May 9 2018, 19:08 Group: Heroes, Master of Mutant Creation
This is great medical news (AI used to help field researchers with the interiors of cells), but it also has some potential for us in other fields.
So what kinds of sensors would require the use of fast computing?
Jun 28 2018, 22:36 Group: Grid Cop, 5éme Corbin
Real-time or near-real-time digital enhancement has lots of promise and it makes plenty of sense to apply AI to more complex images or where recognition has to be very accurate (false positives have bad consequences, are not checked by a skilled human, or there are so many potential hits that it's distracting rather than helpful). Quality of training will be important, as Deep Dream has illustrated.
If it's applied in augmented reality for a moving user particularly, where compute time has to beat human visual response time (luckily, that's relatively long), the AI will have to deliver solutions or guesses moderately fast - which has different requirements, in software and hardware optimisation and in physically carrying the required computer or remotely connecting to it, than crunching big datasets. An observer or sniper might tolerate a slight lag to identify objects of interest.
Quality of recognition might be emphasised for a sensor feed that's going to a user who might not be well able to discriminate between correct or spurious choices, such as a rehabilitation aid for multiple patients with visual or brain injury, a child, or a training device for an animal.