A new startup company is trying to discover the hidden factors that make some scientific experiments succeed and others fail. For example, one lab succeeded in detecting the Covid virus, but another, using the same protocols, failed.
To determine how this happens, the company’s researchers are
closely monitoring scientists as they work. Cameras capture their every move,
and the scientists narrate their work into microphones. Artificial intelligence
adds to the narrative with explanations such as “The operator resuspends the
pellet by pipetting it up and down ten times.” The idea is to capture science as it happens in
such a way that artificial intelligence models can be trained to recognize
every object the scientists used and every action they performed.
When scientists carry out experiments, they keep careful
records in their lab notebooks. Subsequent researchers use those records to repeat
the experiment. Sometimes the repeated experiment doesn’t work. That’s because not
every detail got written down.
Scientists make thousands of minor decisions over the course
of a day, such as how they shake a test tube. Such decisions are based on “tacit
knowledge": the practical know-how, skills, and intuition that are difficult to
write down, explain, or teach to others. Some researchers consistently get
experiments to work, but they don’t know why. According to their colleagues,
they have magic hands.
In reviewing the lab scientists’ audio and video records, researchers
are finding that the workers are performing the same experiment in many
different ways. As one analyst noted, “The variation we see even among
well-trained scientists is pretty jaw-dropping.”
The researchers are going to have to sift through all this
information in hopes of finding the secrets of success. Who knows? Maybe AI can
figure it out.
For an introduction to this blog, see I Just Say No; for a list of blog topics, click the Topics tab.
