Hacker News Reader

Comment by shanghaiaway | Hacker News Reader

Artificial Intelligence does not exist. What we have is Artificial Stupidity.

A charitable reading of your post is that current AI, if judged against an average human personal assistant, would make many more "stupid" mistakes than the human. It would maybe forget less explicit stuff, but it's going to be a long time before AI stops us from going in a meeting having missed a button (or recognize which meetings are important enough to warrant a physical inspection of the participant). Or says, "hey before you commit to that meeting tomorrow, don't forget about this other thing that's maybe more important."

Anyone who's ever had a real personal assistant any good at the job would very quickly look at the best AI as the faint shadow of the real thing. And it's going to be that way for a while. Forever? Probably not, but maybe.

I'd argue that AI will sooner replace many legal and accounting jobs before it replaces good personal assistants. Just like how we thought chess is more difficult than walking and the exact opposite turned out to be true.

Even for the relatively constrained case of booking urban travel that's not 90% the same trip over and over, there's a pretty complicated interplay of price, schedule, convenience of location, quality of hotel, participation in loyalty programs, personal time, etc. etc. [ADDED: And just personal preference.]

A longtime personal assistant who has your preferences really dialed in can probably come close. But given how much I sometimes end up fussing over even my own travel, it's really hard to see virtual assistants getting there anytime soon.

> A longtime personal assistant who has your preferences really dialed in can probably come close

If they don't come close, they can safely filter out 90% ahead of time and present you a table of the remainder, saving you lots of time.

On the other hand, more likely that digital assistant was made by the booking company and is trying to upsell you with high margin items.

I think if we can let user control algorithms themselves, to a certain extends it does really help us to gain higher quality of information.

e.g sentimental analysis , fake news filter ...

I know there is a certain sense of hypocrisy in me saying this but maybe we don't need algorithms for everything.

Users can self-select. We can always just yell at Timmy who shares too much or outright unfollow him on Facebook if we don't want to see what they share.

I suggested that just because someone posts something on their Facebook wall, it doesn't mean I will see it. (This was around 2009). People often looked like I was saying something insulting. I think they'd understand things the other way around but I guess they don't see themselves as noise. There is some user education that the network needs to do.

In any case, be it Facebook or SharePoint or 37 signals, there should be zero obligation for anyone to look at things you post.

If you are having a baby shower and don't invite me personally and just write on your wall, that means you don't want me there.

I still believe in a 100% non-intelligent newsfeed. Ben Evans is wrong and I'd even say malicious for dismissing the idea of a reversed chronological newsfeed.

The only reason this is a problem is Facebook desires to inflate the number of connections. Let people cull their newsfeed and they will. Set expectations for posters that their posts may not get to everyone and they will understand as well.

Hm. Facebook not using algorithmic feeds tuned to your liking is hardly a reason to rule them out entirely. People’s eyeballs keep coming back, so Facebook’s news feed works great for Facebook. It’s an incentive or strategy issue more than a technology issue.

I agree. Facebook's News Feed is a culmination of years of research in retention and attention management. While that isn't really what most users want, it's exactly what Facebook has optimised the News Feed for.

We'd all benefit from a very different system which would structure things as we need it to, but that's sadly not what our economy incentivizes corporations to be right now.

Will the learning curve of "self-select" too steep to certain users ?

There's a hard limit to what I as a user can reliably express in terms of explicit rules. I have a cousin I was pretty close to growing up. She mostly uses Facebook to post anti-vaccine nonsense and occasional bits of thinly-veiled pro-Trump propaganda and BS "stories" from the alt-right. I "unfollowed" her a long time ago, but it would have been nice to see all the updates she posted when her dad had a stroke.

I need an algorithm for that. There's no good way for me to say, "I don't want to see all the stupid shit she posts, but please show me important stuff" because I lack the ability to rigorously define "important stuff". And it's unreasonable to treat her unwillingness to personally text those updates to every single person she knows.

Can't you just rigorously define the unimportant stuff? Like that which is related to vaxxiinwa, Trump or politics in general?

Not really. I could spend time and energy building filters that matched current known patterns, but those filters probably won't be very good. Suppose instead of a stroke it had been a gun accident. I'm not confident I could successfully anticipate and build a filter good enough to recognize that it wasn't a political post about the NRA and gun rights.

Ultimately, that line of thought ends with me developing my own algorithmic feed, which I don't really want to do and isn't scalable anyway.

Reply on news.ycombinator.com