Slashdot Mirror
Google Wants To Use AI To Cut the UK's Electric Bill By 10 Percent (popularmechanics.com)
The Google-owned firm artificial intelligence company DeepMind is in talks with the National Grid about a potential partnership, with the possibility of using the technology to make the supply of energy across the UK more efficient. From a report: Google Deepmind is opening talks with the UK government to use the company's artificial intelligence to reduce energy use by up to 10 percent. Artificial intelligence is highly adept at spotting patterns and making predictions that are much too small and subtle for humans to pick out, which lets AIs to micromanage systems with far greater efficiency than any human engineer could hope to achieve. For instance, Google is currently using Deepmind's AI to control its server rooms, where it manages windows, fan speeds, air conditioning, and more than a hundred other factors to save Google hundreds of millions of dollars in electricity costs.
Bills will go up by 10% to pay for it.
So Google wants UK's energy usage information? Fascinating.
Welcome our electricity-saving overlords!
What are the chances that this optimizes use to zero by turning everything off? Remember the end of I, Robot.
Maybe it can start by identifying and adjusting readings from faulty or noisy meters.
Hell, could claim 100's% of savings. OR maybe just report 10% and tuck the rest of it in profit.
I read it as "Google Wants To Use AI To Increase the UK's Electric Company Margins by 10%, Customers To See No Benefit"
Think about people! How many will lose jobs because of this? /s
In this case, the AI would be used to predict the high and low points of energy usage, as well as supply from renewable sources like wind and solar. Deepmind believes that such a system would increase the country's ability to rely on renewables, cutting energy costs by as much as 10 percent annually. If Deepmind's system is implemented and as successful as they believe, it could save the country billions of dollars a year.
This seems like a very good idea to me. Much better than a brute force solution like selling them more batteries or forcing the use of those silly compact fluorescent bulbs.
Okay, I study grid efficiency in the US and not the UK but assuming the British grid isn't massively worse than the US grid a 10% improvement is probably impossible. 5-6% however is another story. Power cos already utilize intelligent algorithms to manage generation. Google claiming to sit on algorithms 2x more efficient than other utilities seems highly unlikely....They've been doing it since the early 90s and they are constantly improving on their models. Most of any gain will come from deregulation and allowing power from other states to flow more freely. We lose a lot to very dumb rules that force over generation but even here liberalization will likely only yield a 5-6% savings.
No-one else concerned that an AI wants to reduce human use of power so there will be more available for its own processing? No?
In fact from the article itself, we find that Google is not even involved in making this request:
Google Deepmind is opening talks with the UK government
I mean, holy shit!
"There is more worth loving than we have strength to love." - Brian Jay Stanley
So they are able to save money where they control all parameters. Fine.
How are they going to manage windows, fan speeds, air conditioning, and more than a hundred other factors in normal houses?
#DeleteFacebook
Check the weather. Below freezing run overnight on 3. Below 40 run overnight on 2. Otherwise fire it up for an hour or two in the morning if it's chilly.
love is just extroverted narcissism
Here was one I wrote up at the weekend:
http://www.earth.org.uk/Hey-Si...
Guess what could compute a daily forecast ready to upload to those phones and laptops, just for example, as well as some real-time polling?
Some of it could be based on the data used here:
http://www.earth.org.uk/_gridC...
Rgds
Damon
http://m.earth.org.uk/
Google will get to harvest all of that useful metadata about usage, etc.
Artificial intelligence is highly adept at spotting patterns and making predictions that are much too small and subtle for humans to pick out
But all the patterns that AI extracts are historical. They all assume that the events in the future will be caused by, and will act out, the same things that happened in the past.
We have seen this with computerised trading: that all they can do is find a past pattern of actions and try to fit that to what is happening now and will continue into the future. AIs have no ability to understand when the rules have changed, or when new and previously unseen conditions need to be applied.
The UKs electricity generation often runs very, very, close to its limits in the winter. Mainly due to cost-cutting: why spend money on maintaining plant and excess capacity when it won't be used?
To employ AI to shave further percentage points and thereby run even closer to the limits simply reduces the margin for the unexpected. And being unexpected, you can't blame an AI for not spotting those patterns in the past.
A dangerous game.
politicians are like babies' nappies: they should both be changed regularly and for the same reasons
Artificial intelligence is highly adept at spotting patterns and making predictions that are much too small and subtle for humans to pick out
But all the patterns that AI extracts are historical. They all assume that the events in the future will be caused by, and will act out, the same things that happened in the past.
The recent past remains statistically a good guide to the near future. Contingency plans deal with the rest. Using the former better saves money and makes the latter *less* likely.
We have seen this with computerised trading: that all they can do is find a past pattern of actions and try to fit that to what is happening now and will continue into the future. AIs have no ability to understand when the rules have changed, or when new and previously unseen conditions need to be applied.
The UKs electricity generation often runs very, very, close to its limits in the winter. Mainly due to cost-cutting: why spend money on maintaining plant and excess capacity when it won't be used?
To employ AI to shave further percentage points and thereby run even closer to the limits simply reduces the margin for the unexpected. And being unexpected, you can't blame an AI for not spotting those patterns in the past.
A dangerous game.
It's more likely about better scheduling/forecasting than cutting any reserve.
Cover for the largest expected single generator failure were increased when Sizewell (nuke) and then Longannet (coal) tripped in close succession in 2008. Maybe better modelling would have had the increased cover in place *before* then and 500,000 people would not have lost power.
Rgds
Damon
PS. BTW, I worked with low-latency traders. I suspect it doesn't work quite how you imagine.
http://m.earth.org.uk/
Should that really read "electricity bill" or are we talking about some sort of cyborg toucan?
...logical, since the only thing that can stop A.I. is to pull the power, so if A.I. suspect that's the only thing stopping A.I. from total control, it starts with the power of course ;)
What this world is coming to - is for you and me to decide.
Ahhh!!! Micromanagement. Ahhh!!!
This is to compensate for the fact that 10% of today's global electricity bill is caused by training deep learning models.
Getting a good simulation of a city up and running, plus installing the data jack in the back of people's necks. After that 10% should be a cakewalk if that documentary I saw was legit.
Michael Botvinnik's spinoffs from his "Pioneer" chess program optimized energy network planning in the Soviet Union something like 40 years ago.
This is one case where humans might want to be involved - versus being cut off in the middle of a cold winter by AI.
Twitter supports and protects racists - by smearing their critics with the "Hate Speech" label.
They've been doing this since the 1970s. Thats how economic dispatch works for crissakes. Its a giant constrained optimization problem that is solved at 5 minute intervals (I'm over simplifying but the important math is done there)
Google is FOS and I'm sure they know better. It's an excuse to get into the optimization business on their part and try to take a slice out of the US ISO's.
Well said. Speaking as a PJM electricity trader I can tell you that your comments are spot-on.
Neural nets are used all the time for load forcasting and there will always be a limit to their accuracy because every day is a little bit different and you never truly know whats going to happen tomorrow. The same issue holds for unit committment algorithms and contingency screening.
You are just assuming they will be lowering the reserve margin thresholds.
[Disclaimer, I've not read TFS]
Figure out how much of each user's electricity bill is due to google ads and arrange for compensation directly into our bank accounts?
Requiem for the American Dream
They all assume that the events in the future will be caused by, and will act out, the same things that happened in the past.
That's how the human forecasters work too. They check the weather, the TV guide, industrial requirements and look at how they historically affected load. They factor in probability too, e.g. chance of parts of the grid being damaged based on weather. There are never really any unprecedented events, and there is always a balance between cost and reliability when there are major failures.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
Remember Scotland? Lakes at high elevation? Every stream in the damned country blocked with a dam?
Great Britain has many dams that only generate power for a few hours a day or even skip days.
They have at least 100x the potential electricity output they will need for this century.