TECHNOLOGIE SPIEGEL OVER JE: HOE KAN KUNSTMATIGE INTELLIGENTIE VOORKOMEN DAT HET JE GEBEURT?
Experts trained a neural network (an ML system modeled by the brain) to predict the future. It was trained on hours of recorded speech to continuously analyze 2-second audio clips and predict what is likely to happen. This new technology uses what is called a ‘opposite approach’. It’s machine learning, which looks for patterns in data and adjusts sounds so that an AI can mistake them for something else. Essentially you use one AI to trick the other.
Even though the ASR framework was trained to transcribe speech interrupted by Neural Voice Camouflage, it still provided margin. While they were the easiest to interrupt, short words like “the” aren’t the most enlightening elements of a dialogue.
Chiquier says the predictive part of the program has great potential for other applications, such as driving autonomous cars. Forecasting is also an important part of brain function. You feel surprised when your mind guesses wrong. Chiquier says we replicate how people do things.
Andrew Owens says, “There’s something nice about the way predicting the future, which is a classic problem in ML, integrates with another problem in confrontational ML.” Bo Li was amazed at the new approach’s ability to beat the reinforced ASR systems.
Mia Chiquier, a Columbia University computer scientist who conducted the study, said this is just the first step toward protecting privacy in the face of AI. Artificial intelligence collects data about our voices, faces and actions. A new generation of technology is needed that respects our privacy.