User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
1y
So as I went for a walk this morning, I was musing on my switch to Kagi, and hosting my own bookmark manager, and Mastodon, and so-on. One of the key things that make these services better, I think, is that they rely on intentional user signals. Three quarters of the websites I visit are totally unimportant. Most of the YouTube videos I view turned out to be neutral to bad, and I don't want to keep seeing more of them. Half the products I look at I have no intention of purchasing. But big tech slurps up all this low quality data and throws it into a big data soup. However: garbage in, garbage out, right? I think hyper-personalization isn't the exclusive domain of big tech. In fact, without big data, I think we can do it better. Because we're relying on explicit decisions that users made, with the intention of effecting their experience. So it's less data, but the quality is a hundred times higher. For example, I would love a fediverse client that can determine the posts that are most similar to posts I previously boosted, and let me choose to sort those to the top. I follow a few hundred active posters, and there's no way I can keep up with all of you. I know I'm missing super cool stuff. But to surface the cool stuff, we don't need big tech and data from millions of users. We just need to make better use of the personal data we already have about ourselves.
3
7
9
0
User avatar
Nic Roland mastodon @nicr9@techhub.social
1y
@fastfinge I 100% agree about intentional user signals. There's a persistent mindset in big tech that hyper-personalisation is just a matter of extracting latent gold from the sludge.

I like the idea about using personal ML to "be more like ourselves". I think it's a beautiful idea.

We can blame big tech for the current state of the mainstream internet, but strictly speaking I don't think the problem is that they're ACTIVELY trying to shove garbage down their users throats. I think that's a second order effect.

The real problem is SEO. Sure, advertisers could post on Mastodon like anyone else but they'd have to compete as equals. Mainstream platforms provide them with a shortcut. Views go to the highest bidder. The advertisers are now incentivised to shovel garbage (and money) into the system and the platform is incentivised to predict the shortest path between the ads and the users.

Personal ML isn't centralised and offers users more control than advertisers so it should be harder to actively exploit.
1
0
1
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
1y
@nicr9 Also, advertisers want to put everyone into similarity buckets. And that's what AI can be extremely good at, if it's using data of billions of people. It's not recommending the content you would like most. Instead, it's trying to find other people who are most similar to you, and push you the content they like. There's a subtle but important difference. As a blind person, I feel this a lot. I consume a lot of content about video games that happen to be accessible (Forza, Hearthstone, The Last of Us, etc). So big tech pushes me content about other video games people play, but those games aren't accessible. The AI has dropped me into the "video gamer" bucket and assumes I like the same kind of things other gamers do. And advertisers can now target me with generic video game content. Whereas distributed, personalized AI would just be focused on learning about you as an individual, instead of trying to decide what massive bucket you fit into.
1
0
1
0