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

U.S. government limits exports of artificial intelligence software

reuters.com

Post reply on HN
Print view View original

51–60 of 220 posts

Re: U.S. government limits exports of artificial intelligence software

#51

Post by tgsovlerkhgsel »

oefrha wrote: post #46
withinrafael wrote: post #25

Additional details are present in the unpublished rule (document 2019-27649 [1][2]). -- cut -- Geospatial imagery “software” “specially designed” for training a Deep Convolutional Neural Network to automate the analysis of geospatial imagery and point clouds, and having all of the following: 1. Provides a graphical user interface that enables the user to identify objects (e.g., vehicles, houses, etc.) from within geo…

So image recognition APIs like ones provided by AWS and GCP should be affected by this due to #2 and #3, no? Or the “provides a graphical user interface” part applies to all points? Edit: I overlooked “have all of the following”.

The main restriction is "“specially designed” for training a Deep Convolutional Neural Network to automate the analysis of geospatial imagery and point clouds".

Re: U.S. government limits exports of artificial intelligence software

#52

Post by incompatible »

extropy wrote: post #49
incompatible wrote: post #45

Earlier quoted context omitted.

The "graphical user interface" bit is weird. I suppose you are fine if the software is driven entirely by keyboard shortcuts.

IMO that's an indication that this was made for a specific piece of software and trying to limit the collateral damage.

Yes, it seems very specific.

Re: U.S. government limits exports of artificial intelligence software

#53

Post by sgt101 »

roenxi wrote: post #20

We're in an early and explosive growth stage of AI where well-established statistical knowledge is having an unreasonable effect when combined with computing power. I've yet to see any AI platform that is mindbending vs doing basic math with a multivariate normal distribution. The eyewatering stuff is the number of Hz of computing power being thrown into simulating Go games, etc, etc. Assuming that among 1.4 billion…

I'd be interested in how you would do NLP tasks such as those now done with transformers like gtp2 and Bert by working with multi variate normal distributions.

Re: U.S. government limits exports of artificial intelligence software

#54

Post by tomp »

_iyig wrote: post #39
astatine wrote: post #36

Unlike the space race, where Russia and the US were significantly ahead of everyone else, the field is far more level in AI. Arguably there will be areas where the US could even be behind in some areas. Wonder what the real world impact of this will be. Not much, I expect.

>Wonder what the real world impact of this will be. Not much, I expect. Here's a thought experiment I use to imagine the impact of AI: Imagine you've got a million people at your disposal. At zero cost and with no downtime, these people can remotely operate robots, understand text, interact using natural language, or classify objects in images, all with human-level intelligence and accuracy. Now what? Obviously there…

Now imagine doing crowd control. 10 frames per second, 100k people, if you need just 1 second to recognize a face, you’ve just saturated your 1m human-AI. The point of digital AI is that it can scale, almost indefinitely.

Re: U.S. government limits exports of artificial intelligence software

#55

Post by floatingatoll »

incompatible wrote: post #45
withinrafael wrote: post #25

Additional details are present in the unpublished rule (document 2019-27649 [1][2]). -- cut -- Geospatial imagery “software” “specially designed” for training a Deep Convolutional Neural Network to automate the analysis of geospatial imagery and point clouds, and having all of the following: 1. Provides a graphical user interface that enables the user to identify objects (e.g., vehicles, houses, etc.) from within geo…

The "graphical user interface" bit is weird. I suppose you are fine if the software is driven entirely by keyboard shortcuts.

Not so weird. Palantir put a GUI on Hadoop/whatever and that was enough to sell it to every government for citizen inspection. It’s the GUI that gets the contracts, not the technology alone.

Re: U.S. government limits exports of artificial intelligence software

#56

Post by sytelus »

withinrafael wrote: post #25

Additional details are present in the unpublished rule (document 2019-27649 [1][2]). -- cut -- Geospatial imagery “software” “specially designed” for training a Deep Convolutional Neural Network to automate the analysis of geospatial imagery and point clouds, and having all of the following: 1. Provides a graphical user interface that enables the user to identify objects (e.g., vehicles, houses, etc.) from within geo…

Gosh! There are like thousand github repos that do these stuff. This could become a tool to prosecute a lot of unwanted people like DeCSS.

Re: U.S. government limits exports of artificial intelligence software

#57

Post by mihaaly »

roenxi wrote: post #20

We're in an early and explosive growth stage of AI where well-established statistical knowledge is having an unreasonable effect when combined with computing power. I've yet to see any AI platform that is mindbending vs doing basic math with a multivariate normal distribution. The eyewatering stuff is the number of Hz of computing power being thrown into simulating Go games, etc, etc. Assuming that among 1.4 billion…

I sense that they will hurt themselves more with the imposed additional bureaucracy, hindering business and collaboration - the latter being a key aspect in efficient research, not to mention multinational companies already living on the area.

Also, how effective could this regulation be with so much knowledge and open source code already disseminated in the field?

Re: U.S. government limits exports of artificial intelligence software

#58

Post by oefrha »

tgsovlerkhgsel wrote: post #51
oefrha wrote: post #46

Earlier quoted context omitted.

So image recognition APIs like ones provided by AWS and GCP should be affected by this due to #2 and #3, no? Or the “provides a graphical user interface” part applies to all points? Edit: I overlooked “have all of the following”.

The main restriction is "“specially designed” for training a Deep Convolutional Neural Network to automate the analysis of geospatial imagery and point clouds".

And having a GUI. It appears you can simply export a binary framework (with or without source) and let the buyer build an interface on top of it. Building an interface is not terribly advanced work and can be done by fresh bootcamp grads even...

Re: U.S. government limits exports of artificial intelligence software

#59

Post by roenxi »

sgt101 wrote: post #53
roenxi wrote: post #20

We're in an early and explosive growth stage of AI where well-established statistical knowledge is having an unreasonable effect when combined with computing power. I've yet to see any AI platform that is mindbending vs doing basic math with a multivariate normal distribution. The eyewatering stuff is the number of Hz of computing power being thrown into simulating Go games, etc, etc. Assuming that among 1.4 billion…

I'd be interested in how you would do NLP tasks such as those now done with transformers like gtp2 and Bert by working with multi variate normal distributions.

Maybe I didn't make myself clear enough.

There was a time when a software engineer didn't get basic stuff. A time when languages like C were developed where there wasn't an associative data structure baked in for example.

It wasn't because associative data structures are a secret tech that requires great insight to uncover. It is because the field was new and people hadn't cottoned on to how basic and important having access to hash-maps is. Times moved on. Now basically all modern languages have hash-maps as a basic data type.

'AI' is in that early phase where the engineering world is still getting excited over stuff that will basic practice eventually. BERT and GTP2 are signs of how much computer power Google/etc's researchers have access to, not signs that the architectures are fundamentally complicated or somehow hard to work out if you live in China. AlphaGo for instance was breathtaking as a standalone project, but not hard to implement.

Re: U.S. government limits exports of artificial intelligence software

#60

Post by _iyig »

tomp wrote: post #54
_iyig wrote: post #39

Earlier quoted context omitted.

>Wonder what the real world impact of this will be. Not much, I expect. Here's a thought experiment I use to imagine the impact of AI: Imagine you've got a million people at your disposal. At zero cost and with no downtime, these people can remotely operate robots, understand text, interact using natural language, or classify objects in images, all with human-level intelligence and accuracy. Now what? Obviously there…

Now imagine doing crowd control. 10 frames per second, 100k people, if you need just 1 second to recognize a face, you’ve just saturated your 1m human-AI. The point of digital AI is that it can scale, almost indefinitely.

OK, then imagine 100 billion people. Scale isn't the point of the experiment - the point is bounding expectations based on probable (maximal?) capabilities.

Perhaps a post-Singularity AI will have wild capabilities beyond our comprehension, but that is outside the scope of this experiment.

Post reply on HN