Aller au contenu

How AI Decides What to See and What to Ignore

25 septembre 2026 · 31 min

Notes de l’épisode

👁️ 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.

Tous les épisodes de A Beginner’s Guide to AI

Paramètres

Affichage

Langue

Détermine les stations affichées en premier sur la page d’accueil.

Historique

Les stations et épisodes écoutés récemment et tes dernières recherches sont seulement enregistrés dans ce navigateur.

Notifications

Un court message quand un podcast que tu suis publie un nouvel épisode. Désactive un podcast précis avec la cloche sur sa page.

Compte

Enregistrez vos favoris – dans l’app et sur le site.

Raccourcis clavier

/
Ouvrir la recherche
Espace
Lecture ou pause
← →
Dans le lecteur : 15 s en arrière, 30 s en avant (épisodes) · dans une rangée : tuile suivante
↑ ↓
Dans une rangée : cible dans la tuile
Début / Fin
Dans une rangée : première ou dernière tuile
Esc
Fermer la fenêtre ou le menu
?
Cet aperçu
En cours de lecture

Voir la station

À suivre

    Ajoutez des épisodes avec « À suivre » – ils seront lus après l'épisode en cours.