Hacker News Reader

Comment by experimental123 | Hacker News Reader

Kinda interesting that new farming techniques are still being discovered. It seems like someone in the past should have tried this experiment and kept some fields unflooded but no one ever bothered to do the experiment and measure yields.

Something AI might be good at is suggesting new farming techniques and processes for increasing yields. There is probably enough literature in agricultural sciences with data for various experiments that could be used as a training corpus.

View on HN

> Something AI might be good at is suggesting new farming techniques and processes for increasing yields.

Why do you believe it'd be good at this?

I asked ChatGPT for a novel farming idea and it gave me pseudoscientific bullshit. https://chat.openai.com/share/4ba0c013-466e-4bea-ad00-c8ba22...

"Bio-Resonance Farming is a cutting-edge approach that harnesses the principles of bio-resonance and plant communication to enhance crop growth, health, and yield. Bio-resonance refers to the idea that living organisms emit unique electromagnetic frequencies, and by understanding and harmonizing with these frequencies, we can optimize plant growth and overall agricultural productivity."

The current models were trained on a corpus that is essentially all fiction with no basis in reality. If the training corpus has real world data (like experimental results from agricultural experiments with crops and planting schedules along with their yields) then the neural network should uncover some patterns that wouldn't be obvious simply because finding correlations in large data sets is a hard problem but it is very well suited to analysis by large neural networks.

I mean, you can go try this now; feed some agricultural scientific journals into a model. I suspect it's going to be substantially harder than you expect.

I agree it is a very easy to do which is why it's surprising someone hasn't already tried it. Most of what I see are toy projects with LoRA for generative models bolted onto existing LLMs for fiction instead of scientific applications. These models already work for software so I see no obvious obstructions why they shouldn't work for agricultural experiments.

It's not very easy to do. LLMs aren't capable of understanding, they can merely regurgitate what they've read based on statistical analysis of what words appear to be linked to each other. That doesn't help you when you need to do something new; at best an LLM can tell you what someone else has already done.

There are computer programs that do the kind of thing you're thinking about, for example, for protein structure analysis. They're incredibly complicated and generally require a lot of processing power.

How about

https://www.frontiersin.org/articles/10.3389/fpls.2023.11283...

That's a simple application of machine learning algorithms you might find in scikit-learn. Here is a special issue of another alleged "predatory journal" that is full of papers on the subject

https://www.mdpi.com/journal/agronomy/special_issues/E18K759...

LLMs are a type of machine learning. The stupidest type of machine learning.

The OP did not suggest machine learning in generally, they suggested LLMs specifically, which time and time again have been shown to be incapable of this task as a matter of fundamental design. Worse, because LLMs can't understand the their training data, the output of an LLM must be verified, which in situations like this would probably take more time than simply conducting novel research in the form of random experimentation.

Also, you really need to read your citations. The first one found that machine learning was unsuited for the task of agricultural prediction...

>These models already work for software

Do they? You're talking the agricultural equivalent of something like "devise a new sorting algorithm with sota performance on x, y, z", not "write me some crud boilerplate".

I just mean finding correlations in data sets that are hard to find in other ways. The main idea is that there are plenty of data sets on various cultivars and experiments for how to increase yields. There are probably patterns in the data that would be amenable to analysis by neural networks. The article gives an example for how scheduled flooding can increase yields and I bet there are a lot of low hanging fruits like that to pick. This doesn't require discovering anything novel but simply surfacing some patterns in the data that is buried across several papers and hard to uncover by classical meta-analysis and statistical techniques. Neural networks are very good for uncovering non-obvious statistical correlations which can then be verified by experimentation.

After reading the article I'm sure there are plenty of low hanging fruits to uncover in yield optimization by trying different schedules for flooding and soil enrichment with different kinds of fertilizers. A neural network doesn't have to understand anything to point out useful statistical correlations just like it doesn't have to understand code semantics for incomplete code fragments to suggest potential completions which are then verified by the programmer/compiler/type system.

I would speculate, but don't concretely know, that this is what will happen. I know papers in other fields that were just this; analyzing conditions that successful and failed experiments were performed in and then using ML to derive optimal conditions.

People do all kinds of meta-analysis and literature reviews today, I am sure somebody is already applying A.I. to the document handling for this task but doing a quick search it is hard to differentiate it from literature reviews on the subject of A.I. in agronomy such as

https://www.frontiersin.org/articles/10.3389/fsufs.2022.1053...

It's a big problem that ChatGPT has seduced a large number of people into thinking chatbots = AI and those people have convinced most other people that it is a scam.

I find 77,000 or so articles on "rice" in PubAg

https://search.nal.usda.gov/discovery/search?query=any,conta...

Just like many other areas, agriculture responds to knowledge and is a highly competitive international business. For instance, rice is cultivated by very different methods in Louisiana and Bangladesh and rice from either place could make it to your table.

See

https://en.wikipedia.org/wiki/System_of_Rice_Intensification

for a method which is heavy on labor input and light on fossil fuel input.

> I find 77,000 or so articles on "rice" in PubAg

Analyzing this data set with an LLM would be a very good research project.

Exactly, and not that hard. My RSS reader has ingested about 250,000 articles from random sources since the beginning of this year and does a cluster analysis of about 50,000 of them every day in under two minutes.

Most software dev is repetitive as hell, monkey see monkey do within a computer readable language that has well defined syntax. LLMs can do fairly well in this niche.

Research is by definition not repetitive, the text is free form and the data is never formatted in a way that makes comparison between different papers straight forward.

That's exactly the type of data set that can be analyzed by large neural networks. Heterogeneous data with hidden and non-obvious statistical correlations which would be hard to uncover with classical statistical tools and techniques.

Not at all convinced that this is true, the contrary. Do you have a reference, or something similar that did this successful in another field? (No, that's not ChatGPT and writing some limited software).

Facebook's Galactica.

Which failed miserably at this task...

https://www.technologyreview.com/2022/11/18/1063487/meta-lar...

"A fundamental problem with Galactica is that it is not able to distinguish truth from falsehood, a basic requirement for a language model designed to generate scientific text. People found that it made up fake papers (sometimes attributing them to real authors), and generated wiki articles about the history of bears in space "

There's no shortage of ideas for improving real-world processes. Most of those ideas are bunk, and we're constrained by the amount of experiments[1] we are willing to run/fund, and the quality of data[1] that those experiments can collect, and the reproducibility[1] of those experiments.

Having an AI shout random ideas is very easy for software people to grok, but isn't going to help. If you want AI to assist with this, you'd need to build an 'AI' that can run the real-world experiments, and that's a few orders of magnitude harder than feeding a text corpus to an LLM.

'Thinking' about this problem isn't the hard part, the hard part is doing it. Even using an LLM for something like a meta-analysis of existing research is unlikely to find many profitable avenues of exploration.

[1] Experimental research is incredibly difficult, which is a fact that's highly underappreciated by people working in abstract and theoretical disciplines.

I have my doubts an AI will reliably generate results that are scientifically verifiable.

AI interpolates across it's parameter space, but typically performs poorly in extrapolation exercises.

ChatGPT is a glorified CTL+V of loosely connected content available on the internet.

Yes, but even a model trained on a bunch of scientific papers will lack understanding in the same fashion, until there's some new technological breakthrough.

Understanding is not necessary for uncovering statistical correlations.

Neither is AI.

The parent poster wants it to suggest new farming techniques, which is a little more involved than plotting a trend line.

AI is simply about finding correlations in large data sets. Computers don't understand anything, they just shuffle symbols. So training an LLM on agricultural research will likely uncover patterns that would not be obvious to people and these patterns could point to new techniques and processes for increasing yields like scheduled flooding (as explained in the article). LLMs don't understand code but they consistently can complete code fragments which end up being correct more often than not. A model for yield optimization doesn't have to understand farming to suggest techniques and processes for increasing yields just like LLMs do for code fragments.

LLMs can complete code fragments because they're just matching them up based on having seen these code fragments paired together in they're database.

Yeah, just like a car is a CTRL + V of parts available on the car parts market.

You prompted it in exactly the wrong way. It also says:

> It's important to note that the concept of Bio-Resonance Farming is speculative

“Speculative” is… charitable.

Why do you expect an LLM would be the tool for this job? There's plenty of “actually smart” AI (well, it's legit, so call it ML) out there that can do mathematical/scientific analysis better than we can.

When all you have is a hammer.

(and a hype cycle)

It’s not a new technique as i understood it from the article. It was just abandoned by the introduction of fast growing hybrid rice. Nonetheless it’s very interesting the experiment has not been done before, couldn’t have been discovered at a better time!

“Lansing, an ecological anthropologist, has studied Indonesia’s rice fields since he arrived in Bali in 1974 to work on his Ph.D. His focus was subak, a rice irrigation system managed by water temples, which had been in place since the 9th century until it was disrupted by the arrival of the Green Revolution in the 1960s and 1970s. Like their counterparts across the globe, Balinese farmers were encouraged to swap slow-growing local varieties for fast-growing hybrid rice, fertilizer and an extra harvest.”

I think the flooding of the paddies is also related controlling weeds and pests.

So the viability of the technique may depend on other technology or resources being available, compared to peasant farmers of the past.

Experiments are difficult in farming. Most land holdings are not big enough and that land is the single source of income. So risking it with experimentation, when profits are anyways not high, becomes problematic

Weeds grow crazy fast here in the tropics. With plenty of rainfall, flooding the rice field is the easiest way to prevent weeds from growing.

Aye, if you grab a home gardening in the tropics book if it's a good one it's going to give you info on how to deal with the amazingly fast carbon turnover.

There are all kinds of innovations being made in farming, and many more valuable practices from history that have been left by the wayside. Big Ag is and has only ever been interested in the bottom line.

Reply on news.ycombinator.com