AI Can't Predict Black Swans—Here's How to Prepare

Direct answer: AI cannot predict a true black swan, because such an event sits outside the data and assumptions any model was built on.

And the loss ran at roughly $10 million per minute ($460 million ÷ 45 minutes). For each system that acts without a human approving each action, estimate the worst-case loss per minute and divide the loss you can absorb by it. Knight’s rate was about $10 million a minute.

Key takeaways

• Enforce limits outside the system.
• The SEC and Knight’s filing differ on the loss, and I report both.
• Do this next: Pick the one automated system in your organization with the largest worst-case loss rate.

Knight’s order, share, and loss figures come from the SEC’s announcement and Knight’s 10-K.

Read the full article: https://www.neuralgrimoire.com/ai-cant-predict-black-swans/