this post was submitted on 13 Jun 2024
607 points (96.5% liked)
Technology
59148 readers
1986 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Finally a use case where AI/Machine learning would absolutely make sense. If we can have AI that can generate text or images, imitate people's voices or write code, we can also have a lightweight model that can detect ads and skip them during playback. There's a model trained on SponsorBlock data for detecting sponsored segments https://github.com/xenova/sponsorblock-ml
I'm sure that we can have something similar but for embedded ads.
It's called a classifier and it could easily detect an embedded ad. The issue is now everyone needs to run it on their hardware to detect and this will cost some electricity.
Fine with me
Well I'm not happy about potentially adding new type of load on the electrical grids around the world.
Ads have definitely added more load on electrical grids in aggregate than locally hosted and lightweight models, especially given that ads are fucking everywhere all the time. Websites, apps, the servers, even 24/7 electric billboards. I'm not worried about a few nerds using slightly more electricity sometimes for their own benefit and joy (it's still less power than gaming), as opposed to a corp that burns through power and breaks their climate pledges (Microsoft) for the benefit of their bottom line and nothing else. Corps don't get to have a monopoly on AI that was built with our data, only to have it fed back to us to pull more data and siphon more money.
So basically fuck Google and fuck ads.
Do you understand what we're still talking less energy than the monitor it displays on. I would bet even untuned VGG16 could do that without even a fine tune. Advertising is starkly different to content and the output is a "ad=yes/no" signal. It's a very small amount of data, probably less than the plain hardware video decoder. It's also not a new type of load, it runs off the same power supply as any computer, a slight capacitive load, it won't even change the grid powerfactor.
It already exists. Although it's not AI, and mostly works best when using channel logos to work out the ad breaks.
It's hard to run AI on phones and low-end hardware though. It's not that easy.
Give it 5 more years in hardware performance improvements and software/model optimization and I don't see a problem. The important part is that improvements are made public for everyone to use and improve upon instead of letting openai and microsoft take the whole cake