Automation and experimental work

[safkjldfk]

With automata replacing everything in this world, including a lot of laboratory functions, would experimental researchers be obsolete as well, with computational researchers as the only ones remaining? Or am I looking at it too simplistically?

I prefer experimental work to computational, but I can be flexible enough.

 

[Vanadium 50]

How do you automate something that has never been done before?

 

[cgk]

There is also the issue that someone has to develop the automation systems. This is itself a major point in many areas of study (for example, look up "lab on a chip" in the context of chemistry, or computer algebra in physics or automatic theorem proving in math...). Automation and ready made tools are never a problem---they just allows you to focus on better problems on a higher level (e.g., no physicist has to waste time with manual numerical computations any more[1]--so they can do more important stuff). This applies to both experiment and theory.

[1] Being proficient at that was still an important factor for physicists until the 1960s!

 

[gravenewworld]

It's going to be harder IMO to automate physical science work, but automation absolutely will make many researchers obsolete in the future. Microfluidics is going replace tons of expensive lab equipment, the enormous amount of work hours to run mundane assays, drastically reduce the space needed and the number of people to do sample prep/runs. 50 years from now scientists are going to laugh their behinds off that people were crazy enough back in the old day to do something like laborious PCR pipetting when a microfluidics chip will get much faster and better results from a a biological sample lysate in 1/10th the time. Huge rooms for high throughput drug screen platforms? Laughable. They'll soon be doing single molecule testing on a specific drug target with microfluidics and will be able to screen millions of compounds in a fraction of the time it currently takes.

A physical science like chemistry will be harder to automate, but it WILL come, for example see this example of total synthesis utilizing flow chemistry techniques:

http://pipeline.corante.com/archives/2014/04/15/total_synthesis_in_flow.php

Imagine just needing 1 scientist to pack a few columns and initiate setup of a flow system to synthesize anything that comes to mind in 1 day instead of needing a team of 14 scientists each synthesizing an intermediate and putting them all together. This company in the UK-- Cyclofluidics--claims it can synthesize, purify and assay a compound in 90 minutes using flow chemistry. Then a computer plugs the structure and IC50 data into an alogorithm and chooses the next compound to synthesize, and the process repeats over and over. It's plug and play SAR; all the chemist needs to do is keep the solvent reservoirs full and the waste barrels empty.

http://www.cyclofluidic.co.uk/en-gb/

I'm sure it's Cyclofluidics technology doesn't work as well as they claim, but this is type of automation the future is going to bring which is going to make many scientists and their skills obsolete.

Data analysts will be in higher demand since they'll be needed to sift through the massive amounts of data that next gen higher throughput techniques are going to generate. Software that is being developed, however, is going to do the data analysis for you, but you still need someone at the end to interpret the results.

 

[f95toli]

This is missing the point a bit. Today much of the actual "hands-on" experimental work in the lab is all about automating measurements (e.g. writing Labview programs). Hardly anyone makes "manual" measurements nowadays, most measurements involve computer controlled setups where the data is automatically acquired and saved to a computer HD. Hence, nowing how to design, implement and control such setups is one of the most important skills you can have as an experimentalist.

One consequence of this is the amount of data acquired has increased a LOT; it is not at all unusal for even rather simple measurements to generate gigabytes of data. This i partly because it is easy to do once everything is working and the measurement is automated (it is always better to have too much data), but also because many measurements involve "needle in a haystack" searches for very subtle effects and you don't want to risk missing something just because you couldn't be bothered to take enough data.

 

[gravenewworld]

This is missing the point a bit. Today much of the actual "hands-on" experimental work in the lab is all about automating measurements (e.g. writing Labview programs). Hardly anyone makes "manual" measurements nowadays, most measurements involve computer controlled setups where the data is automatically acquired and saved to a computer HD. Hence, nowing how to design, implement and control such setups is one of the most important skills you can have as an experimentalist.

One consequence of this is the amount of data acquired has increased a LOT; it is not at all unusal for even rather simple measurements to generate gigabytes of data. This i partly because it is easy to do once everything is working and the measurement is automated (it is always better to have too much data), but also because many measurements involve "needle in a haystack" searches for very subtle effects and you don't want to risk missing something just because you couldn't be bothered to take enough data.

Maybe in physics, but you'd be surprised how prevalent something like the Western Blot still is in biomedical science. The high throughput that is coming to the biomedical science arena is absolutely, without question going to make many biomedical scientists unemployed. Western blots will finally be relegated to the past. They have systems now where all you have to do is insert a microfluidic chip, basically just inject cell lysate, and walk away. What kind of expertise do you need for a setup like that? You could probably hire a lab technician with an associates degree to do that work and only 1 scientist to analyze the data instead of 10 scientists that would still have to do prep work for automation. Automation in the biomedical sciences still requires laborious sample prep, but once we can do away with much of the pre-experimental work, many scientists in the biology/medicine fields are going to find themselves unemployed. I don't think we'll see the majority of this impact during our careers, but 30 years from now, it will be completely different.

 

[AlephZero]

You could probably hire a lab technician with an associates degree to do that work and only 1 scientist to analyze the data instead of 10 scientists that would still have to do prep work for automation. Automation in the biomedical sciences still requires laborious sample prep, but once we can do away with much of the pre-experimental work, many scientists in the biology/medicine fields are going to find themselves unemployed.

I suspect you are being too shortsighted there. If you can do a procedure 10 times as fast, you don't need 10% of the staff. More likely you will want to do 100 times as much work, and need 10 times as many staff to do it.

Certainly in engineering, the key factor is how long it takes to complete an analysis cycle. If you need to get a product to market in 1 year and it takes 6 months to evaluate a proposed design, you only have two chances to get it right. if you can cut the 6 months down to 6 hours, you can look at 5 different options in a week.

That magnitude of speed up is realistic, and I haven't seen it cause much unemployment, except for the small minority who can't or won't adapt. Is the world overrun with unemployed women, now that millions of them don't work as copy typists or operate manual telephone exchanges any more? I don't think so.