHdCSWNzWmtiOWlpM1VnSTdBSGdDMWoxa18zejQwYTlSckc1OU5RclZjNFI3MDlMTTF4UE1IbXhUR19DTWVLbzZKZ2xR?oc=5
by Csongor Fekete | Sep 25, 2025 | AI, Business, Machine Learning
China-based AI company DeepSeek has made headlines by announcing that its latest large language model (LLM), DeepSeek-V2, was trained at a remarkably low cost of just $294,000. This undercuts industry norms significantly, where similar models often require tens of millions of dollars in cloud compute resources and data engineering efforts.
DeepSeek-V2 is a 236-billion-parameter Mixture-of-Experts (MoE) model. MoE architectures intelligently switch between specialized parts of the model depending on the input, allowing fewer parameters to be active per task and, consequently, reducing training compute costs. In comparison, industry leaders such as OpenAI, Google, and Anthropic spend exponentially more to achieve equivalent or marginally better performance levels. DeepSeek’s cost-efficiency challenges the myth that state-of-the-art AI is limited to organizations with nearly unlimited budgets.
From a business and martech perspective, this revelation unlocks promising opportunities. Mid-sized enterprises and innovative startups can now explore building custom AI models or deploying fine-tuned variants aligned to unique business needs — from personalized marketing campaigns to intelligent customer service automation — without incurring traditionally prohibitive costs.
A key takeaway for HolistiCrm and other AI consultancy or AI agency players is the importance of leveraging efficient architectures like Mixture-of-Experts to democratize access to AI. This not only improves marketing performance but also enhances customer satisfaction through tailored interactions, while maintaining cost-effectiveness. Businesses want highly specialized Machine Learning models trained on their domain-specific data. With advancements like DeepSeek-V2, delivering such solutions becomes increasingly viable and economically rational.
As AI becomes more accessible, embracing holistic AI strategies that coordinate marketing, data infrastructure, and custom modeling can deliver significant return on investment — both in cost savings and enhanced customer value.
Read the original article: https://news.google.com/rss/articles/CBMirAFBVV95cUxNMkZhUFVGdDNURURHa3I5UlNySUJnamtFNWc4U29VaUhNTzNBRExHQ244cWJrUDNfTWJHeEZNanViRkJLanpGNWRMSDJYWG1ab244c0lnWENVN3NwX2pxaGo1NWhKMDNRcmRlbUFhbFNjWEZUb04zZHRRd0xFUGthZnltOWk3X09UaXBYVVY4bksxZ1BOcUNiXzUxN19lZmRLUzhhakdocjNFay13?oc=5
by Csongor Fekete | Sep 24, 2025 | AI, Business, Machine Learning
The accelerating fusion of AI and SEO is reshaping digital marketing strategies, and 2025 promises a major inflection point. In “AI & SEO: How to Prepare in 2025” from Exploding Topics, several critical shifts are highlighted that marketers, martech professionals, and AI experts cannot afford to ignore.
The article outlines how Google’s Search Generative Experience (SGE) is fundamentally changing how content is ranked, prioritizing AI summaries and drastically reducing organic traffic for traditional SEO tactics. Additionally, SEO is becoming more holistic—less about keywords and backlinks, and more about authority, context, and trustworthiness. The shift toward zero-click searches and AI-curated summaries means that brands must pivot from pure visibility to experience-based, value-driven interactions.
From a business value perspective, companies can future-proof their marketing by deploying Machine Learning models that optimize content performance in real time, adapt dynamically to shifts in ranking signals, and personalize user search experiences. Customized AI models, designed by an AI consultancy or AI agency, can analyze user behavior, intent signals, and topical authority to enable brands to maintain prominence in AI-driven search environments.
A real-world use-case could involve a retail brand leveraging a custom AI model to automatically adjust content strategies based on SGE behavior. This system could analyze how generative search displays their products and then recommend optimizations across product pages, blog content, and customer reviews—improving visibility and customer satisfaction simultaneously.
The convergence of AI and SEO demands not only technical strategy but also a holistic business approach. Organizations investing in tailored AI solutions for martech will gain a competitive edge in performance, user trust, and long-term customer loyalty.
Original article: https://news.google.com/rss/articles/CBMiVEFVX3lxTFBFQXBRXzBBc1pnVzhoMHNtYTJ0VXFXY01BR3p1R1BHRUVLU0tSdVoxUFFXNjUyNnlURkhsX1ZidHJXTXZ3cUxvNmhPSERmRDZoUGg0NQ?oc=5
by Csongor Fekete | Sep 24, 2025 | AI, Business, Machine Learning
China’s leading open-source large language model, DeepSeek, has been unveiled in a groundbreaking research paper that provides rare transparency into the inner workings of advanced AI systems. Unlike many Western models, DeepSeek’s developers from Shanghai-based startup DeepSeek-Vincilab laid out architectural details, training strategies, and performance benchmarks in full—a move that adds depth and dimension to the global LLM landscape.
The paper outlines DeepSeek’s hybrid architecture that blends transformer networks and retrieval-augmented generation (RAG) techniques, enabling higher factual accuracy and coherence in generated content. Trained on a mix of Chinese and English datasets, the model’s multilingual capacity demonstrates China’s growing investment in AI self-sufficiency. Moreover, DeepSeek is optimized for efficiency, using a "Mixture of Experts" layer that selectively activates parts of the model—reducing energy use without sacrificing quality.
For AI experts and martech-driven businesses, this transparency is a valuable resource for benchmarking state-of-the-art architectures and adapting them to build custom AI models. Businesses using holistic approaches to AI deployment can strategically evaluate how advanced LLM designs like DeepSeek deliver on cost, performance, and adaptability.
A relevant use case includes upgrading existing marketing platforms with AI-powered content generation. By integrating similar architectures, a marketing team can support more accurate, language-diverse customer communication while lowering infrastructure costs. An AI agency or consultancy can further maximize business value by tailoring such machine learning models to specific industry needs—whether it's customer satisfaction analysis, lead scoring improvement, or localization across regions.
Deploying tailored models inspired by cutting-edge architectures like DeepSeek adds measurable value in performance, customer engagement, and martech ROI—especially when implemented through a Holistic framework of data strategy, customer experience goals, and AI scalability.
original article: https://news.google.com/rss/articles/CBMiqwFBVV95cUxPd1hKZF9WWVI5OGRUbHpHdk42cUJodzAtZ1lvQnhrY2dOZUM4WVhyS0I2aHhyUVdiVlNZNl9mOUkwTG4yY00wbHVEdUV6dlJGb2FwOEdna2FYV2F6UDN0Y0FPbFhyUVNNNTBPandTVWhTWU0wSnJ0NTFOOGxPMnRieDRPV25ZM19JT0M4TXh0SEoydW5hV2I0bHlqTzdXMWdjS2NzSVRtTndiMXc?oc=5
by Csongor Fekete | Sep 23, 2025 | AI, Business, Machine Learning
Ohio State University President Carter’s recent State of the University address highlights the university’s strategic focus on affordability, AI research, and workforce development. The address emphasized making education more accessible, while also investing in future-proof domains like artificial intelligence to strengthen the academic and economic future of both students and the broader community.
Key takeaways include:
- Deepened commitment to affordability through expanded student aid and cost control.
- Strategic prioritization of AI research to elevate Ohio State’s stature as a national leader in tech innovation.
- Workforce readiness and expanded career opportunities through interdisciplinary programs and industry alignment.
These priorities underscore the growing importance of integrating AI into higher education frameworks. For organizations in the martech and CRM industry, this shift represents a pivotal opportunity. When universities build AI capacity and produce skilled graduates with AI fluency, companies gain access to a stronger talent pipeline that can fuel innovation.
A potential use-case in this context is a collaboration between CRM platforms and academic institutions to develop custom AI models for analyzing student engagement and predicting enrollment trends. Such Machine Learning models enhance performance forecasting, allow more targeted marketing strategies, and ultimately improve customer satisfaction across educational services.
HolistiCrm’s holistic approach, as an AI consultancy and AI agency, aligns well with this evolving academic-business intersection. Encouraging partnerships that tap into university-level AI research fosters innovation while creating scalable business value.
Read the original article: President Carter focuses on affordability, AI research, expanding career opportunities in State of the University address
by Csongor Fekete | Sep 23, 2025 | AI, Business, Machine Learning
A groundbreaking development in AI has emerged with the creation of a new Machine Learning model capable of forecasting an individual’s lifetime risk of developing various diseases, as reported by The Economist. This medical AI model, trained on the anonymized data of over 450,000 patients, analyzes a decade of clinical records to predict future health outcomes. Unlike traditional diagnostic tools, this model can assess long-term risks across 13 diseases, including diabetes, lung cancer, and heart conditions, offering a proactive and holistic approach to health.
Key takeaways include the model's ability to:
- Detect early warning signs from patient history and health metrics.
- Predict disease risks years in advance, improving preventative care.
- Handle multiple diseases at once, addressing patient complexity rather than singular diagnoses.
This represents a significant shift from reactive to preventative healthcare, powered by custom AI models. From a martech and business perspective, this presents a compelling use-case for customer satisfaction and retention strategies in sectors like health insurance, telemedicine, and pharmaceutical marketing. AI agencies and AI consultancies supporting health-oriented clients can leverage similar custom Machine Learning models to enhance personalization, optimize outreach, and improve overall service performance.
For example, a healthcare CRM embedded with predictive analytics can alert users of critical health risks based on behavioral data, enabling timely, targeted interventions and automated marketing workflows. This not only drives better health outcomes but boosts brand loyalty and patient engagement—demonstrating how AI can transform raw data into business value with a holistic strategy.
original article: https://news.google.com/rss/articles/CBMizAFBVV95cUxOZWcwQ2JsYXM1OXNwV1hVZnF4dmg2U3NtNFFzUlNVb1UyWDlSSXVWSkRzM2VWNmxZUXlhdl9jLW4xVGtuRlMwQW5BVko1a201cVYxXzdZcHlySU5UUFFrZ2hGN3B1emhLOEJhQmJPa2tYaE9pbm5VMnRabk9hQ0swUEVXbE1LVHlCRnBHeS1rdWhZUk9fWEVPX3ltZDFEYTFxdXVjOEpwVFJNZWcySmlKblRib2VoRVFGbl80MkdJN1k3OWJxd0Y1bXhXRFI?oc=5
Recent Comments