A recent medical trial involving AI therapy bots has produced encouraging results for patients experiencing anxiety, depression, and eating disorders.
Over 100 people spent in excess of 100,000 hours developing the system, which was trained through the use of specific data sets.
Although there are inherent risks, such as the bot failing to correctly identify signs of self-harm, the key to the AI’s success appears to be its training data.
It is crucial that this is both wide-ranging and based on evidence; AI that is trained on unproven or biased data could prove to be ineffective or even harmful.
The next step is for AI therapy bots to be trained on high-quality data sets, and for their efficacy to be evaluated by the FDA.