The key to AI is always context. The more it can know about you, the more it can predict about you. This helps it respond to what you intend to do, and draw the correct conclusion.
The constant risk is a question of bias. If a system is connected to Apple 24/7, there is always the likelihood of a "promoted tweet" or some statistical variation thereof slightly biases a result towards a profitable outcome for Apple.
It's fantastic that such progress has been made in AI, but without complete transparency we risk falling into the same traps as have happened with other forms of proprietary software.
Key to AI is clear. But, which is the challenge? Are we talking about predicting actually? Or is it rather about building a huge case-based reasoning system that can "replicate" known patterns?
Most recent research is focused on neural nets and deep-learning to that end, this answers the question wrt prediction, but this seems to be a partial solution.
6 comments
[ 3.6 ms ] story [ 22.9 ms ] threadThe constant risk is a question of bias. If a system is connected to Apple 24/7, there is always the likelihood of a "promoted tweet" or some statistical variation thereof slightly biases a result towards a profitable outcome for Apple.
It's fantastic that such progress has been made in AI, but without complete transparency we risk falling into the same traps as have happened with other forms of proprietary software.
Most recent research is focused on neural nets and deep-learning to that end, this answers the question wrt prediction, but this seems to be a partial solution.