Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

We were promised Strong AI, but instead we got metadata analysis

calpaterson.com

Post reply on HN
Print view View original

101–110 of 427 posts

Re: We were promised Strong AI, but instead we got metadata analysis

#101

Post by Causality1 »

carrozo wrote: post #73
soarfourmore wrote: post #49

The following quotes are fairly interesting and ironic: > Larry Page and Sergey Brin were originally pretty negative about search engines that sold ads. Appendix A in their original paper says: >> "we expect that advertising-funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers" > and that >> "we believe the issue of advertising causes enough mixed incentives…

Everybody’s got a plan until they’re punched in the face (with hundreds of billions of dollars).

Yep. Money is the one unpatchable zero-day for every platform and service on earth.

Re: We were promised Strong AI, but instead we got metadata analysis

#102

Post by marcosdumay »

cbm-vic-20 wrote: post #55
fouc wrote: post #2

> There are woolly intimations that self driving cars will read roadsigns to work out what the speed limit is for any stretch of road but the truth seems to be that they use the current GPS co-ordinates to access manually entered data on speedlimits. I actually didn't know Tesla cars relied on GPS + map data w/ speed limits until recently. What a disappointment, apparently it causes all sorts of issues with sudden br…

It would be cool if the signs could have a low density error correcting code on them (maybe even only using a paint that can be seen in infrared) that would let the computers have more confidence in what they're seeing.

You mean some codes like a standard shape and color, colored borders, standard fonts, and standardized messages?

Re: We were promised Strong AI, but instead we got metadata analysis

#103

Post by bildung »

cblconfederate wrote: post #37

The problem is that google created a negative feedback loop with the web. People have a tangible interest to game google, which worsens the quality of their AI datasets, which makes their AI suggestions terrible. There was the expectation that google's AI suggestions and info boxes would improve over time, but i 've noticed them getting worse and consider them a permanently broken gimmick now. They probably have the…

I've noticed something similar (garbage in, garbage out, essentially) with translation services by google and co. People use these services to translate their sites, which then get fed into these models. There are a bunch of translation errors of terms of trade that are endemic in the german/english translations and apparently originate from these wrongly-trained models, which are then used to build new translations.

Re: We were promised Strong AI, but instead we got metadata analysis

#104

Post by PoignardAzur »

jacobolus wrote: post #86
cblconfederate wrote: post #37

The problem is that google created a negative feedback loop with the web. People have a tangible interest to game google, which worsens the quality of their AI datasets, which makes their AI suggestions terrible. There was the expectation that google's AI suggestions and info boxes would improve over time, but i 've noticed them getting worse and consider them a permanently broken gimmick now. They probably have the…

> google created a negative feedback loop with the web Friendly note: this may be a "negative" (i.e. bad) effect, but it is not a negative feedback loop. A negative feedback loop is a part of a system that self-corrects back toward a stable position. https://en.wikipedia.org/wiki/Negative_feedback

I think OP meant "the system always gets back to the status quo where suggestions are near-useless", hence a negative feedback loop.

Re: We were promised Strong AI, but instead we got metadata analysis

#105

Post by bjourne »

virgilp wrote: post #9
balia wrote: post #5

This seems it will be applied to self driving technology in a huge way. All of these companies doing advanced AI vision detection, classification etc... they’re really hard problems. But the whole challenge will eventually become nullified when every single road sign, landmark, and the road itself are tagged internally with metadata. Instead of trying to decipher how does this PNG of a speed limit sign translate into…

My car already does a pretty good job of showing the speed limits (and it's definitely based on computer vision/ works offline and in places where map metadata is poor). And it doesn't have "self driving", just a limited "pilot assist". Not sure why reading road signs is used as and example of "extremely hard thing to do" - there are other way harder things that cars can't currently do. E.g. figuring out the right sp…

How does your car handle snow-covered road signs?

Re: We were promised Strong AI, but instead we got metadata analysis

#106

Post by JohnJamesRambo »

soarfourmore wrote: post #49

The following quotes are fairly interesting and ironic: > Larry Page and Sergey Brin were originally pretty negative about search engines that sold ads. Appendix A in their original paper says: >> "we expect that advertising-funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers" > and that >> "we believe the issue of advertising causes enough mixed incentives…

https://imgur.com/t/funny/fjqdCZz

I like the whole quote. There’s one more gem in the middle. It shows they were good once. That they know. Any chance it can return?

Re: We were promised Strong AI, but instead we got metadata analysis

#107

Post by lifthearth »

Causality1 wrote: post #101
carrozo wrote: post #73

Earlier quoted context omitted.

Everybody’s got a plan until they’re punched in the face (with hundreds of billions of dollars).

Yep. Money is the one unpatchable zero-day for every platform and service on earth.

There is a patch thanks to contributor Marx but everyone keeps pressing "remind me later".

Re: We were promised Strong AI, but instead we got metadata analysis

#108

Post by jcfrei »

criddell wrote: post #71
Mediterraneo10 wrote: post #45

Earlier quoted context omitted.

Google page ranking depends on how long people spend on the webpage, so if you have to wade through longform text, it mistakenly assumes that the content is quality. Note that most recipe blogs are fake. Someone wrote the original content long ago. Then, someone else hired a copywriter off a freelancing platform to change the words of the text just enough to avoid copyright violation, then put up a new copycat websit…

> Google page ranking depends on how long people spend on the webpage, so if you have to wade through longform text, it mistakenly assumes that the content is quality. Is this true? I have a hard time believing it because it's a pretty naive assumption. A site that lets me get what I'm after quickly is generally going to be what I want.

Google analytics is running on lots of pages. Would be pretty surprising if they didn't use the insights they get from there, such as time spent on the site, buttons clicked, how far they scrolled etc.

Re: We were promised Strong AI, but instead we got metadata analysis

#109

Post by sseagull »

soarfourmore wrote: post #49

The following quotes are fairly interesting and ironic: > Larry Page and Sergey Brin were originally pretty negative about search engines that sold ads. Appendix A in their original paper says: >> "we expect that advertising-funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers" > and that >> "we believe the issue of advertising causes enough mixed incentives…

Wasn't this also the story of a dating site (whose name escapes me. OKCupid? PoF?)? Original owner wrote an article about how paying for a dating site is a bad idea. Money is offered, article disappears.

Searching is failing me at the moment

edit: Was OKCupid: https://www.themarysue.com/okcupid-pulls-why-you-should-neve...

Re: We were promised Strong AI, but instead we got metadata analysis

#110

Post by jeffbee »

hoppyhoppy2 wrote: post #91
darkerside wrote: post #14

Earlier quoted context omitted.

Maybe it uses both? My Honda can tell me the speed limit, and I'm sure it's not using the internet to do it.

>I'm sure it's not using the internet to do it. How sure are you? >Daniel Dunn was about to sign a lease for a Honda Fit last year when a detail buried in the lengthy agreement caught his eye. Honda wanted to track the location of his vehicle, the contract stated, according to Dunn — a stipulation that struck the 69-year-old Temecula, Calif., retiree as a bit odd. [...] >There are 78 million cars on the road with an…

>> I'm sure it's not using the internet to do it.

> How sure are you?

Pretty sure. The Honda Sensing display of the local speed limit changes on the dashboard at the instant you pass the sign, it's too exact to be using GPS. Nobody has geocoded the location of every sign in America, that would be too much work. Also, it's often wrong but in ways that you would be wrong if you were just reading the signs. For example, it will switch to 55 MPH speed limit in a 70 MPH zone on the interstate after it sees a sign that is intended for trucks only.

Post reply on HN