Man and technology: AI from human cells and mind reading

Scientists from Indiana University have built a hybrid biocomputer in which they integrated electronics with human brain tissue grown in the laboratory. Researchers used living brain cells to build an artificial intelligence system capable of recognising different people’s voices with up to 78% accuracy. That achievement could usher in a new era of powerful “Brainware” computers and significantly reduce the modern AI system’s energy demand. Current AI hardware uses approximately 8 million W to power a neural network that the human brain could power with just 20 W.

A team from the University of Technology in Sydney has developed a cap that fits over the head and is used to read a person’s mind. In the study, participants silently read text passages while an electroencephalogram (EEG) recorded their brain’s electrical activity. An AI model called “DeWave” then decoded the thoughts, transcribing them with 40 to 60 per cent accuracy. Thanks to new technology, it is possible to translate thoughts into text non-invasive, thus helping people who cannot speak due to illness or injury and enabling people to interact directly with machines.

In the largest-ever study of deep neural networks trained to perform auditory tasks, MIT researchers found that these models exhibit an internal organisation similar to the human auditory cortex. Computational models that mimic the structure and function of the human auditory system can help experts design better hearing aids, cochlear implants and brain-machine interfaces.