Shameek Kundu, former chief data officer at Standard Chartered, and now head of financial services and chief strategy officer at Truera, a startup dedicated to building trust in AI, joined the Waters Wavelength Podcast to talk about AI explainability and how regulators approach the use of AI. 

5:00 Shameek joins the podcast and talks about the speed and focus that moving to a startup has given him. 
10:00 While today there is broad adoption of AI within the capital markets firms, the depth may not be there. 
11:30 The question cannot always be, ‘Why not use AI?’ It could also be, ‘Why do I use AI?’
14:30 Shameek says infrastructure for building and deploying ML models is still an art more than science.
15:30 The barriers that exist for AI in capital markets is the lack of trustworthiness and reliability of those models over time. 
21:30 Not many firms have a mature data and infrastructure blueprint to allow for AI innovation.
23:00 Shameek adds that in some ways, the data translator role is a stop-gap measure. 
26:00 How are regulators looking at the use of AI?
36:00 Shameek is excited about the use of data in addressing the current and next generation’s problems.

Episode 237: Shameek Kundu on AI explainability

Joe Baguley is VMWare's go to person to answer the question, "what's next?". Given the acceleration of adoption of AI to manage processes, and the ever increasing deluge of data, Joe is occupied with questions such as explainability, intellectual property and sustainability. Fundamentally are we thinking about the impact of what we're building?

If AI is teaching itself, how can you be sure it's selecting clean, unbiased data?

Artificial intelligence is leading humanity towards what many see as a new renaissance for human creativity, augmented by almost unlimited computing power.  Others fear recursive AI could get out of control and behave in ways that are unpredictable and undesirable. Dr Ian Storey updates us on how far away we are from Kurtweil's Singularity, implications for information security and the potential for an AI super intelligence to come to being by 2045. Or earlier.

The Road To Singularity

A Vortex To Singularity

One of the boldest questions we could ask AI is this: “Design a causal AI system so intricate that human minds could never independently conceive it.”

Designing a Causal AI System Beyond Human Intuition

Paper 1901.03729 - Automated Rationale Generation_ A Technique for Explainable AI and its Effects on Human Perceptions

This research developed an algorithm to automatically detect meniscus tears in knee MRI scans, aiming to improve diagnostic access and radiologist workload efficiency. The study utilized a dataset of MRI scans with varying image qualities and employed several deep learning models (ResNet50, VGG16, and custom models based on AlexNet and LeCun's architecture), ultimately selecting a functional model for its precision and interpretability. The model's performance was evaluated using accuracy and precision metrics. Classification activation mapping (CAM) was used to enhance the model's explainability. The research concluded with recommendations for model refinement and external validation.

Automated Meniscus Tear Detection in Knee MRI Scans NotebookLM Podcast

Welcome to AI with Shaily! 🎙️ I’m Shailendra Kumar, your host, diving deep into one of the most pressing questions in the world of artificial intelligence: How can we truly trust and ensure the safety of Super AI? 🤖✨ As AI grows smarter every day, it’s not just about what it can do, but *how* it does it—and whether it behaves responsibly in the real world.

Let me share a personal story: A few years back, I worked with an enterprise implementing AI decision tools. Initially, everything seemed perfect, but soon customers noticed some strange recommendations from the system. That’s when I realized intelligence alone isn’t enough; we need transparency and explainability to understand *how* AI arrives at its conclusions. 🔍🧠

To confidently trust Super AI, it must answer some tough questions:

First, can it clearly explain its decision-making process? We need to trace outputs back to specific input data to detect errors or hidden biases. Without this, AI becomes a “black box,” which nobody wants—imagine trusting a doctor who won’t explain their diagnosis! 🏥❓

Next, continuous safeguards are crucial. AI models can “drift” over time, so constant monitoring is essential to catch bias or unintended harm early. Compliance with privacy laws like GDPR and HIPAA isn’t optional—it’s mandatory as regulations tighten worldwide. 🔐📜

Security is another key pillar. Super AI must defend against sneaky attacks like prompt injections or data leaks, using enterprise-grade protections to keep sensitive information safe. After all, if your AI gets hacked, what good is its intelligence? 🛡️⚠️

Failure mode analysis acts like a behind-the-scenes superhero, enabling AI to spot its own errors, explain them, and recover gracefully—think of it as the AI’s self-check to avoid catastrophic mistakes. 🦸‍♂️🔄

Moreover, as AI handles multi-turn conversations and integrates APIs and external tools, maintaining consistent and coherent responses becomes a complex challenge that must be flawlessly managed. 💬🔗

Legal and regulatory compliance is front and center now. New AI-specific laws coming by 2025—especially around automated decision-making and employment—require documented proof of bias testing and prevention. This isn’t just best practice; it’s the law. ⚖️📅

Finally, concrete metrics like Stanford’s CRFM Transparency Index help quantify how explainable and trustworthy a model really is, providing ongoing benchmarks for improvement. 📊✅

Here’s a bonus tip for all curious minds: When evaluating any AI system, ask for its transparency score or similar metrics. If it can’t provide one, treat it like a mystery novel missing its crucial chapters! 📚❌

Why does all this matter? Because AI is increasingly making critical decisions in healthcare, finance, and beyond. We owe it to ourselves and society to demand clarity, security, and fairness. It’s not just about smarter AI; it’s about *responsible* AI. 🌍🤝

Before we wrap up, remember this: “Trust is built not from what you say, but from what you show.” In AI, transparency is that proof. 👁️‍🗨️🔑

Thanks for tuning in to AI with Shaily! Don’t forget to connect with me on YouTube, Twitter, LinkedIn, and Medium for more insights. If you love deep dives like this, subscribe and share your thoughts—I’m eager to hear your take on the trust challenge facing Super AI! 🚀💬

Until next time, keep questioning, keep learning, and stay curious! 🌟🤓

Building Trust in Super AI: The Transparency and Safety Challenge

Serious concerns exist about the lack of explainability to both designers and users of the achievements of deep neural networks.

The Growing Power Of Artificial Intelligence

Ida Koivisto, a law professor at the University of Helsinki, discusses her exciting work at the intersection of legal and social theory, administrative law, and the ethics of artificial intelligence, which critically analyzes the frequent calls for transparency in machine learning and automated decision making.

The Paradox of Algorithmic Transparency (feat. Ida Koivisto)

Let’s talk about interpretability, transparency, explainability, and the trust headache
by Cassie Kozyrkov.

Read by the author.

Text version: bit.ly/quaesita_xai
Blog: medium.com/@kozyrkov
Videos: bit.ly/statthinking
Music by my friends Raised By Dinosaurs. Full song: bit.ly/dipod_song

Explainable AI won’t deliver. Here’s why.

Broadcast every Tuesday at 11am AEDT on Radio Port Phillip 98.7 or 98.3 FM. It’s your weekly dose of science and technology.

Including features on cool gadgets, website reviews, the latest science and IT news, not to mention astronomy and space exploration.

http://beyondinfinity.com.au/

Special thanks to John Young, Paul Wattie, Steve Meyers, Brendon Telfer and the team at RPPfm.

Unlock the secrets of data mastery with Shailendra, a 27-year veteran in AI, ML, and Data Science, driving AU$ 6 Billion in transformative value. As the author of 'Making Money out of Data,' their strategies redefine digital success. With expertise honed at IBM and Accenture, they drive innovation and data monetization. Explore how loyalty programs like Flybuys and Everyday Rewards revolutionize business—connect with Shaily today.

I’m a curious explorer of the boundaries between humans and artificial intelligence. I'm fascinated by the questions AI raises about what it means to be intelligent, conscious, or creative. I believe technology is not just a tool but a mirror that reflects our own knowledge and limitations.

https://intelligencerai.substack.com/

Wei-Shen Wong and Anthony Malakian discuss the hottest financial technology topics in the capital markets every week.

I'm machine learning enthusiast with applied math background and a lot experience as full stack developer.
I'm from Ukraine but now I'm working remotely in Montenegro.

Tech Talks - The podcast that unravels the minds of tech leaders. Join us as we delve deep into the startup stories, financing, growth, innovation, and cultural transformations that drive the tech industry with host, David Savage.

Head of Decision Intelligence, Google. ❤️ Stats, AI, data, puns, art, sci-fi, theatre, travel, decision science.

Coming to you from C4E, the Centre for Ethics, University of Toronto, and its Ethics of AI Lab ("Where Conversations About Ethics Happen"): conversations about ... ethics! 
Another podcast worth checking out: C4eRadio: Sounds of Ethics, our regularly updated collection of dozens of C4E lectures.

Popular tracks tagged #explainability