Show Notes
👁️ How does artificial intelligence decide what to see? Your eyes can look directly at something without your brain ever noticing it. AI faces a similar problem. A camera may capture every pixel, but the system must still decide which parts of an image matter and which parts it can safely ignore. In this episode of A Beginner’s Guide to AI, we examine spatial attention in humans and visual attention in artificial intelligence. You will learn how the brain uses a mental spotlight, why seeing is not the same as noticing, and how attention mechanisms help computer vision systems process complex images.
We also investigate the limitations of AI attention. A model can identify the correct object for the wrong reason, use backgrounds as shortcuts, or create a convincing heatmap without truly understanding the scene.
🏥 Our central case study follows the collaboration between Google DeepMind and Moorfields Eye Hospital. Their medical AI system analysed three-dimensional OCT retinal scans, created detailed tissue maps, and recommended how urgently patients should be referred. It performed at a level comparable with leading specialists in a retrospective test. Then a different scanner caused its accuracy to fall dramatically. The anatomy had not changed. The machine’s view of it had.
🔍 Key highlights: How spatial attention filters human perception How AI decides where to look Spatial attention compared with self-attention Why vision transformers connect distant image regions The limitations of saliency maps and AI heatmaps How AI retinal scans can support medical specialists Why machine vision fails when devices or environments change How humans and AI can compensate for each other’s blind spots
📧💌📧 Tune in to get my thoughts and all episodes, and don’t forget to subscribe to our newsletter: beginnersguideto.ai 📧💌📧
Quotes from the Episode “Spatial attention begins with a simple problem: there is too much world and not enough brain.” “The anatomy had not changed. The machine’s view of it had.” “Every spotlight reveals something. Every spotlight also leaves something in the dark.”
About Dietmar Fischer Dietmar is a podcaster and digital marketer from Argo.berlin. If you want to get your AI or digital marketing moving, contact him at argoberlin.com Hosted on Acast. See acast.com/privacy for more information.
More from A Beginner's Guide to AI
AI agents are becoming more capable, more personal and much harder to ignore. They can use connected apps, prepare work in the background and act on your behalf. So why are Big Tech companies making them look like cheerful little creatures with giant eyes?
🤖 Why AI Transformation Needs Intentionality, Not More Technology
🤖 AI has moved beyond technical teams and innovation labs. At Berlin’s BIG BANG KI Festival, Dietmar Fischer found packed stages, long queues, and business professionals searching for practical ways to implement artificial intelligence.
🤖 AI fluency is becoming a core business skill. But being AI fluent means much more than knowing how to prompt ChatGPT. In this episode of Beginner's Guide to AI, Dietmar Fischer talks with entrepreneur and growth marketer Solomon Thimothy about the transition from using AI as a chatbot to using AI as part of the workforce.
⚡ AI is getting cheaper, faster, and more efficient. So why are data centres preparing to use far more electricity? In this episode of A Beginner’s Guide to AI, Professor GePhardT explores Jevons’ paradox and its uncomfortable lesson for artificial intelligence: using less energy per task does not guarantee that we use less energy overall.