Researchers were able to decode patients' internal thoughts in real-time with an accuracy of up to 70 percent.