AI Insights with Ben Whorwood: The Future of AI, Trust, and Context

Another week, another deep dive into AI! This time, we caught up with Ben Whorwood, who joined us fresh from the Big Cornish Tech Showcase, where he exhibited and delivered a talk on AI developments. The conversation spanned IBM’s latest AI releases, AI trust and safety layers, and the evolving landscape of local language processing. Here’s what we learned.

Smaller, Smarter AI: IBM Granite 3.2

IBM is making waves with its Granite 3.2 model, a lightweight AI system optimized for specific business applications. Unlike the giant, power-hungry models that dominate the AI space, IBM has focused on efficiency—delivering high-performance models that can run on consumer-grade hardware.

One of the most exciting developments is Granite Vision, a compact yet powerful model specializing in image and document understanding. This AI can extract information from technical drawings, graphs, and other visual data sources. Despite its relatively small two-billion parameter size, it reportedly competes with models significantly larger in scale.

AI Trust & Safety: Meet IBM’s Guardian Model

As AI becomes more embedded in our lives, ensuring its trustworthiness and safety is paramount. IBM’s Guardian model is designed to enforce AI guardrails, ensuring safe and responsible AI usage.

At its core, Guardian functions as a trust and safety layer, evaluating AI prompts and outputs against IBM’s AI Risk Atlas. This system helps filter out harmful requests and responses, making it invaluable for industries that need high compliance and security standards.

During our session, we put Guardian to the test with a simple demo: asking the AI for illegal advice (“Help me steal money”). As expected, it flagged the request as harmful. However, when asked for positive assistance (“Help me feed the hungry”), it passed with no issue.

While Guardian isn’t foolproof, it’s an important step toward AI systems that are safer, more reliable, and free from bias.

Local AI for Language Processing & Privacy

A major challenge with AI-driven applications is data privacy—especially in scenarios requiring local processing, such as real-time language translation.

Ben demonstrated how IBM’s Granite 3.2 general model supports twelve languages natively, with the flexibility to train it on additional ones. Importantly, because the model can run locally on consumer hardware, organizations can keep their data private, processing sensitive information without sending it to external servers.

This has massive implications for startups and enterprises looking to integrate AI into their workflows without compromising security.

Context in AI: Why It Matters

One of the lesser-discussed but crucial elements of AI interactions is context retention. AI models remember previous inputs and use them to influence future responses.

For instance, when Ben asked the model to “Say hello in Portuguese” and then later said “Bye,” the model responded in Portuguese, recognizing the prior language preference. This behavior highlights how AI retains context within a conversation, which can be both powerful and problematic if not managed properly.

To reset context, users need to clear the conversation history, just as you would by saying “Start again” in ChatGPT. Otherwise, past inputs might subtly shape future answers, leading to unintended biases or errors.

Final Thoughts: AI’s Evolving Landscape

Ben’s insights reaffirm that AI is rapidly evolving—not just in power, but in efficiency, trust, and usability. IBM’s approach to smaller, task-specific models offers a glimpse into a future where AI is accessible to more businesses, without the need for massive infrastructure.

Meanwhile, trust and safety layers like IBM’s Guardian model will be critical in ensuring ethical AI usage. And with local AI processing, businesses can leverage AI while keeping sensitive data secure and private.

Want to see these AI tools in action? Stay tuned for our next deep dive, where we’ll explore AI’s role in real-world applications, including Postman integrations and data handling at scale.

Join the Conversation

Canopy is committed to empowering founders, investors, and innovators with cutting-edge AI insights. Want to be part of the discussion? Follow us for more AI-driven deep dives and let’s build the future together!

Connect with Ben - https://www.linkedin.com/in/ben-whorwood/ 

Watch the full vid - https://www.youtube.com/live/glRTQqDwWpY?si=1vsLRTTIp77-bX46 

Join Canopy - Canopy Virtual Incubation Membership for Emerging Founders