I would like to fully agree (well almost) with David Zilli's suggestion that artificial intelligence needs to work smarter, not harder (Comment, April 2022). The most successful project on the planet must be the six-million-year-old development of a carbon-based computer, the human brain. In all projects, there is normally a lessons-learnt phase. Artificial neural networks stem from a concept formed over 60 years ago. Only recently, in around the past 15 years, has computing technology been able to make good use of the Al concept. Feeding a neural network lots more training data does not necessarily make it smarter. It will start to identify coincidences and draw parallels that are incorrect, or, worse, solutions will be misleading.
展开▼