Scientists losing their jobs to automation?

Two articles have appeared in the popular science media in the last week talking about software that automatically made new scientific knowledge. In one case, the knowledge was completely new, while in the other case, the software autonomously learned something we already knew.

Here are the first and second articles.

It makes sense to me that this type of technology will do more over time. Science is 1% inspiration and 99% perspiration. That 99% often involves running large numbers of experiments, processing reams of data and designing slightly changed experiments based on the previous results. When the execution of the experiments can be automated, e.g., when a simulator is used, software is ideally suited to run the experiments.

Artificial intelligence is sometimes described as the science of search. Thus, the techniques developed in that field for 50 years, should be ideally suited to performing routine experiments. The design and processing of experiments and their results is essentially search over some complex, sometimes physical space.

But getting software to do that "1% inspiration" is likely a long way off. That leads to two possible outcomes: (a) Scientists starting actually working 1% of the time (about 20 minutes a week) or (b) Scientific discovery speeds up 100 fold. I know where my money is.