4 ms·
> “becoming ever more reliable with time, as the algorithms learn“ I mean considering Amazon has put in kajillions of dollars on their learning algorithms and
by np- 4y ago
> “becoming ever more reliable with time, as the algorithms learn“
I mean considering Amazon has put in kajillions of dollars on their learning algorithms and yet now in 2022 those algorithms still aren’t even smart enough to figure out trivial stuff, like that I’m probably not gonna buy a second air fryer immediately after buying a first one, and stop advertising them to me. What hope do we have for this new algorithm?
- BlargMcLarg 4y agoIt just showcases either the ones making the decision are naïve or they have an ulterior motive. They tried to justify it with the following footnote: >For example, Microsoft reports that the accuracy of its grooming detection tool is 88%, meaning that out of 100 conversations flagged as possible criminal solicitation of children, 12 can be excluded upon review and will not be reported to law enforcement; see annex 8 of the Impact Assessment. Of course, anyone could poke a dozen holes in this statement. "Possible", so there still needs to be a ton of review. Microsoft reporting on its own detection tool. Etc.
- wsc981 4y agoThe ulterior motive is gaining total control over citizens in EU member states. Which is also why the EU digital ID [0] will soon be introduced and a Euro CBDC will follow in the coming years. It's quite likely the Euro CBDC wallet will be linked (part of) the EU digital ID. I expect at some point the EU digital ID will be required to use when accessing the internet or perhaps when signing up for email, creating a Twitter account and such. This EU digital ID will EU member states will become a carbon copy of China. People will worry what they can say out of fear that they will not be able to spend their hard-earned Euro CBDC. --- [0]: https://www.politico.eu/article/eu-europe-digital-id/ https://www.politico.eu/article/eu-europe-digital-id/
- koonsolo 4y agoAs long as we have democracy, we can vote on any party we want, also on the ones that would take our own country out of EU. So unless you can show me how a move away from democracy can happen, we won't turn into china soon.
- raxxorraxor 4y agoDemocracy didn't really protect use from intelligence agencies sniffing communication data, so I don't see how it would protect anything. Not an intrinsic fault of democracy since the EU still has deficits here.
- 2000UltraDeluxe 4y ago> I expect at some point the EU digital ID will be required to use when accessing the internet or perhaps when signing up for email, creating a Twitter account and such. I'm not going into the debate about whether there's an ulterior motive, but current eID solutions in various EU countries are considerably less draconian than what you're describing. The current solutions are going to tie into the new one, after all. In Finland, as an example, we simply have a chip with a cert/key on ID cards. People without card readers or who don't want to use them are free to use their netbanking logins instead. Being able to use the same solution in other EU countries is of rather limited value to most, but it will ease up paperwork submission when travelling, moving to another member state, and so on.
- raxxorraxor 4y agoProblem is that if an ID system is ever established and becomes ubiquitous, a quick change in legislature can force online services to leverage the ID. Protection of intellectual property, CSAM, a pretext is easily found. Happens faster than you can blink and than the free internet in Europe would quickly die. The best protection against that is to have alternative ID systems or just anonymous usage if no transactions are involved. And a lot of services would be fine to use a state ID because they can tie their advertising to a real and unique person. To secure against that the smart move is to reject the little convenience here for longtime benefit.
- unionpivo 4y agoPretty much whole of EU already has some form of government ID (not sure if there are any exceptions, but if they are they are not many). People are used to them, so it's really just a matter of time, before it goes digital. I don't disagree with your observations, I just think its a forgone conclusion at this point. That said I think the value lies in creating new non centralized (and less popular) solutions. They will probably never have mass appeal, but at least they will be there for people who want/need them.
- raxxorraxor 4y ago
- raxxorraxor 4y agoThe little fact that Microsoft then has the possibility to read random interactions that have been flagged by its own blackbox should really induce doubt here.
- ClumsyPilot 4y ago> It just showcases either the ones making the decision are naïve or they have an ulterior motive Its the same AI coolaid silicon valley has been peddling for years and using it to justify various crap. So you should not be surprised
- bliteben 4y agoMaybe they aren't optimizing for conversions but instead for running out budgets? Amazon often has anti-consumer practices, that likely evolved that way because it improves bottom line, you'd think how hard they make it to search reviews or to filter reviews for a specific version of a product would be simple fixes, but likely testing has shown those things decrease sales.
- bryanrasmussen 4y ago>like that I’m probably not gonna buy a second air fryer immediately after buying a first one, this comes up all the time, but one of the primary indicators that {person} will buy a second {item} of type {X} increases when they have already bought one. However it does seem to me that I have never done this, even when I am unsatisfied with item, I soldier on for at least a year or two before saying aw screw it, perhaps the machines would be more impressive if they could recognize what people will and will not ever buy a new expensive item immediately again after buying one.
- np- 4y agoYeah that is a pretty naive algorithm though, especially if you can explain it in one sentence. After all that, that’s the best algorithm they got? If you buy one thing you might buy two. Pretty sure most of us could implement that in a line or two of code and save billions in research. And even if this whole research exercise was necessary to discover this fact, it’s still not a “smart” algorithm if there are a ton of false positives. And it’s still a relatively trivial use case—-if it comes to potentially miring someone in legal issues, I hope the “algorithm” can do better than that.
- ldng 4y agoI've been told (I don't know if it's) that showing you an ad for something you just bought reinforce your satisfaction and are thus significantly less likely to return the product. I really wonder if it's true.