Friday 21 June 2019

Unsettling AI, Right Here & Now

I'm sure we are all fully seized with the potential of AI and related developments, positive & negative, and have all read what internet uses the Chinese intend for it.  Much of it seems a little way in the future.  But here's a first-hand tale from this week.

I belong to a UK-based organisation that has a branch in the US.  I correspond by email very frequently with the UK Hon Secretary, we'll call him Martin; and only occasionally (maybe once a quarter) with the US Hon Sec whose name, coincidentally, is also Martin.  Recently I was told, verbally only and as a piece of breaking news not generally disclosed anywhere, that Martin (US) proposed to float an idea for a new type of activity involving his US members - and (before he told anyone else) what did we think of it?  

Having slept on the idea (but not communicated it further) I emailed the person who'd relayed it to me that I thought it was a good one.  I referenced 'US members' but no names.  I then hit the cc button to Martin (UK), something I'd done a thousand times before, because he'd been party to the original verbal discussion.

My email system immediately flashed up: "Did you mean Martin (US)?"

FFS !  If its algo was (as one might expect) based on past behaviours by volume, and indeed context at a basic level (i.e. comms within the organisation in question), the simple probabilities were very heavily stacked in favour of my intending it to be Martin (UK), as I'd actually intended and clicked.  But presumably it had "seen" a reference to US members, and "knew" Martin (US) was associated with US members; or "seen" an email on the same subject between Martin (US) and my addressee ...   This is clever and useful, but ... well.

I recall the first time when I was "asked" whether I meant to append an attachment when I'd mentioned such a thing in the text, but hadn't appended anything before hitting Send.  That seemed quite clever at the time.  Things seemed to have moved on a bit ... 

And this is all real-time stuff with (obviously) no human intervention whatsoever.  Cunning bastards.  God alone knows what the real bastards (China, FB, ...) are doing with this stuff.

Any stories you'd care to share-and-scare?



BlokeInBrum said...

I was chatting with a colleague yesterday about traffic cameras and number plate recognition. Birmingham is due to impose a clean air zone in the city center soon, which will naturally need a whole lot of very sophisticated cameras and I.T. systems to implement.
As in London, everyone traveling in and out of the city will be tagged and tracked.
As this technology becomes ever cheaper and more ubiquitous and more linked-up, it will effectively mean the end to privacy.
Link it up with facial recognition or AI that can recognize people simply from their stature or gait and I can easily envision a future system in which the activities of people can be tracked from cradle to grave.
Can you seriously imagine that Government wouldn't use this info to entrench its power and use it to control people?
Wouldn't like to be a political dissident in the future. I can easily see us heading down the Chinese social credit/control way.

Anonymous said...

Not really AI, but I see (by its recommendations etc) that Youtube/Google knows who I am even in a private browsing session, when I'm not signed in to anything on any open tab. IP address plus browser fingerprinting plus cookies I guess.

Similarly an Android phone scans my incoming texts and suggests quick replies.

Google's pretty scary, though most young people don't seem to see it. And it's nowhere near as good a search engine as it was way back in 2005!

Quite a few search engines have been neutered over the years. The BBC and Guardian both used to be pretty good, now very poor. Deliberate strategy.

Anonymous said...

BlokeInBrum - a lot of ANPR in the UK now, not everywhere but enough that given a registration of a car the police can tell you "it was seen two hours ago heading out of Wisbech". In London or Brum the detail will be a lot finer.

The good news is that it won't automatically flag you up for a police stop or a letter through the door if your MOT ran out a week ago, unless you're on a list of bad boys or bad cars.

I wonder if the location of ANPR cameras is publicly available?

Anonymous said...

You have to assume there will be an ANPR camera soon on every motorway bridge. Why wouldn't they?

Anon 11:08, any thesis on why BBC/Guardian have neutered their search?

Jag said...

I have three different browsers.
Those three give different ‘ google only’ searches, for the same thing.
Each has been tailored to try and guess what I might want.

So, first search in each one for Roses.

1. Rose garden
2. Stone Roses cd
3. War of the Roses


All three are also trying to sell something. Far more now than years ago when asking usually brought up information first.

Raedwald said...

I'm fascinated as to how automatic speed limiters - compulsory in new cars from 2022 in the EU - will work here in Austria. They work both by reading speed signs and GPS. As for GPS, half the country here is at 45deg and altitude changes can also be very rapid - meaning even my own marine standard better than +/- 3m GPS sets are frequently way off, particularly on forest roads. Speed signs are just as problematic - the speed limits of side roads being posted only on main roads, same size and format as main road speed limits but with a small plate beneath. So on the drive to the Spar a couple of klicks away, the signs go 50 - 70 - 30 - 30 - 50 - 70

Now if they've already developed an AI brain cheap enough to be fitted to production cars that can work these things out for not one but twenty-seven national data sets of road signage and speed conventions, I am seriously impressed.

Al said...

Instructing Google to do something has different results week by week e.g. navigate to (sometimes starts navigation, note to self sometimes goes to Google keep sometimes to other apps).

Anonymous said...

"any thesis on why BBC/Guardian have neutered their search?"

“Who controls the past controls the future. Who controls the present controls the past.”

People like the Biased BBC site used to do news searches for "far-left" and "far-right" on BBC news and see the enormous disproportion between the small number of news items mentioning "far-left" and the massive numbers for "far-right".

(Far left organisations and individuals generally being called "campaigners", as in "campaigners for social justice".)

Similarly it was easy to compare the news coverage of racist murders where the victim was black, and the far smaller coverage of those (the majority) where the victim was white. Not very easy now.

More generally, Brexit and Trump in 2016 shocked elites on both sides of the Atlantic - hence the ongoing purges/demonetisations at Youtube/Facebook/Twitter/Paypal and the struggle to re-establish control of the news narrative.

BlokeInBrum said...

Yes, I remember the times when Altavista rules the roost and Google was the young upstart.

Googles homepage was famously simple in order to minimise loading time and resource use. This was back when they were mean and lean and were trying to usurp the old favourites by delivering more accurate search results and indexing a much larger breadth of web sites.

Fast forward to today and Googles homepage is a constant propaganda assault against conservatives. They diddle their web results constantly to push a leftist agenda. Most people now live in an individually tailored web bubble created by the likes of Google and Facebook, who now have gathered an enormous amount of data on every individual and who are in the business now of influencing people rather than helping them.
This is only going to get worse in the future.

During the last Presidential election, Trump was media savvy enough to engage directly with voters, thick skinned enough not to pay attention, or give ground to the conventional media.

The conventional media and political operators were caught flat footed and didn't have a response to him or the army of meme lords who ridiculed the left on the likes of 4chan and it was sufficient to pull him over the line.

Everything happening since then is a response to prevent the same thing happening again. Every Hollywood film a promotion of Diversity and a celebration of the end of 'Pale Stale Males'. Every HR department in every major corporation wedded to the notion of diversity, promoting and pushing anyone who isn't a white male, talent be damned. Every major educational institution infested with SJW's pursuing with absolute fanatacism the end of any influence by people holding right of center views . The Democrats absolute determination to prevent Trump from securing the borders, as they know that every immigrant, legal or otherwise will vote for them.
Online, every major platform is defenestrating anyone who doesn't kneel at the altar of diversity and leftism.
You would think , given the influence exerted by these people, that they were a large majority. But this isn't the case. They are a small minority who have insidiously inserted themselves into positions of power.
You only have to look at our own situation wrt. Brexit to see how a small minority can usurp the will of the people and get their own way.

BlokeInBrum said...

Sorry, bit of a rant, the last post.

My point is that the State & major corporations and the media can control everything you learn (at school/uni), everything you see on the telly, everything you hear on the radio and, more and more, everything you see, hear and do online.
The Internet was once the last bastion of freedom, where people could seek out and find a genuine diversity of views and information and communicate with whom you wanted to.
Not so any more.

The longer this goes on, and the better they get at controlling it, the smaller the chance of future generations having any real freedom. In fact they won't even be able to comprehend the freedoms they've lost.

Manipulation is the name of the game now.

Nick Drew said...

We like your rants, BiB - no apols necessary

Anonymous said...

Yeah, I shared that link with my friends. I was amazed by the reaction I got. Basically "that site is well known for sharing links to aggravate anti-muslim feeling"

Totally missed the point and jumping to the usual knee-jerk "poor muslims" response.

My issue was that the government indulges in rampant propaganda, and as evidenced by the reaction, it bloody works.

NigeleastofBRUM said...

Mr Drew, I don't like to post comments, normally, because normally reading your wisdom, I feel very much like a small boy on the edge of a grown up conversation. However, this is one that I do have personal experience of. I read on a daily basis, a couple of websites that are agglomeration of news media stories that the BBC and their ilk would not touch with a barge pole, and that the great and the good would really prefer that us oiks were unaware of. Last year I booked a flight to Dulles, Washington, via Dublin from the same IP address that I access these sites from. Very peculiar things happened. I could not print out the boarding card from Dublin to Dulles, in Dublin I was given a boarding card and Dear Lady Wife and I were given two seats in the middle of an otherwise empty row. My boarding card was marked SSC, which I later discovered meant special security check. The check was everything short of the rubber glove treatment. Then immigration/passport check prior to the flight was much more rigorous than the usual hard stare. Then, when we got to the gate, the airline employee, after checking our passports and boarding cards on his screen, called out our names and seat numbers to a young, fit looking chap who was standing at the side of the tunnel down to the aircraft, he wrote this information down in a notebook. I thought this was all the usual airport bugger factor until I saw that we were sitting in an empty row on the plane, in glorious isolation. I pondered this for most of the flight and came up with the idea that it might be an imbecillic algorithm at either GCHQ in Cheltenham or the NSA in Maryland and the young fit bloke was the air Marshall. I subsequently mentioned this to acquaintances and had the theory dismissed as preposterous. THEN, this January I transited LAX on a trip to New Zealand and on both the outward and return through the US, I was targeted for the same security treatment. What is it they say? once is happenstance, twice is coincidence and the third time the buggers are out to get you. I'll let you know what happens the next time I go to the States.

Wildgoose said...

With reference to BlokeinBrum's comment: "Wouldn't like to be a political dissident in the future. I can easily see us heading down the Chinese social credit/control way."

Welcome to the Future! It is already here.

If you're not Left-wing you are a dissident right here and now. If your social media posts gain any notice and you don't follow the typical left-wing narrative then "social justice warriors" and their willing compatriots in the main-stream media will actively harass you, dox you, and try and get your employer to sack you. And of course get your Youtube videos demonetised and you banned from Facebook, et cetera.

The only difference is that the Chinese Social Credit system has clear rules to follow, applies to everybody rather than being partisan, and is not suddenly arbitrarily applied because of a deliberately misinterpreted comment made on a website several years ago.

The Chinese Social Credit system appears to be trying to engineer the creation of a High Trust society for the long-term benefit of all as well as also conveniently cementing the Communist Party in power.

Our own version? Just the opposite. We have a system that appears to be deliberately undermining social cohesion and atomising our society into competing groups. And weaponising interactions between those groups.

I am a life-long opponent of "multi-culturalism". I prefer to celebrate what different people have in common rather than emphasise how they differ. And yet all you will hear (or be told) is that I oppose multi-culturalism with no explanation but plenty of inference that I am some kind of far-right racist.

And with advanced data-mining it is increasingly easy to tie together all aspects of a persons life. Not just for advertising and selling purposes but also for identifying political opinions in order to find the most appropriate reason for shaming/bullying and attacking them in order to keep them in line.

It's not the Intelligent Algorithms we have to worry about. It's who is in control of them.

Nick Drew said...

N.e.o.Brum - don't ever hesitate to comment here! (well, so long as you don't attract the attention of ...) One day they'll close us down, but not just yet

I fear that if you've got yourself on the US watch-list you might be there for quite a while. A business acquaintance of mine, upright US citizen and highly decorated Vietnam veteran, once made a sarcastic comment to a US border guard at airport immigration (he was tired and pissed off about something bureaucratic)

he immediately got frogmarched off for a large helping of 'extra bureaucracy'; and manifestly they've stuck something fairly permanent on his records because he's had the same treatment frequently when re-entering the USA after travel abroad. No sense of "humor", these types

I am inclined to agree with Wildgoose (and your near-neighbour B.f.Brum). Remember the days when politics on the www meant Guido, ConHome and Ian Dale?!

It all started to go wrong when C@W was silently blacklisted from BBC blogs' BTL comments ... (true)

andrew said...

For me facebook beats all for spooky ai associations. OH is a big facebook user and it worked out we were in copenhagen (rammstein) and wanted us to buy a copenhagen card. She had not posted anything about copenhagen at that time.

It is not that clever, we are in berlin now and it has not worked that out.

AI is a thing that is the frontier of what you think machine intelligence is.

Back in the 90s that was context sensitive help.
Then the post office managed to get addresses recognised.
Now google and ibms attempts. (Deliberately ignoring tesla)

Yes, one plays go well and the other won at jeopardy but underneath it when you come to commercial implementations outside a very narrow problem space (phillips screwdrivers work great with phillips screws) now so, apparantly in the case of watson.

This is not surprising.

These wins have been gained using immense amounts of computational power, yes all v.v. impressive but to get traffic across the english channel you can build a tunnel or if you are solving using ai techniques, fill the english channel in.

Underneath it all it is inevitably just a massive pile of if - then - else (the google attempt seems to be stateful and so has some promise)

In another sense an ant has about 300 neurons and i (from what i know) do not think there is ai out there that can emulate an ant (vision,movement,comms,eating etc)

DeepThought said...

Cityunslicker has scored a <5> on the appropriate sentiment scale.
This is an improvement from <3>
Congratulations, your black mark is now a grey mark. Please keep the hard work up. Once you score 7 you will get a gold star and will form part of the inclusive.

Anonymous said...

We're far from actual AI, it's machine learning. It's clever stuff, but the real danger is governmental trust and belief in what is being peddled.

Example - some well meaning types in the US decided to train a system to deal with justice. They were a bit horrified at the outcome, as it turned out their system was really rather racist. If a black person and a white person committed the same crime, the white person could expect more lenient treatment.This was because it had been trained using the results of the US justice system. Garbage in, garbage out.

Meanwhile tech-ignorant politicians and journalists are salivating at the claims of companies, bullshit-detectors left at home. Witness to clapping over Musk claiming they'll be producing JohnnyCabs in a couple of years. No Elon, you won't.

I'm also more concerned about FB, I think it was ND who claimed he had no FB account. Sorry, but you do. They buy financial data, phone data and have 'accidentally' hoovered up contact details of their app users over the years. So if you've spent using a credit card, owned a mobile phone in your own name or someone with your number has had the FB app on their phone then they've got a shadow profile of you which gets used, along with the likes of the FB pixel on websites, to sell your data profile to advertisers. An actual FB account is just another data source to build up your pre-existing profile.

@Wildgoose - what you are seeing ties directly into what we want as a society. Freedom is messy.

Do we really want to go the Chinese route? You're reliant on whoever is in charge being on "your" side to a much greater extent than we are in that case. What we see with historical tweets is a generation with a complete lack of context, no ability to comprehend individual growth.

Went watching a very left-wing, feminist, anti-capitalist band the other week (I like their music, not necessarily their politics). They had the same concerns, genuinely pissed off at other left-wingers who use historical social media to score points.

So it isn't a political wing thing - witness Guido doing the same thing with a different demographic - and the papers have always relied on people just hearing about something that was said/done years ago and treating it as current.

I think the millennials and Gen Z are worse than Gen X and the Baby Boomers with it, but not by much, and the popularity of social media just amplifies it by the sheer volume.

E-K said...

I thought you meant Al... Johnson. (Unsettling)

D'oh !

MyComputerN.A.I.m ;) said...

There is no such thing as AI.

What youre all talking about is VFIP - Very Fast Information Processing.
It takes all inputs and calculates a statistical probability against 'historical behaviour'. This is why ND got his email flagged.

However, things are far, far more dangerous than mass 'surveillance' - thats already in place.
Whats far more worrying is the concept of AI; *this* is the poison pill.

If we accept the concept of Artificial Intelligence, we accept the idea that a machine can make 'decisions'; as with NDs email, above.
You can look at this two ways;
1. The AI 'noticed' a potential mistake and intervened to spare your blushes or
2. The computer registered a statistical anomaly and brought it to your attention

You are being trained to imagine that (1) happened while in fact (2) occurred.
It is important to understand the difference.

The aim of this 'training' is for you to accept that 'AI' makes decisions independent of its programming and programmers and is therefore responsible for its own actions.
Once this notion has been established in the social mindset, we must move to the legal sphere whereby 'AI' must be enshrined in Law as a Legal Entity, separate from its creators and responsible for its own actions.

Once achieved 'AI' can be put into all sorts of machinery - Police, War, Welfare, Banking - you name it!!!! Imagine the fun that can be had when nobody is responsible - or more specifically liable - for anything?!?!?!

Here be dragons....

Anonymous said...

Once this notion has been established in the social mindset, we must move to the legal sphere whereby 'AI' must be enshrined in Law as a Legal Entity, separate from its creators and responsible for its own actions.

The driverless car issue....

I agree with VFIP and not AI though it might be argued that our own intelligence is VF(Human)IP based on glutamatergic receptors.

So VFIP comes down to lies (news), damned lies (fake news), and statistics.

We are all so gullible.

Nick Drew said...

@ MyComputerN.A.I.m

for my money that's the most interesting comment we've had around here since 2011 when anon suggested that the riots would get really interesting when the pool of disgruntled, underemployed graduates gave birth to an officer-class

the distinction you draw is very telling. I'm not (yet) sure I go with the line of thinking you develop from it, but I think we'll be back on this one in a short while

andrew said...

Short answer
Right now it is the driver behind the wheel.
And (imo) it will stay that way for a long time.
Medium answer
What naim said and a road is a very open and uncontrolled environment and product liability and so there is a chance that the car will do X somethings wrong every Y million km.
If it is a washing machine that means your clothes are not dry. If it is a car it means somone might might die.
It does not matter if it is safer than a person.
So level 5 autonomous cars wont happen for some time.
The problem is the uncontrolled environment.
For me there are two issues
"Fat man on the bridge" do you kill the fat man or let the people in the runaway tram die - how does tha ai choose?
Failure and spoofing. There are stories of teslas changing lanes dangerously due to random paint marks on the road. Those random paint marks will appear on t shirts.

Long answer
I think there are govt committees etc working on this and have seen a number of stories on ai and driving on theregister.

The more chilling ideas involve making roads fit for ai by banning other road traffic and pedestrians

I pointed out that a uk motorway is a close fit to that.

Anonymous said...

"The more chilling ideas involve making roads fit for ai by banning other road traffic and pedestrians. "

The railways have been like that for a long time, and have the added advantage that the steering is done by a passive mechanism requiring no computer.

Don Cox

Anonymous said...

@andrew - also been mooted is having everyone being legally required to own a device that emits their current location so that automated vehicles know people are in the vicinity.

Now most of us have devices that announce our location - mobile phones - but we're not required to have one. Yet.

andrew said...

My mum absolutely refuses to carry a mobile.
My dad has a mobile. He was given a new free android because they switched off the 2g signal. But he keeps it switched off until he wants to call (always my landline...)

Most children do not have their own phone until over 11 or so.

People who advocate this have no elderly relatives or children and never ever let their battery run down and never forget to take their mob with them.
And arent foreign.
And do not have dementia.
And and and.....

Otoh think of the chaos when just a few hundred people just leave their mobiles at home and then stand in the middle of the road for 5 mins every 15 mins

Wildgoose said...

It's not just mobile phones that allow your location to be tracked - there are a host of other technical means that can be tied together to track your location.

Which is why Hong Kong protestors used cash rather than their Metro Cards for travelling to the recent protests. And why it is important we retain physical (anonymous!) cash as a means of payment.

Jan said...

Regarding FB, I'm proud to say I only have 5 "friends" on there! I've noticed if one of them looks up something eg timetable for Metro bus I then get an advert for Metrobus popping up so there is no privacy there at all. This sort of thing happens often which is why I hardly use it.

Also I recommend Ch4 "Hunted" to see how the state can track your every move and know all about you. It's not on at the moment but there will be a new series.

I'm quite keen on RT and Aljazeera channels for other news/discussions etc which don't appear on British Channels.

hovis said...

@Jan: Re' "hunted" the purpose seems to be to show just how intelligent and clever the experts are. In other words resistnce is futile.

Wildgoose: agree about cash - use at all oportunities - doesnt create the data. Why crypto is not the panacea that it is positied to be - expecially if itis dominated by a virtual corporate fiat like "Libra".