How AI Could Solve Our Renewable Energy Problem

Video script and citations:

Get my achieve energy security with solar guide:

Follow-up podcast:

Join the Undecided Discord server:

👋 Support Undecided on Patreon!

⚙️ Gear & Products I Like

Visit my Energysage Portal (US):
Research solar panels and get quotes for free!

And find heat pump installers near you (US):

Or find community solar near you (US):

For a curated solar buying experience (Canada)
EnergyPal's free personalized quotes:

Tesla Referral Code:
Get 1,000 free supercharging miles
or a discount on Tesla Solar & Powerwalls

👉 Follow Me
Mastodon

X

Instagram

Facebook

Website

📺 YouTube Tools I Recommend
Audio file(s) provided by Epidemic Sound

TubeBuddy

VidIQ

I may earn a small commission for my endorsement or recommendation to products or services linked above, but I wouldn't put them here if I didn't like them. Your purchase helps support the channel and the videos I produce. Thank you.

[timoluetk]

It's really nice to hear something hopeful in these turbulent times. Much love Matt!

[TwilightMysts]

Semi-related, I vaguely remember a story where a group of pigeons were trained to identify malignant cancers in medical scans. The process was similar to machine learning; they started out with a sample of known pictures, any time the pigeon selected a picture with malignant cancer, it was rewarded with food, thus training the birds to select pictures with malignant cancer. I want to say that any given pigeon had less than 80% accuracy, but when they ran it as a group (presumably something like requiring at least 60% to say yes) the accuracy climbed to over 98%.

[namAehT]

The problem with using a neural network to determine grid functionality is that they're effectively a black box. If something goes wrong, there's no debugging because we genuinely don't understand how the network works.
If you want to make a change, you have to retrain the model.

[CurtGodwin]

Nvidia: "We build hardware that can help model and predict global warming."

Also Nvidia: "Look at this gargantuan GPU we made that sucks up nearly 1, 000 Watts!"

[AsciiSmoke]

One potential concern of machine learning when it comes to other things like designing products that work best for humans is that they can potentially further alienate people who don’t fit the majority model.

For example, people with autism struggle in shopping centres because those buildings are designed to encourage echos and pipe music to give the feel of a bustling busy environment. That can cause painful sensory overload for some on the spectrum.

Similarly, supermarkets use eye-level shelves to market some of their biggest deals, but those in wheel chairs don’t share the same eye level as those who are able to walk.

[deanwight]

Having installed a Sense monitor in my home about 2 years ago, I've been disappointed in its ability to identify individual devices...an ability that Sense markets pretty strongly. The system still classifies about 60% of my consumption in the catch-all categories of Always On and Other. Not the fine-grained detection I'd hoped for.

[johnmoncrieff]

I am still troubled by the fact that all these high powered renewable companies fail to recognize the fact that wind and solar are two of the least efficient forms of renewable on the planet and can never produce the level of efficiency or continuous base load of HYDRO POWER! Through out the world there are areas of land that are subject to consistently high levels of rainfall on terrain that is ideal for hydro electric dams, ( primarily in sparsely populated deep valleys with over 3 meters of rain per year!) Then there are run of river units that can range in size from small individual units that power only a single off grid dwelling to ones that power a city! All requiring minimal maintenance over the extremely long life span of them compared to that of wind and solar!
Then there are wave and currant power generation, along with tidal on numerous coastal sites around the world that should be developed

[jimdob]

If people invested in passive homes and then did this as well and we did this for an entire nation then I truly wonder how much power would we actually use in America. Biggest issue will be the factories and corporations since they are the biggest offenders anyway.

[lazytsfarm]

just talk to a business chatbot to learn how far ai has progressed.

[robertfoertsch]

Excellent Analysis, Deployed Worldwide Through My Deep Learning AI Research Library…
Thanks Matt

[vidurchandra]

Thank you Matt! As a data scientist aspiring to work in the field of renewable energy, this video is a gold mine. Do you have any resources of analytics and data science projects that I can tap into and learn from?
Not sure if you'll get a chance to see this but, thanks a ton.
You're a rock star! \m/

[Riverpines]

Machine learning would be a huge help in winter road maintenance, predicting how much snow and ice response will be needed over a large area on the roads how much sand and chemical yeah I sure would be needed for each winter. Currently we only have past weather data to go by and passed expenditures. A lot of money and grief could be saved by knowing how much to allocate for a system and when.

[richardcarpenter]

Matt, I was underwhelmed with "Sense" AI service. I had it installed for about 6 months, but it never seemed to identify many of the devices installed in my house. There was just a pretty always on or showed that we had consumed X number of watt hours. This included t6hings like refrigerator and standalone freezer. When it did register a device, I do not think it ever identified it as the actual device. I have since installed an Enphase system with batteries and gave up on trying to identify items consuming electricity.

[mmcowan]

Hey Matt, there seems to be a categorical blame of landfills whenever there is some sort of discussion of our environment. However, there is never any proof, data, or arguments presented to back up the assertions. I believe that the evils of landfills are grossly blown out of proportion. For example, what's the issue of burying discarded plastic in a landfill until it breaks down? Plastic came from the ground and now it's going back. Are we actually "clogging up and filling up" all of our landfills, or is this a fallacy? Can you please do a video on the subject? Check out the episode of Penn & Teller Bullsh*t to see what I'm talking about.

[dannydenison]

Videos from the clean energy creators like this are such a nice thing to look forward to. Thanks

[sunkicked]

GE's Podcast Theatre did a story called LifeAfter regarding digital twins. Really cool seeing the ideas presented there being discussed as real world tech. I think GE needs to do more of those, as it's a fantastic way to get concepts out to the public.

[generalharness]

Medical researchers where using machine learning to check if moles where cancer or not. They feed alot of pictures of moles though the machine and it ended up with a 100% accuracy. As in it picked up on every cancer mole.

They looked into it and it ended up that the machine was not even looking at the moles, it was looking for a tape/ruler as all the cancer moles had one in the picture.

[sophiophile]

Neuroscientists eloquently summarize Hebbian learning as:

Those that fire together, wire together. Those that fire out of sync, lose their link

[jaxolotle]

The solar lab at my school has transitioned to almost all data science specialists to try and advance PVs faster

[kerradeph]

A few years back, I pseudo-coded how I could set up a machine learning driven system for when to turn on and off a fan in my house at night based on predicted weather. I never got around to it, but this just reminded me that I should have pushed more on that.