The Perfect Milk Machine
May 1, 2012 11:11 PM   Subscribe

The Perfect Milk Machine: How Big Data Transformed the Dairy Industry
posted by Joe in Australia (38 comments total) 18 users marked this as a favorite

 
Yes, but all the cows will get the blight and rot in the fields and it will be the end of the Irish all over again.
posted by Foam Pants at 11:54 PM on May 1, 2012 [3 favorites]


The weird thing about this... you could conceivably store semen in a freezer somewhere, a bull could continue to have offspring long after the bull dies. I wonder if you could freeze a bulls balls have them continue to generate semen invitro. I'm sure these guys will try it at some point.
"[In the past], we combined performance data -- milk yield, protein yield, confirmation data -- with pedigree information, and ran it through a fairly sophisticated computing gobbledygook,"
Oh yeah, the gobbledygook. That's the good stuff. You always want to run your Big Data through the gobbledygook - it gets the best predictions. Make sure it's the sophisticated gobbledygook though. You don't want the crude stuff - you end up with little bits of dygook all up in your eigenvectors.
posted by delmoi at 12:26 AM on May 2, 2012 [9 favorites]


Why is Alexis Madrigal so consistently interesting.
posted by psergio at 12:45 AM on May 2, 2012


Delmoi: I wonder if you could freeze a bulls balls have them continue to generate semen invitro.

This makes me think of s cheesy horror movie that should be made about huge killer bulls balls that are unfrozen and come back to life and terrorize the countryside indiscriminately shooting off bull semen. It could happen. Especially with such magical self-propelled intrinsic balls.
posted by Skygazer at 1:54 AM on May 2, 2012 [1 favorite]


This definitely does seem like breeding in weakness, as Foam Pants alludes to upthread.

While this is an interesting article, I do wish we could find a better term than "big data". It's turning into the new "in the cloud" as a catchphrase for a group of only loosely-related technologies.
posted by KGMoney at 3:25 AM on May 2, 2012 [1 favorite]


Holy cow, I'm mentioned by name in this article. When I gert to work and I'm not typing on my phone I'll be happy to answer any questions you guys might have, even the ones about frozen bull balls. Awesome post!
posted by wintermind at 3:34 AM on May 2, 2012 [4 favorites]


Metafilter's own Gregor Mendel
posted by unknowncommand at 3:59 AM on May 2, 2012 [5 favorites]


This is the true first step toward Gattaca. People have tended to think about this in extremely reductionist terms — that is, science works out exactly how each gene affects phenotype, including all the epigenetic stuff, and you end up with some deterministic description of the human genome, as seen in the film.

Much more likely, and much sooner, we'll see a combination of vastly increased data gathering on the actual human genome coupled with the already-existing and ever-increasing data gathered about peoples' bodies and health, and with statistical analysis we'll get something close to what's described in Gattaca without the need to understand all of the details of the molecular biology.

Being able to accurately predict seven or so characteristics of milk production in dairy cows from a huge mass of genetic data without knowing what's actually going on at the molecular biology level proves that this can be done with humans (though probably not regarding milk production). We only need the data and the analytical techniques. The latter already exists and the former is already beginning.

Anyone who worried about a Gattaca world but was comforted by just how difficult the genetics problem is from a truly reductive standpoint should still be worried.

And the thing is, there's no stopping this. We shouldn't want to — the science involved and the whole of what it will make possible have far too many benefits. And we couldn't stop this, anyway. What we should want to stop, however, is the codified injustice seen in a Gattaca society. We're not going to stop that world from being born by preventing the related science being done, we can only stop that world from being born by political and legal action designed to protect justice.

In my opinion, this is very much like the discussions about privacy (indeed, the underlying issue in this is privacy, really). There's no stopping the technology or the data-gathering that is eroding contemporary privacy. Attempting to protect a just society with regard to privacy by opposing the technology is foolish. What really needs to be done is social, political, and legal change that protects privacy where it otherwise is easily invaded.

Just so with this sort of genomic data-gathering and analysis. We can't and shouldn't stop the data-gathering and analysis. What we can stop is it being put to uses that create a grossly unjust society.
posted by Ivan Fyodorovich at 4:41 AM on May 2, 2012 [6 favorites]


Have they solved the problem where it doesn't stop until 50 gallons have been withdrawn? YEAH! WITHDRAW-AW-AWWWN!
posted by (Arsenio) Hall and (Warren) Oates at 4:50 AM on May 2, 2012 [2 favorites]


Somebody looking for a sockpuppet account name could do a lot worse than "Badger-Bluff Fanny Freddie," especially if you're looking to ask a certain type of question.
posted by .kobayashi. at 4:53 AM on May 2, 2012 [1 favorite]


KGMoney, you should probably skip this article of mine, then. :-)
posted by wintermind at 4:55 AM on May 2, 2012 [1 favorite]


Ivan Fyodorovich, we actually compute genomic evaluations for 31 or so traits for six different breeds of cattle (it varies a little by breed). We're also working on understanding the molecular biology underlying those traits, but it's a hard problem because it's very expensive to do basic biology with large mammals, and from a genetic prediction perspective you don't really need to know the biology for the process to work. Many of us DO want to know because we're curious people, and we've found some interesting things [self-cite].

I would do a study on human lactation if I could identify a good source of funding to support it, but that's surprisingly hard to do. There's a lot we don't know about human lactation, particularly when it comes to the shape of the lactation curve. The best use of that knowledge wouldn't be to select humans for increased milk yield, but to better understand when women need to pay particular attention to their diet (although humans almost certainly don't experience negative energy balance for very long) and when supplementation with formula or stored breast milk might be necessary because production hasn't yet reached peak yield, or yield is decreasing over time.

I'm not an expert in human lactation, and I hope that haven't said something phenomenally stupid, but it's possible that I did. If so, I hope someone will correct me.
posted by wintermind at 5:05 AM on May 2, 2012 [7 favorites]


Nice post, fascinating subject, and generally good article. However, the presence of a parenthetical "heh" in The Atlantic was kind of depressing.
posted by Trurl at 5:34 AM on May 2, 2012


Wintermind:Actually, that looks pretty interesting. I'll have to read it when I'm not at work. The fact that you define what you mean by "Big Data" in the title of the paper makes it all OK. It's just that people say "Big Data" when they really mean "Large dataset analysis", the way you use it, "Large amount of data storage and retrieval", or even "Distributed storage and retrieval", and I wish we would settle on different names for these technology sets.

OK, I'm done with my derail.
posted by KGMoney at 5:35 AM on May 2, 2012 [1 favorite]


Let's hope that milk cows breeders are smarter than beef cattle breeders. The pursuit of more and more meat per cow has contribuited to dystocia for calves, with death from birth approaching 10%.
posted by francesca too at 5:36 AM on May 2, 2012


francesca too, we scientists at AIPL have done a lot of research on dystocia and stillbirth, which is a problem for Holstein heifers in particular. We've identified a major gene associated with calving traits in Holsteins, but we're still working-out the biology there. The papers on our Scientific journal articles page should link to PDFs on our website, not to paywalled versions. Unfortunately, the Atlantic article links to one of my papers on the publisher's site, which is not accessible to lots of people.

It's probably not really a secret at this point that I'm the Cole in Cole et al., right?
posted by wintermind at 6:00 AM on May 2, 2012 [6 favorites]


The idea of an algorithm that takes the genetic data of a mammal as input and gives a dollar figure as output is really sort of sad. Living, conscious beings deserve our respect and empathy, even if their ultimate use is food production. The article ends:
He is, for all intents and purposes except for his own, genetic material that comes in the handy form of semen. His thousands of daughters will never smell him and his physical location doesn't matter to anyone.
It really ought to matter
posted by crayz at 6:52 AM on May 2, 2012 [1 favorite]


This is not new. The data used to be kept in a "recipe" card file and digested by a MK1 human brain. http://www.dhia.org/default.asp

My father worked on a farm that did this as early as 1950. He did the data processing by shuffling cards.
posted by Jumpin Jack Flash at 6:57 AM on May 2, 2012


For example, Kruglyak said, human population geneticists want to figure out how to explain the remarkable lack of genetic variance between human beings. "The typical [genetic] variation among humans is one change in a thousand," he said. "Chimps, though they obviously have a much smaller population now, have several fold higher genetic diversity." How could this be? Researchers hypothesize that human beings once went through a bottleneck where there were very few humans relative both to the current human population and the chimp population. Few humans meant that the gene pool was limited at some point in the pre-historical but fairly recent past. We've never recovered the diversity we might have had.

OMG - so there really was a great flood!
posted by b33j at 7:12 AM on May 2, 2012


This list of top bull names is Mefi sockpuppet material gold.
posted by euphorb at 7:18 AM on May 2, 2012


I'll be happy to answer any questions you guys might have, even the ones about frozen bull balls.

Do I need to defrost them before breading, or can they go straight into the deep-fryer?
posted by Greg Nog at 7:22 AM on May 2, 2012 [1 favorite]


Hope this isn't a derail, but I wanted to give a shoutout to a local business in my dairy community working on micro-dairy machinery. A MeFite is the lead engineer on the "low input-low impact pasteurizer" but I don't have his permission to "out" him here. So while the article does state that there are fewer farms with larger herds, there are people trying to keep family-run farms alive ... at least here in Vermont.
posted by terrapin at 7:26 AM on May 2, 2012


It's best if you defrost them first, Greg Nog, so that you don't get a fried-frozen-bull-ball apocalypse in your kitchen.
posted by wintermind at 7:33 AM on May 2, 2012


Agreed, Jumpin Jack Flash. A few years ago AIPL, which was originaly established to work with the national milk recording system, celebrated its centennial.
posted by wintermind at 7:36 AM on May 2, 2012 [1 favorite]


fried-frozen-bull-ball apocalypse

What was that you were saying about sockpuppet names, euphorb?
posted by Rangeboy at 7:50 AM on May 2, 2012 [1 favorite]


OMG - so there really was a great flood!

Only if that great flood somehow managed to selectively drown only humans, because the chimps apparently weathered the flood just fine...
posted by saulgoodman at 7:51 AM on May 2, 2012 [1 favorite]


wintermind: "...we scientists at AIPL...

In my mind that stands for Animal Intestines Propulsion Laboratory.
posted by symbioid at 8:01 AM on May 2, 2012 [2 favorites]


wintermind, Is there a concern that by selecting bulls for success on a particular set of genetic markers that are known to us, we are significantly reducing the genetic diversity of all cows? When I read that one bull is responsible for an estimated 14% of the genome of a particular breed, I wonder if that isn't inheriently risky.
posted by Lame_username at 8:20 AM on May 2, 2012


We worried about that even before we started using genomics, Lame_username, but we have a couple of ways of dealing with the problem. We use a statistical adjustment to account for the effects of future inbreeding in the population, so the genetic evaluation for, e.g., Freddie already has been reduced to account for future loss of variability. The research says that the use of genomic information can support increased rates of genetic gain while preventing the loss of significant variation, but that is not how things are necessarily working out in the field. The challenge is that producers seeking to maximize genetic gain will use the bulls with the highest genetic merit, and those animals tend to come from a small number of families.

In grade (average) cattle, there also are lots of errors in pedigrees, which can effect our estimates of genetic variation. We have found the genomic information to be very useful in identifying and correcting pedigree errors. This will not increase heterozygosity per se, but it will help us avoid matings based on incorrect parentage.

It's not really discussed in the article, but it's also important to understand that we absolutely must continue to collect phenotypes (performance data) from which we calculate the effects associated with the DNA markers. If the data start telling us that a particular chromosomal segment is associated with low production, poor fertility, reduced longevity, etc. then animals carrying that segment will be penalized, and as a result other families may become more attractive for selection.
posted by wintermind at 8:34 AM on May 2, 2012


*...affect our estimates...
posted by wintermind at 8:36 AM on May 2, 2012


Wintermind, how is semen "data" extracted from a big scary living bull?
posted by Skygazer at 9:05 AM on May 2, 2012


Wintermind, how is semen "data" extracted from a big scary living bull?

There's always some jerk off willing to do it.
posted by orme at 9:58 AM on May 2, 2012


Using a device called an artificial vagina. I'll have to Google up something suitable when I'm not at work. Basically, it's a special sleeve with a tube on the end. The bull's equipment goes in and swimmers come out.
posted by wintermind at 9:59 AM on May 2, 2012 [2 favorites]


b33j: I know you were making a joke about the Genesis flood, but there is a theory that an eruption of a supervolcano in Indonesia reduced the human population somewhere between 1000 and 10000 breeding pairs. Totally unrelated to the article, but I think it's interesting.
posted by Turkey Glue at 10:48 AM on May 2, 2012 [1 favorite]


The BBC did a good history of farming recently. It was bit of an eye-opener just how much tech had influenced the business in recent years.
posted by fearfulsymmetry at 11:00 AM on May 2, 2012


I really hope we're not on the way towards Gattaca because I really dislike swimming and overwrought metaphors.
posted by flaterik at 11:34 AM on May 2, 2012


This is all I know about milk machines.
posted by Ducks or monkeys at 12:12 PM on May 2, 2012


Or take a look at the how its made episode with the robot milker that the cows use when they feel like it.
posted by Iax at 7:32 PM on May 2, 2012


« Older "It's a Good Life"...  |  How to assemble a happy kitche... Newer »


This thread has been archived and is closed to new comments