Getting Positive Results from My Most Recent Experiment

Miscellaneous and Personal Stuff I Must Share
1

This is very exciting for me. I’m building an application of artificial intelligence that has, as far as I can tell, never been done before. I did a literature search and found nothing, so I had some colleagues do a search and they found nothing. One of the world’s leading experts in this research domain told me that he’s never heard of such a thing being done. He also said “because it is impossible.” And today, after much development, bug fixing and verification, it works. It freaking works!

The ultimate intent of this research relates to plant modelling and simulation. This helps crop physiologists produce better crops cheaper and faster. A secondary benefit is that growers will be able to select the best possible crops for their local environment. That means increased crop production. Now, my piece is of course a very small piece of a very large process, but how exciting to think that I might be helping to feed people some day.

Anyway, that’s it. I’m just excited. Homer Simpson can perhaps sum it up a bit better.

2

Congrats! Just hope you didn’t trigger the singularity! :stuck_out_tongue:

3

Very cool. :slight_smile: Cautiously excited for you.

4

Would you be able/allowed to describe your work in (fairly) layman’s terms?

5

Woohoo! What a great feeling. Congratulations. :slight_smile:

6

Sure.

Imagine you take a 3D image of a plant. Then you take another image as the plant grows. So after a period of time you have this series of 3D images. You convert the images into symbolic instructions on how to draw them. For example, F is a line, + is turn left, - is turn right, [ is start a branch, ] is end a branch, L is a leaf, etc. So now you have this time series of instructions for how a plant grows. For example:

F
FF[+FL]
FF[+FL]F[-FL]
FF[+FL]F[-FL]F[+FL]

(if you picture this in your head you’ll see a basic branching structure)

Now the question is what is the mathematical model that produces those instructions?

In this case, you have:

A -> F[+FL]B
B -> F[-FL]A

This are read as replacement rules. So it means a symbol A is replaced with the symbols F[+FL]B and B is replaced with F[-FL]A.

A and B represent underlying mechanisms of growth. So if you start from FA you get:

FA
FF[+FL]B
FF[+FL]F[-FL]A
FF[+FL]F[-FL]F[+FL]B

Erase the non-physical symbols (the As and Bs) and you have the original instructions back to draw the plant. So, the two rules above (the model) now describe how a plant grows.

There’s been a lot of work on finding such models but for real plants they are found by teams of human experts because they are very complex (the example above is extremely simple). They include parameters such as temperature, avg amount of light, soil conditions, etc. In other words, it isn’t so simple that A -> F[+FL]B but that this only occurs when the temperature is above a certain point, and there’s been enough light and the soils conditions are correct, etc. here’s been lots of work on finding models automatically but none have succeeded, or even come close, to finding models for real plants.

My algorithm uses artificial intelligence to find such models and for the first time ever has found a model for a system that incorporates such complexity. Eventually this can be used to create predictive models of plant growth and that’s how it helps plant physiologists and growers. Physiologists will be able to modify the parameters of the system (which will eventually include the genetics of the plant) and make a reasonable estimation of how it will grow. Now it isn’t a case that they would fully trust the prediction, but it allows them to eliminate the clearly bad choices. For growers, it allows them to input their conditions and find the plants (or particular breed of plant) that will grow best for them.

Hopefully, that all makes sense. If you have any questions, other than exactly how I do it, for that you’ll have to wait for the inevitable publication(s), I’d be happy to answer them.

7

Quit bragging, hush up, and start patenting. You’re sitting on a large oilwell if you don’t give it away first. *Then *come back and fill us all in. :smiley:

Beyond that, big congrats for a big event. That’s got to be an amazing feeling.