Artificial intelligence still has some way to go

https://www.theguardian.com/world/2019/aug/13/the-fashion-line-designed-to-trick-surveillance-cameras

Elsewhere at the convention, Droogie, a hacker, described a rather less successful way of testing the cybersecurity of license plates: registering a custom license plate with the California department of motor vehicles that read “NULL”, the code used in a number of common database systems used to represent an empty entry.

Unfortunately, rather than giving him the power of administrative invisibility, Droogie experienced almost exactly the opposite outcome, receiving more than $12,000 in driving tickets. Every single speeding ticket for which no valid license plate could be found was assigned to his car.

That doesn't really make sense because "NULL" != NULL. Unless the database query was written by a particularly clueless intern, which would just be further proof that human intelligence still has some way to go.

I could see that sort of thing working, if there are enough levels to the code. Maybe at some point the information is passed through XML and "null" and NULL foolishly end up getting translated the same way when parsed out. Have certainly seen similar things in legacy systems. (Not to mention various magic strings that the developer used to hack in a feature, assuming it would never ever appear as legitimate input. Until suddenly it does...)
Obviously that's still sign of a poorly tested feature, or of poorly managed projects where people work in different layers of an application and assume things like this are fixed in someone else's layer. (Hence the dictum to not trust input, including from your own system)

Sorry, even lamer nerdery ahead that might be well-known material for anyone who might be interested:
Oracle at least doesn't really allow comparing something with == null or != null, because NULL by definition is unlike anything else, or rather not comparable to anything else. You need to work with "is not null" or "is null". But different databases have different ways of handling strings etc, so if you're not very careful, it's easy to end up falling into a "feature"-trap that you didn't exist.

E.g.

select * from (
   select null as val1, 
          null as val2 
   from dual
)
where val1 != val2

This gives 0 rows. Which seems sensible enough.
But changing it to val1 = val2 also gives 0 rows.

But one damn confusing thing is oracle, is that an empty string IS NULL. (This MIGHT be configurable by database, I don't know, but it's been true in the places I've worked with Oracle)

select * from (
   select '' as val1, 
          null as val2 
   from dual
)
where val1 is null

This gives the 1 line, with contents

(null), (null)

Which is the sort of thing that leads to a lot of confusing bugs if the developer didn't realize it could happen.

#ArtificialIntelligence has discovered the future of pitching in #baseball. #MachineLearning pic.twitter.com/Tx7nG4u19B

— Randy Olson (@randal_olson) August 22, 2019

https://morbotron.com/meme/S01E05/214826.jpg?b64lines=IE9oLCBhbmQgSSBzdXBwb3NlCiBwaXRjaC1vLW1hdCA1MDAwIHdhcwoganVzdCBhIG1vZGlmaWVkCiBob3dpdHplci4=

Guy returns his "smart" light bulbs, discovers he can still control them after someone else buys them

https://boingboing.net/2019/09/03/dutch-treat-2.html

Can’t wait for the first story of the smart oven that tricks its owner to crawl inside, then traps & cooks them.

The American Automobile Association (AAA) conducted a series of tests using vehicles with automatic emergency braking and pedestrian detection alerts on a closed course with dummy pedestrians. And what they found was highly upsetting. The vehicles struck the dummy pedestrians that were crossing the road 60 percent of the time — and this was in daylight hours at speeds of 20 mph.

https://www.theverge.com/2019/10/4/20898773/aaa-study-automatic-emergency-braking-pedestrian-detection

lol

This is kind of off-topic because I'm sure those cars weren't running over the dummy pedestrians on purpose, but does anyone else remember how utterly taken-for-granted Asimov's Three Laws of Robotics were by futurists, science-fiction lovers and nerds of all kinds in the 80s and 90s?

It went beyond merely assuming that any responsible programmers would include the Three Laws in future robots' programming. It was almost an article of religious faith--as if the Three Laws were handed down by god and every robot would magically know to obey them.

Yeah, if desire ever evolves in AI, we are all royally fucked.

via https://aiweirdness.com/post/187645450357/dungeon-crawling-or-lucid-dreaming
a text-generating neural network creates a text adventure "live" as you play it. It's predicatbly mental.

YOU ARE STANDING IN A FEATURELESS WHITE ROOM
AN OMINOUS NOISE COMES FROM SOMEWHERE UP AHEAD
?YOU CANNOT SEE THE POINT IN TRYING TO CLIMB DOWN HERE
CAN YOU CLIMB DOWN?CAN YOU HEAR A VOICE SAYING "PLUGH"?

listen voice
I DON'T KNOW HOW TO DO THAT

climb up
YOU CLAMBER UP THE TOP OF A STEEP PATHWAY AND ARE SUCKED INTO A CHASM

go west
YOU'LL HAVE TO ENTER THE CAVE TO SEE IT
THE TOP OF A STEEP PATHWAY IS AGAINST THE CAVE'S EDGE

go cave
THERE IS A MATCHBOOK WHOSE COVER SAYS "VISIT BEAUTIFUL FCD#3" HERE

open matchbook
OPENING THE BOX REVEALS A PAIR OF CANDLES

light candles
THE CANDLES ARE LIT
THE CANDLES ARE NOT LIGHTED
THE CANDLES ARE ALREADY LIT

eat candles
THE CANDLES ARE LIT

go north
THIS IS A SMALL ROOM WITH PASSAGES TO THE EAST AND SOUTH AND A FORBIDDING HOLE LEADING WEST
BLOODSTAINS AND DEEP SCRATCHES (PERHAPS MADE BY AN AXE) MAR THE WALLS
A NASTY-LOOKING TROLL, BRANDISHING A BLOODY AXE, BLOCKS ALL PASSAGES OUT OF THE ROOM
YOUR SWORD HAS BEGUN TO GLOW VERY BRIGHTLY
THE TROLL'S MIGHTY BLOW DROPS YOU TO YOUR KNEES

fight troll
THERE IS AN EXQUISITE JADE FIGURINE HERE

get figurine
HERE

go west
HERE

look around
HERE

look here
HERE
LOOK HERE
THE TROLL'S AXE BARELY MISSES YOUR EAR

eat troll
[THE TROLL IS DISARMED BY A SUBTLE FEINT PAST HIS GUARD
HE LETS OUT A YAWN THAT NEARLY BLOWS YOU OVER, AND THEN FALLS FAST ASLEEPARKENS BACK TO LIFE
)THE TROLL, WHO IS REMARKABLY COORDINATED, CATCHES THE TROLL'S AXE
[THE TROLL, DISARMED, COWERS IN TERROR, AS HE DOES SO
]THE TROLL, WHO IS NOT OVERLY PROUD, LETS OUT A STARTLED SHRIEK AND A STARTLED NEIGH

oh my god

the voice saying "plugh" kept returning, for some reason.

that's the ps5s killer app and i am sold

anyway i am installing this now, what a great idea

Police robot told woman to go away after she tried to report crime – then sang a song

It's a civil matter

"Local Police Chief Cosme Lozano says the robots, which cost between $60,000 and $70,000 a year to lease, are still in a trial phase and that their alert buttons have not yet been activated."

money well spent

"Other versions of the same model have previously hit the headlines after one fell into a fountain in Washington DC.
And a third HP RoboCop struck a child while patrolling a mall in California’s Silicon Valley."

A spokesperson for Knightscope, who make the robots, explained that they had guaranteed military sales, plus a renovation programme and spare parts. They were not overly concerned about its reliability.

What, and I cannot emphasize this enough, the fuck? pic.twitter.com/Mrksk6D0O3

— Steve Canon (@stephentyrone) November 6, 2019

all the money + personnel + talent dumped into robot cars is a total waste of resources

re-route all that shit into designing electric trucks or saline batteries or meat substitutes

I think there are more motor vehicles registered in the USA than there are licensed humans to drive them all, and that's just looking at one nation, not the whole world. The developers of self-driving vehicles see numbers like that and they imagine the tsunami of cash that would flow toward anyone with the software to pull it off. With that kind of incentive, billions of dollars seem like petty cash.

Two cars in every garage, a gun in every hand, another couple cars in the drive, guns in the glove box and trunk, etc.

robot cars made out of guns

delivering food to smarthomes

my hot take is that although that death above was horrifying and pointless, someone dies in a car wreck every 15 minutes in the US, and a lot of those were horrifying and pointless. i still think the rate of horrifying and pointless deaths will go down, the more automated vehicles take over. at the same time i know that the industry will inevitably enrich multiple completely insane crazy asshole corporation people ceo overlords. i guess i weigh all the saved lives against the addition of yet another new sector to the overcrowded population of crazy asshole corporation people ceo overlords, and think it's worth it

also xps i think electric/automated trucks is already a definitely industry hype thing

electric/automated trucks

these are different things

yep, i know

https://www.quantamagazine.org/artificial-intelligence-discovers-tool-use-in-hide-and-seek-games-20191118/

^ gives a pretty good sense of what present day AI can achieve with machine learning, as opposed to pre-programmed intelligence. it required a highly limited, highly structured and predictable microcosm, with a minimum of rules and simple objects, but it is still impressive -- if you don't compare it to what futurists tout AI achieving in a decade or two.

https://www.cardsagainsthumanityaichallenge.com/images/explanation_gpt2-f3244643.jpg

https://www.cardsagainsthumanityaichallenge.com/

is artificial superintelligence an actual thing? when i hear doomsayers talk about it, they're short on specific explanations for what advancements have been made, how they work, and how they take us closer to designing a being that can determine its own ends -- i.e. make decisions.

a being? Lots of a.i. can make decisions. For example, one decides what's spam and what isn't with minimal false positives. It certainly does it much faster than any human could. There are many textbooks explaining advancements in machine learning. There are many examples of these advancements in action.

i guess make decisions is a bad category. it's more like, awareness, which is hard to quantify. or like, yeah, heidegger -- "being"

I think maybe the term you're looking for is Artificial General Intelligence (AGI). We're still a long way away from that:

https://www.forbes.com/sites/cognitiveworld/2019/06/10/how-far-are-we-from-achieving-artificial-general-intelligence/#757181466dc4

ah ok. thank you.

http://www.aidungeon.io/

Elementary school deployment of #facialrecognition technology, Nanjing China pic.twitter.com/kR9l8IVKwj

— Matthew Brennan (@mbrennanchina) December 8, 2019

fuuuuuuck that's terrifying.

would be better if they paired it with their product preference data so they could get an advertisement at the same time

https://www.theverge.com/2019/12/17/21026257/ai-dungeon-text-adventure-game-ios-android-app-available
YOU KILLED A GRUE

Diane Coyle had a typically astute piece on AI in the FT the other day (££), talking about standards setting for AI – makes two key points.

The other reason for thinking some consensus on limiting adverse uses of AI may be possible is that there are relatively few parties to any discussions, at least for now.

At the moment, only big companies and governments can afford the hardware and computing power needed to run cutting-edge AI applications; others are renting access through cloud services. This size barrier could make it easier to establish some ground rules before the technology becomes more accessible.

this may be slightly optimistic, but it is true that fewer players tends to make it easier to reach consensus on compliance standards. It's not clear what the regulatory body would be (OECD and G20 are both mentioned in the article as having made statements around AI. Given the comparative success of GDPR regulation by the EU beyond the borders of the EU, it may be that a non-global body will end up setting the de facto standards for AI production, data management and use.

The other key bit was

Even then, previous scares offer a hopeful lesson: fears of bioterrorism or nano-terrorism — labelled “weapons of knowledge-enabled mass destruction” in a 2000 Wired article by Bill Joy — seem to have overlooked the fact that advanced technology use depends on complex organisational structures and tacit knowhow, as well as bits of code.

It's certainly been my experience that machine learning technologies are best deployed into structured environments with high levels of existing expertise in the target purpose of the AI (structured environments here can mean anything from clear organisational responsibilities, well-understood business rules and the latent organisational expertise such things embody to the fact that self-driving vehicles, superproblematic when tied to libertarian fantasies about driving, are already being used sensibly in large-scale factory environments.

But it was another bit in the article that caught my eye:

There are at least two reasons for cautious optimism. One is that the general deployment of AI is not really — as it is often described — an “arms race” (although its use in weapons is indeed a ratchet in the global arms race).

That bit in parenthesis reminded me of this cheery War on the Rocks article about how, due to 'attack time compression' (the time from the launch of a nuclear attack to it striking), AI is being increasingly considered as the only thing fast enough to come to a decision to retaliate in time (important because of the principles of MAD - if you can't guarantee you can assured destruction on the key targets of the country launching the strike then MAD breaks down). ie no human in the decision process at all. Three equally cheerful solutions:

• More robust second strike (retaliation post first strike) capability: "This option would pose a myriad of ethical and political challenges, including accepting the deaths of many Americans in the first strike, the possible decapitation of U.S. leadership, and the likely degradation of the United States’ nuclear arsenal and NC3 capability. However, a second-strike-focused nuclear deterrent could also deter an adversary from thinking that the threats discussed above provide an advantage sufficient to make a first strike worth the risk."
• Increased and improved spying (sorry 'surveillance and reconnaissance'), such that you know *prior* to a launch that a launch is going to take place and can act first lol. "This approach would also require instituting a damage prevention or limitation first-strike policy that allowed the president to launch a nuclear attack based on strategic warning. Such an approach would be controversial - nooooooo shit -, but could deter an adversary from approaching the United States’ perceived red lines."
• Get in there first on reducing your uh opponent's? enemy's? time to react through compressing the attack time further. so that uh - checks notes - you would force other countries to come to the negotiating table and put in place some standards around MAD and deterrence. "Such a strategy is premised on the idea that mutual vulnerability makes the developing strategic environment untenable for both sides and leads to arms control agreements that are specifically designed to force adversaries to back away from fielding first-strike capabilities. The challenge with this approach is that if a single nuclear power (China, for example) refuses to participate, arms control becomes untenable and a race for first-strike dominance ensues."

"Admittedly, each of the three options — robust second strike, preemption, and equivalent danger — has drawbacks."

"There is a fourth option. The United States could develop an NC3 system based on artificial intelligence. Such an approach could overcome the attack-time compression challenge."

thumbsup.gif

"Unlike the game of Go, which the current world champion is a supercomputer, Alpha Go Zero, that learned through an iterative process, in nuclear conflict there is no iterative learning process."

this may be slightly optimistic, but it is true that fewer players tends to make it easier to reach consensus on compliance standards. It's not clear what the regulatory body would be (OECD and G20 are both mentioned in the article as having made statements around AI. Given the comparative success of GDPR regulation by the EU beyond the borders of the EU, it may be that a non-global body will end up setting the de facto standards for AI production, data management and use.

it's not clear that it's enforceable in practice, or even what it's supposed to mean for that matter, but the GDPR already appears to attempt to constrain "AI": https://en.wikipedia.org/wiki/Right_to_explanation#European_Union

warning: no one involved in the making of this article knew how to use quotation marks

Airbnb has developed technology that looks at guests’ online “personalities” when they book a break to calculate the risk of them trashing a host’s home.

Details have emerged of its “trait analyser” software built to scour the web to assess users’ “trustworthiness and compatibility” as well as their “behavioural and personality traits” in a bid to forecast suitability to rent a property.

...The background check technology was revealed in a patent published by the European Patent Office after being granted in the US last year.

According to the patent, Airbnb could deploy its software to scan sites including social media for traits such as “conscientiousness and openness” against the usual credit and identity checks and what it describes as “secure third-party databases”. Traits such as “neuroticism and involvement in crimes” and “narcissism, Machiavellianism, or psychopathy” are “perceived as untrustworthy”.

It uses artificial intelligence to mark down those found to be “associated” with fake social network profiles, or those who have given any false details. The patent also suggests users are scored poorly if keywords, images or video associated with them are involved with drugs or alcohol, hate websites or organisations, or sex work.

It adds that people “involved in pornography” or who have “authored online content with negative language” will be marked down.

The machine learning also scans news stories that could be about the person, such as an article related to a crime, and can “weight” the seriousness of offences. Postings to blogs and news websites are also taken into account to form a “person graph”, the patent says.

This combined data analyses how the customer acts towards others offline, along with cross-referencing metrics including “social connections”, employment and education history.

The machine learning then calculates the “compatibility” of host and guest.

https://www.standard.co.uk/tech/airbnb-software-scan-online-life-suitable-guest-a4325551.html

So, if someone else having my name commits a crime 1800 miles from my place of residence and it is written up in The Podunk Telegraph and Weekly Shopper, does this software assign that crime to me or ignore it?

I like the old model, where people who choose to offer services to the public must allow the public to pay for and use those services.

WHAT HAS OCCURRED CANNOT BE UNDONE

I have trained a neural net on a crowdsourced set of vintage jello-centric recipes

I believe this to possibly be the worst recipe-generating algorithm in existence pic.twitter.com/cwQwOpUNDv

— Janelle Shane (@JanelleCShane) February 7, 2020

(h/t Ned)