Hacker News new | past | comments | ask | show | jobs | submit login
Germany enlists machine learning to boost renewables revolution (nature.com)
99 points by vezycash on July 14, 2016 | hide | past | favorite | 25 comments



This article in Nature, which the Popular Mechanics article appears to be based on, is much more informative: http://www.nature.com/news/germany-enlists-machine-learning-...

It's a little misleading to say they're using AI. Most of this work (and all the specific examples listed on the project webpage) is more appropriately called statistics.


Nowadays there's no clear difference between statistics and machine learning. But the latter gets you more grant money. :-/


> no clear difference between statistics and machine learning

Yes, there is. When your model has thousands/millions of parameters, or when your model has many more parameters than the number of examples in your training set, it's called Machine Learning.


That's not true. Linear and logistic regression are like the first things you learn in a machine learning class. Many popular ML models do not have millions of parameters, including decision trees and shallow neural networks.


Technically you call overfitting AI...


Machine learning is setting it up to find the correlations for you these days :p It is annoying when people choose things due to a more exciting name/buzzword rather than actual merit...


Thanks, we updated the link from http://www.popularmechanics.com/technology/infrastructure/a2..., which does point to this.


Popular Mechanics has become nothing but vapid clickbait nonsense these days. Thanks for the source.


> Since wind and PV-power forecasts do not follow exact theoretical error distributions, the probabilistic power forecasts will be based on quantile regression coupled to artificial intelligence or data-mining methods. Additionally, methods such as kernel density estimation, ensemble dressing, or Bayesian model averaging, which have proved useful in meteorological applications, will be evaluated in this subproject.

Source: http://www.projekt-eweline.de/en/subproject2.html

AI is only mentioned as a possibility in the context of one subproject of three subprojects. Based on titles alone they've made only a single publication[1] involving AI and they've been publishing for a few years already. Saying that AI is used here may be technically correct but seems like a massive overstatement.

[1]: http://www.projekt-eweline.de/en/publications.html


Great. Can we start doing this for the economy, already?


This does happen, but in the US (and most(?) countries) we don't have command economies. Consequently, this sort of model-based approach to economics (logistics) is done by businesses. The big, successful ones are very good at it. Walmart knows how much of what good to send to what region on what day. They've been collecting and correlating data for decades now.

Where it's done by the government it's applied at the level of bonds and currency production and such. More abstract levels than directing purchase or production of specific items (generally, more direct intervention does still happen of course, but usually masked).


Germany doesn't have a command economy either, it's a combination of government incentives, private enterprise, and (semi-)public utilities all going in this direction.


Has been tried with previous generations of decisionmaking technology: https://en.wikipedia.org/wiki/Project_Cybersyn


Anyone not familiar with this, do go to the link and check out the Control Room. Beautifully retro-futuristic, looks like a '70s sci-fi film set, but they were actually using the thing.


A tangential question related to carbon-free energy production:

Are there companies that do software for nuclear power plant construction? Like modern project management software for nuclear power plants? It seems that a few bigger nuclear projects in Europe (namely in Finland and France) are in trouble mainly due to project management reasons.


Siemens. They are building the plants, too.


Renewable energy supply strongly depends on weather forecast so essentially one needs a good algorithm for weather forecast. Yet, forecasting energy supply is much easier if all generators are integrated into one power grid because it is necessary to predict average supply for a (large) reagion.


Why is this news? We've been doing the same thing at deregulated US RTOs for at least 6 years.


AI is the new Nanotechnology


I thought it was the new carbon nano tubes?


I also considered graphene as a candidate. Or "biofuel made from X" (with very rare and expensive catalyst).


There's no such thing as new news.


That name yields some rather unfortunate Google image results.


Looks like a product of German engineering, so what?


More like Swedish.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: