Managing AI News

This data was originally featured in the March 19th, 2025 newsletter found here: INBOX INSIGHTS, March 19, 2025: Data Governance, Managing AI News

In this week’s Data Diaries, let’s talk about evaluating AI news. There’s a very nearly literal firehose of AI news spewing information at you every day. Since we last spoke a week ago, these things have happened:

• New Alibaba VACE video generation model announced
• New Baidu ERNIE 4.5 and X1 models
• New Cohere Command A model
• New DAPO algorithm
• New Google Gemini 2 Flash Image model
• New Google Gemini 2 upgrades – audio overviews, canvas, and enhanced Deep Research, with Veo support announced as coming soon
• New Google Gemma 3 models
• New IBM BeeAI agent framework
• New LG Exaone Deep models that beat Deepseek
• New MCPs for Claude that allow 3D asset generation in Blender
• New Microsoft Phi-4 Multimodal model
• New Mistral Small 3.1 Multimodal model
• New NotebookLM mind maps announced
• New NVIDIA hardware – DGX Spark (formerly Project Digits) – announced
• New Perplexity MCP API
• New Sesame CSM 1B speech generation model
• New SmolDocling OCR model

And this is just in the last 7 days.

Each of these is a pretty big deal in some fashion. Each of these contributes to our overall capabilities with generative AI in some way. But the question is, how? And what do you need to pay attention to?

Katie often asks me to “tell her when I should care”, and that’s a useful request from a stakeholder to filter down the news. So, in true data nerd fashion, let’s create a scoring rubric for AI news that can help us filter the firehose. We’ll call this the TellKatie™ rubric:

• Could Katie use it today? This is binary. Is it something that, with little to no preparation, Katie could use?
• Does Katie have a use case, as a strategic but non-technical user? This is also binary.
• Finally, how impactful is it to the work of Trust Insights? Something might not be valuable to Katie but might be valuable to customers and clients, because there are a huge number of different use cases. This is a sliding scale, like 1-5.

From there, we can think through more broadly how close a piece of news is to the user:

• An announcement with no product is unusable and furthest away
• An algorithm or paper on a methodology is very far away; the non-technical user will not be implementing DAPO or GRPO, for example
• A raw model requires lifting to get up and running, especially for a non-technical user
• A SaaS model requires much less lifting to get up and running; this is usually a menu item or a dropdown
• A fully baked product is the closest to the user; they just log in and start using it

In your own filtering of AI news, this is a handy rubric. How close is the news to the work you do? And this will vary by person. Katie’s not going to be hunting down iQuants of GGUF models. I am, because I have different use cases.

Once we score and sort the news, this is what we see:

From this, I should definitely let Katie know about the changes to Google Gemini, and how we might be using the new Gemma 3 model in production. I might make mention of the upcoming changes to NotebookLM, but that’s about it. The rest can wait until it’s closer to the user or more impactful to the operations of Trust Insights.

Consider adopting a similar practice. Any time you see a new piece of AI news, ask yourself whether you could use it, whether you have a use case for it, and how impactful the news will be to your work. You’ll find that a lot of the AI news is actually AI noise for you to screen out.

Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.