AI is increasingly influencing decisions about military targets, access to public services and outcomes that can directly affect people’s lives.
Last week I wrote about Palantir’s expanding role in Britain’s digital state and the risk of vendor lock-in across critical systems, including the NHS and MoD.
This week Britain becomes the first foreign country to gain access to Ukraine’s Avengers Labs battlefield dataset.
It is built from millions of real combat observations used to train AI for targeting and defence. The MoD is reportedly particularly interested in the acoustic sensor data.
This is another major dataset feeding into the UK’s rapidly developing defence AI capability.
When AI starts influencing decisions with consequences this serious, what happens when someone says the system is wrong?
How do you challenge an AI-assisted decision if you can’t see how it reached its conclusion?
How do you prove an error when the institution relying on the system may understand it better than you do, or may not fully understand it either?
Who has the power to overturn the decision?
If the technology becomes too complex, too embedded or too dependent on outside suppliers to meaningfully audit or replace, control has already shifted.
AI capability is accelerating fast, but will accountability keep up?
Read the full article:
https://www.ft.com/content/063dbdd2-fb5a-4ece-9152-74e791e4835c
Last week I wrote about Palantir’s expanding role in Britain’s digital state and the risk of vendor lock-in across critical systems, including the NHS and MoD.
This week Britain becomes the first foreign country to gain access to Ukraine’s Avengers Labs battlefield dataset.
It is built from millions of real combat observations used to train AI for targeting and defence. The MoD is reportedly particularly interested in the acoustic sensor data.
This is another major dataset feeding into the UK’s rapidly developing defence AI capability.
When AI starts influencing decisions with consequences this serious, what happens when someone says the system is wrong?
How do you challenge an AI-assisted decision if you can’t see how it reached its conclusion?
How do you prove an error when the institution relying on the system may understand it better than you do, or may not fully understand it either?
Who has the power to overturn the decision?
If the technology becomes too complex, too embedded or too dependent on outside suppliers to meaningfully audit or replace, control has already shifted.
AI capability is accelerating fast, but will accountability keep up?
Read the full article:
https://www.ft.com/content/063dbdd2-fb5a-4ece-9152-74e791e4835c
14❤️3👍1🖤1
