| Submit your comments on this article |
| Science & Technology |
| Army AI Task Force Works to Massively "Speed Up" Weapons Attacks |
| 2019-10-26 |
| [DefenseMaven] |
| Posted by:Skidmark |
| #9 #8 The neural net learned to identify goats in the Middle East. Posted by Clyde Dribble8052 And still the goats laugh at us while staying just out of range. |
| Posted by: jpal 2019-10-26 15:14 |
| #8 The neural net learned to identify goats in the Middle East. |
| Posted by: Clyde Dribble8052 2019-10-26 15:04 |
| #7 Yes, It's extremely hard to tell what your AI neural net has actually learned... |
| Posted by: Bright Pebbles 2019-10-26 14:18 |
| #6 Heard a funny anecdote from the early days of neural networks. The task was telling trucks from tanks in photos, something the military was quite interested in. The neural net showed astounding accuracy and everyone was giddy with success... until someone noticed all the tank pix were taken in the shade and the trucks were in sunlight. Bottom line, the network had learned to tell light pictures from dark pictures and was basically a rather baroque light meter. Choose your datasets wisely, O my brothers! |
| Posted by: SteveS 2019-10-26 12:57 |
| #5 The Air Force, meanwhile is doing this: USAF arms B1-B with long range mach 5 weapons |
| Posted by: Slats Flaith8307 2019-10-26 12:28 |
| #4 ^ But with massive amounts of data, they'll learn. Fast. |
| Posted by: Lex 2019-10-26 11:18 |
| #3 All they're doing is pattern recognition. It will still be less sophisticated than the vision of most land animals. |
| Posted by: Rob Crawford 2019-10-26 10:19 |
| #2 Prompt Global Strike. |
| Posted by: M. Murcek 2019-10-26 09:33 |
| #1 Skynet smiles. |
| Posted by: g(r)omgoru 2019-10-26 09:05 |