Contemplating the potential for personal redundancy as a result of too much daydreaming, Dr Matthew Partridge risks another few hours musing over the benefits and limits of laboratory processes run entirely by super-efficient, highly focused AI and robotics...
Ever since a student of mine rigged up a robot arm she had built from a kit to threaten anyone that came into the lab with a 3D printed sword, I’ve wondered if I could be replaced by AI and robotics.
There is no doubt that within just a generation the number of things robots and AI can do has radically increased. Not to date myself too much but, when I was younger, I was excited when I could get ready for school with my yellow Sony tape deck playing the one tape I owned. This morning my son discovered that he can program his Google Home to start his day by playing any music he wants while also sending a recorded message to me ordering his breakfast, all before he gets out of bed. My son also discovered that if he ‘orders’ his breakfast, he’ll find his breakfast served at high speed from a disgruntled chef.
From a more lab-based perspective, the same has happened. When I first started in research, the idea of having the computer record solution temperatures or pH was practically science fiction. Now buying pH probes without an iOS app or integrated web interface is actually difficult. And from all those little sensors has come ‘adaptive’ experimental control systems and ‘intelligent monitoring’. The last gas flow control system I looked at had the kind of smart control that would normally take training an intern twiddling valves months to pick up.
To call some of those innovations in lab automation ‘AI’ would, I think, be charitable. Many are just complex versions of IF Temp = 30 THEN Fan = TURN. But they are getting ever smarter and being coupled to experimental design computer models or machine learning programs which do allow for AI to plan and then control the whole experiment!
Personally, I’m fine with allowing AI to suggest experimental points but I do have some reservations about allowing it control of my lasers. I wouldn’t last even a day on the ENCOM game grid (ED: for anyone born after the mid 1980s this is a reference to the movie ‘Tron’) and I don’t trust it to not use billboard mirrors to shoot down missiles (ED: okay I THINK he’s referencing ‘Spies Like Us’) or even using it to - (ED: nope no more obscure references for you, get to the point)…
Despite these highly accurate and very relevant warnings from movies researchers, just like my student, I have been jumping for the chance to plug all these adaptive AI lab systems into robotics. I am prepared to bet good money that within a page or two of this article is an advert for a robotic auto filling pipette or plate loading system.
One company I met at the last trade show I attended (hey remember when we could actually meet people in real life?) built robotic arms that could do essentially every single thing I had to do in my first job but without coffee breaks or griping about its boss spending all day golfing. But if my first job had hired the robot instead of me then who would have worked out how to make a self-checking reader (this reduced my workload by about 30%) or who would have discovered that one batch of antibody was weaker than it should have been when he spilt £4000 of it and we had to get an emergency batch sent out?
For now, I think I’m fairly safe. But if they ever make a robot that can daydream as much as I do, and with the same level of clumsiness, then my days in research are numbered and I’ll go sign up for a training course in robot repair.
Author: Dr Matthew Partridge is a researcher, cartoonist and writer who runs the outreach blog errantscience.com