As a kid I had problems with Foundation (Asimov) premise that loss of scientific knowledge can be the trigger not just the result of civilizational collapse - not anymore.
More pernicious than loss of knowledge imo is loss of trust into the scientific process. „Research“ on social media is seen as superior to expert consensus by a frightening number of people.
I have a PhD in Ecology and a BS in CS. I find the bifurcation portrayed here exaggerated. The best modern ecologists merge rigorous fieldwork with advanced modeling; we need to harness vast, underutilized datasets, not just generate new ones.
The 'computer scientist' quote illustrates a frustrating trend: tech-centric 'drive-bys' that lack the ecological context required for good science. On the flip side, the 'old guard' who ignore modern data assimilation are leaving massive potential on the table. The field is rightfully shifting from site-specific anecdotes to foundational, broad-scale work, but we need both skillsets to do it justice.
Ecology PhD turned data scientist, I was looking to respond and you summed up my thoughts really well!
I will add that funding can complicate things a bit, funding sources often get wowed by more "advanced" methods, while the underlying science might be less than stellar. There are important questions that can be answered by small, elegant field studies, and there are questions that require larger datasets and more computation. When we start putting the methodological cart before the scientific horse, that's where we run into problems.
I did marine biology field work almost 5 decades ago as a lowly junior lab tech. Work always has downsides, for me it was not really the Scots winter, cold feed, chapped hands, the land-rover having to reverse up steep icy roads to get back from the harbourside: it was washing the glassware and dealing with sodium hydroxide weighing (it absorbs moisture from the air so its a fools game). But, field work also brought amazing experiences, I visited the seaside 70+ times over a year, and got an insight into what a time series really means when you cover the tidal and weather and seasonal cycles.
It's also always error-prone. Nothing in the field is perfect. Reality is a bad approximation for your model at times, if you take a model centric view.
I would be immensely skeptical that field work is ever going away. There may be aspects of truth in this around cost of travel, risk, seniority.
Machine learning and data science are not new things in science. It's great that we have the ability to share and work with existing data sets, collect data remotely with sensors, and build software to create models, but we'll always need people to go out and collect updated data, place censors and verify that what models predict is actually happening.
> Scientists who run long-term ecological studies, in particular, report that they struggle to find funding.
It's cheaper and easier to do stuff sitting at a desk. In theory that's a good thing if it means more work gets done, but field work has to happen too. For many people it's the best part of the job, for others it's a pain that has to be suffered through to get the data they need. Hopefully there's room (and funding) for both kinds of people to do the work they want.
I am involved in both botanical field work and ML, unfortunately most of the data that I have gathered and analyzed in the last 12 years indicates how quickly many ecosystems are degrading. Often I wonder why I do the analysis, simply taking a photograph in the same spot ~7 years apart allows any average person to see that things are not on a positive trajectory.
Attempting to convince people to change course and focus on restoration has mostly been a losing battle, with much larger forces behind the main detriments that make local changes feel inadequate.
Working in geology, I find the opposite problem. Field work is so highly valued that we're at a place where we have so much data and not enough people really working and analyzing it. My general impression is that in some subfields work that's done exclusively using preexisting data is kind of looked down on. In my opinion tons and tons of money is essentially wasted collecting new data - and then it's poorly catalogued and hard to access. You typically have to email some author and hope they send you the data. People are fiercely protective of their data b/c it took a lot of effort to collect and they want credit and to be in on any derivative work (and not just a reference at the bottom of a paper)
I would say the main workflow is collect some new data nobody has collect before, look at it and see if it shows anything interesting, make up some interesting publishable interpretation.
It feels like it'd be smarter to start with working with existing data and publish that way. If you hit on some specific missing piece, go collect that data, and work from there. But the incentive structures aren't aligned with this
The AI angle is really shoehorned in, but irrelevant to the larger problem. Sure, it allows you to annotate more data. Obviously it's more fun to go do field work than count pollen grains under a microscope. If anything AI make it easier to do more fieldwork and collect even more data b/c now you can in-theory crunch it faster
There is still field work that happens. AI can not replace that. You'd need to literally simulate the whole world before AI can even get close to gather all data obtainable here. For instance, on a given ant hill: which plant species will be more prevalent there? (For those not knowing a lot about ants: some ant species carry specific plants or defend plants against invaders. The most usual example is for leaf-cutting ants, but there are many additional examples, and for various reasons you will also find different plant species to be more prevalent close to an ant hill in a forest area, than other plant species.) AI can steal existing data, but there is no way it can gather real data UNLESS you are able to monitor this. This is possible via machines, e. g. drones, but AI does not understand what it is doing and even with instructions you still may be able to just hallucinate data. So perhaps one day this may all be fully automated (sensor systems can do all humans can do too, of course), but right now this is simply not the case. And this is just one example for many more.
For example often if you have a blackfly infestation, the first sign is often a steady stream of ants - they are feeding of the honeydew created by the blackfly, they protect the blackfly from predators like ladybird larvae, and they will even transfer the blackfly to new plants.
Once I knew this I found the best way to tackle blackfly was not to go after the blackfly, but distract the ants - a bit of jam works a treat.
Though you could then argue the ant's have then moved on, from farming, to a protection racket.
"Others have expressed concern about ‘AI colonialism’, a practice in which data, collected remotely in poorer countries, are siphoned off for analysis in well-equipped labs elsewhere."
I'm a touch confused, the lead in has "plants shift their flowering times to cope with the heat, rather than adapting through natural selection." How is that not just part of natural selection? Plants that are unable to shifting their flowering time will die off. No?
I feel the same with experimental biology with animals. Nowadays, you need fancy computational models with loads of data to publish in good journals. The reason: because everyone else is doing it. The days of elegant, simple experimental design are gone for the better or the worse.
23 comments
[ 3.2 ms ] story [ 42.9 ms ] threadThe 'computer scientist' quote illustrates a frustrating trend: tech-centric 'drive-bys' that lack the ecological context required for good science. On the flip side, the 'old guard' who ignore modern data assimilation are leaving massive potential on the table. The field is rightfully shifting from site-specific anecdotes to foundational, broad-scale work, but we need both skillsets to do it justice.
I will add that funding can complicate things a bit, funding sources often get wowed by more "advanced" methods, while the underlying science might be less than stellar. There are important questions that can be answered by small, elegant field studies, and there are questions that require larger datasets and more computation. When we start putting the methodological cart before the scientific horse, that's where we run into problems.
I always felt like one of the primary motivations to pursue science was being able to bail out of the office for the entire summer for "field work"...
It's also always error-prone. Nothing in the field is perfect. Reality is a bad approximation for your model at times, if you take a model centric view.
I would be immensely skeptical that field work is ever going away. There may be aspects of truth in this around cost of travel, risk, seniority.
90% of the time it is spend analyzing data or writing up proposals/grants/papers. i don't think AI was the turning point.
> Scientists who run long-term ecological studies, in particular, report that they struggle to find funding.
It's cheaper and easier to do stuff sitting at a desk. In theory that's a good thing if it means more work gets done, but field work has to happen too. For many people it's the best part of the job, for others it's a pain that has to be suffered through to get the data they need. Hopefully there's room (and funding) for both kinds of people to do the work they want.
Attempting to convince people to change course and focus on restoration has mostly been a losing battle, with much larger forces behind the main detriments that make local changes feel inadequate.
I would say the main workflow is collect some new data nobody has collect before, look at it and see if it shows anything interesting, make up some interesting publishable interpretation.
It feels like it'd be smarter to start with working with existing data and publish that way. If you hit on some specific missing piece, go collect that data, and work from there. But the incentive structures aren't aligned with this
The AI angle is really shoehorned in, but irrelevant to the larger problem. Sure, it allows you to annotate more data. Obviously it's more fun to go do field work than count pollen grains under a microscope. If anything AI make it easier to do more fieldwork and collect even more data b/c now you can in-theory crunch it faster
that's the original title before editorializarion
There is still field work that happens. AI can not replace that. You'd need to literally simulate the whole world before AI can even get close to gather all data obtainable here. For instance, on a given ant hill: which plant species will be more prevalent there? (For those not knowing a lot about ants: some ant species carry specific plants or defend plants against invaders. The most usual example is for leaf-cutting ants, but there are many additional examples, and for various reasons you will also find different plant species to be more prevalent close to an ant hill in a forest area, than other plant species.) AI can steal existing data, but there is no way it can gather real data UNLESS you are able to monitor this. This is possible via machines, e. g. drones, but AI does not understand what it is doing and even with instructions you still may be able to just hallucinate data. So perhaps one day this may all be fully automated (sensor systems can do all humans can do too, of course), but right now this is simply not the case. And this is just one example for many more.
For example often if you have a blackfly infestation, the first sign is often a steady stream of ants - they are feeding of the honeydew created by the blackfly, they protect the blackfly from predators like ladybird larvae, and they will even transfer the blackfly to new plants.
Once I knew this I found the best way to tackle blackfly was not to go after the blackfly, but distract the ants - a bit of jam works a treat.
Though you could then argue the ant's have then moved on, from farming, to a protection racket.
Someone is jumping the shark.