elrik @ elrik @lemmy.world

Posts

1

Comments

221

Joined

2 yr. ago

You're in the wrong Rochester!

You may be underestimating the role of targeting in conversion optimization, and I'm not sure how you could better target individuals than based on what they're thinking at any given moment (literally).

For instance, it's not hard to imagine a future where gen-ai inserts product placement for a drug like ozempic into your favorite show, just for your view, while you're actively paying attention, even though you didn't realize at that moment you're still a bit upset about a negative comment someone made about your weight earlier that day. An advertiser didn't have to select this scenario, but instead you were targeted by an ML algorithm at that moment based on brain activity correlated with others who ultimately were successfully nudged to have a conversation with their doctor about their weight. Simultaneously, another ML optimized the product placement generation to minimize viewer disgust while maximizing its visibility. Your behavior becomes immediate feedback to further optimize these algorithms, as you're tracked for how much attention you paid to the placement, what was your emotional state before and after, did you schedule an appointment with your doctor over the next 3 days, were you prescribed ozempic, etc.

That is just a simple example which isn't that far removed from advertising approaches today. I'm certain there are plenty of clever techniques to turn your thoughts and perceptions into conversions far more effectively once advertisers have real-time access to your brain.

Advertisers would absolutely love to augment your reality with ads or even just the ability to accurately confirm you've actually watched a traditional ad along with how you "felt" about it.

At that point people would absolutely sign up for free implants so they can access ad supported services that may otherwise become unaffordable within a society further strip mined of wealth by the then trillionaire class.

You have about 25 items too many on this list.

.. because we live in one of the worst timelines.

Agreed, although ranked choice with some form or proportional allocation instead of "single winner take all" might be even better.

I don't actually store any documents in the "Documents" folder, and it's so obviously a dumping ground for lazy devs.

Thanks for the clarification and I believe I misunderstood your original comment.

To add to your list there is an often underutilized feature of GitHub for discussions too.

You mean pretty much a single GitHub account?

Also your quick question may have already been asked and answered but difficult to find on Discord. Or if it hasn't been asked yet, now a future person can't discover the same question easily. So either way you're just wasting other people's time.

The distinction is web workers and offline mode.

It means your PWA can preload everything it needs to run offline, and you can actually use it offline. That is different from a "cached website" which can only cache the pages you've already visited and otherwise does not allow you to update data locally.

Yes, although that recently changed in the EU (only) with the Digital Markets Act.

It's paywalled, but no worries, we all know the answer: they're going to carefully examine the policies each candidate puts forth, consider their respective past job performances, and make a principled decision with their vote balancing their own interests as well as the country as a whole.

Just kidding! They're going to be "uncomfortable" voting for the orange asshat who will do absolutely nothing for them and whose goals include revenge and "absolute immunity" so that he avoids any consequences of being a fraudster, rapist and traitor.

If the goal of training is to produce output that users "like" or engage with, then yes, upvoted content is higher quality. The definition of quality here will certainly depend on their goals.

My point is a bunch of spammed content intended to poison AI training is unlikely to gather upvotes, and so it could easily be filtered out if they're also okay with discarding some human generated content that was not upvoted.

No, because the upvote ratio on posts and comments will be used to signal higher quality content.

It would take considerable effort and coordination to generate low quality content and give it an upvote history that isn't obviously suspicious and do that for enough content that it actually matters to the training.

Even if you could accomplish that, you can't backdate this activity, so they could simply filter out posts and comments after a recent date and still have an enormous amount of data to train.

Where's the spam?

Really interesting! I wonder what would happen if you combine these two properties. Suppose some length of the middle is all walls, and the hooks are infill, or vice versa. Is there an optimal mix that maximizes the weight it can support in your testing, or have you found the optimal configuration (with infill along the entire length) already?

The only way to accomplish what you propose is to switch our elections to ranked choice voting.

Otherwise, I'm sorry but you're stuck with the two major party choices.

the DNC will also never change, if their actions have no consequences.

What consequences do you propose?

You're famous for being photographed shirtless on a horse. Is that because horses can't judge you, or is it a new Russian policy for reducing laundry costs?