The wonders of AI in Government systems

[Mike HurleyParticipant @mikehurley60381]

What fun. Needed to renew my passport, no rush, followed the really clear instructions on the web site. Decided to do it on-line. Went to a photo booth in local supermarket and had a ‘passport approved’ pic taken. For £9 you end up with a few prints on one sheet plus a ‘code number’ which you need to enter in the online application.

Filled in the online form, got to the photo bit – entered the code – few seconds later a copy of my picture appeared. At this point I will mention that the picture was clear, normal front view of my face without glasses, a hat, hair over face etc. etc. (all as designated no-no’s in the instructions).

The AI then verifies the picture as fitting all the appropriate criteria. With mine (tried this 3 times in all) it says it failed verification as ‘you do not appear to be facing the camera’. ?????? Its a picture of my face! which way am I supposed to face? Totally bizarre.

Ended up sumitting it anyway, so only time will tell.

Mike

[Grindstone CowboyParticipant @grindstonecowboy]

Mother-in-law’s was rejected due to her grey hair not contrasting enough with the background…

In other AI developments, I heard on the radio news a few days ago that the police were claiming AI had enabled them to clear 65 year’s worth of work in 6 months! Seems a little hard to believe?

Rob

By the way, I’m liking the latest layout on the forum

[peak4Participant @peak4]

When I renewed mine, I went to the local Timpsons, who supply a similar service to the photo booth, albeit slightly more expensive.

Bill

[Grindstone Cowboy]

On Grindstone Cowboy Said:

Mother-in-law’s was rejected due to her grey hair not contrasting enough with the background…

In other AI developments, I heard on the radio news a few days ago that the police were claiming AI had enabled them to clear 65 year’s worth of work in 6 months! Seems a little hard to believe?

…

My DVLC photo was rejected because of the background despite being taken by the Post Office.   I thought it was fussy!

65 man years of work cleared by AI is entirely possible.  At the moment UK police employ 147,430 Full Time Equivalents, so 65 man years is only 0.044% of their annual workload.

One application of AI is clearing up old cases by extracting more from the available evidence than was possible in the past.  People scanning many hours of video for faces have a high failure rate, but AI is good at it and tireless.  Also collecting additional detail such as number plates, and doing more intensive correlations of old case data.    Basically AI is good at spotting patterns that people miss even when the manpower is available, and manpower is always in short supply.

In the US, an AI system does face recognition on a collection 30 billion uploaded photographs.  No way could a human do that, but it sometimes allows a photo of a unknown criminal to be linked to an area, family or individual.    Being able to focus an investigation saves a lot of resources.

Dave

 

 

 

[JohnFParticipant @johnf59703]

The ‘new” title AI makes me smile, being a county lad the original AI = Artificial insemination aka The bull in the Bowler hat !!

Had the same problem a few years ago with the passport photo, it kept saying there was a reflection — my grey almost white hair — sent ti anyway and no problems.

As an aside the forum is running unbelievably slow this morning and last night would not load at all !

John

[Michael GilliganParticipant @michaelgilligan61133]

This seems as good a Topic as any for today’s News item:

.

.

.

Righteous indignation from Adobe could be the start of something big !

MichaelG.

[densleighParticipant @densleigh]

The reference to  AI and its farming origins made me think that in your part of the world the AI must have been more expensive than up here in dairy areas of UK

Ours if memory serves had either a cloth cap or none at all – just the very long plastic arm length gloves.

No Bowler hats!

Also he arrived in either a black Morris Minor or a Ford  Pop!

[dodmoleParticipant @dodmole]

I still think that they are too ambitious rushing to use Artificial Intelligence for all these important systems and uses when they didn’t start from Artificial Stupidity and learn from there.   (AS, not to be confused with BS)

[Howard LewisParticipant @howardlewis46836]

IF only the Post Office (And Fujitsu) had remembered GIGO (Garbage In, Garbage Out)

Just because it is appears on a computer screen does not mean that it is correct!

What happened to doing dry runs before launching into widespread use?

Howard

[dodmole]

On dodmole Said:

I still think that they are too ambitious rushing to use Artificial Intelligence for all these important systems and uses when they didn’t start from Artificial Stupidity and learn from there.   (AS, not to be confused with BS)

I suspect Natural Stupidity will totally outclass Artificial Intelligence. 😀

(“Against stupidity gods themselves struggle in vain” – Schiller [German])

[Howard LewisParticipant @howardlewis46836]

As the late, lamented, Rudi Michetslager used to say, “Common sense isn’t that common”

My fear is that Artificial Intelligence may not be as intelligent as all would hope. After all it the programs, initially, are written by fallible human beings.

Poor logic and programming, or data input, might result in catastrophic outcomes.

Howard

[NealebParticipant @nealeb]

…and Artificial Intelligence isn’t really intelligence! It does a pretty good imitation of something that looks like intelligence, but the large data models used for training what we tend to think of as AI systems is just using very sophisticated text analysis to make a stab at “what word would come next in this context?” kinds of questions. It looks at all the text it has ever seen to make the decision. Train a system purely with material from the Flat Earth Society and it will respond to “Is the earth flat?” with reasoned – apparently – answers to say yes, it is. Given some of the nonsense that one sees reported from social media, then the reason that we should support the social media companies who don’t want their material used free of charge to support what is effectively a commercial AI offering is because we don’t want such systems trained on the kind of nonsense that is freely uploaded and then copied across social media. The more times it is copied, the greater weight that an AI system might give it. In fact, it does not have the intelligence to distinguish fact from plausible fiction. Although, having said that, neither do a lot of social media users…

Someone mentioned above, in the context of the current Fujitsu/Post Office fiasco, the “GIGO” principle – garbage in, garbage out. Not entirely relevant to the Fujitsu situation – where it is more like good data in, muck it around with faulty algorithms, garbage out – but it is certainly true in the case of today’s AI systems.

They are astonishingly good at what they do and how well they do it – but that doesn’t mean that you can trust them in all situations. Just think of them as “distilled essence of something that I read on the Internet” and whether you would bet the farm on that!

[Nealeb]

On Nealeb Said:

…and Artificial Intelligence isn’t really intelligence! It does a pretty good imitation of something that looks like intelligence, but the large data models used for training what we tend to think of as AI systems is just using very sophisticated text analysis to make a stab at “what word would come next in this context?” kinds of questions. It looks at all the text it has ever seen to make the decision. Train a system purely with material from the Flat Earth Society and it will respond to “Is the earth flat?” with reasoned – apparently – answers to say yes, it is. Given some of the nonsense that one sees reported from social media, then the reason that we should support the social media companies who don’t want their material used free of charge to support what is effectively a commercial AI offering is because we don’t want such systems trained on the kind of nonsense that is freely uploaded and then copied across social media. The more times it is copied, the greater weight that an AI system might give it. In fact, it does not have the intelligence to distinguish fact from plausible fiction. Although, having said that, neither do a lot of social media users…

Someone mentioned above, in the context of the current Fujitsu/Post Office fiasco, the “GIGO” principle – garbage in, garbage out. Not entirely relevant to the Fujitsu situation – where it is more like good data in, muck it around with faulty algorithms, garbage out – but it is certainly true in the case of today’s AI systems.

They are astonishingly good at what they do and how well they do it – but that doesn’t mean that you can trust them in all situations. Just think of them as “distilled essence of something that I read on the Internet” and whether you would bet the farm on that!

My intelligence disagrees with Neil’s!  It’s because, just like an AI, we are both struggling to understand a complicated dataset.  It consists of what we have seen, been taught, read, listened to, eaten, and where and by whom we were brought up.   Everything depends on how our brain interprets sense inputs.   Computers are analogous.

One model suggests people have two brains.  The original is primitive and it deals with challenges requiring quick reaction: fight or flee.  It’s response is emotional, not logical – when attacked by a bear, there isn’t time to out-think the beast.  The logical brain evolved much later, and it is mostly responsible for what we call intelligence.

Humans are not unique in displaying intelligence.   Crows can count up to 4, my cat learned how to open doors by jumping on the handle, and dogs train their human slave to pick up poo. In addition, human intelligence takes many forms.  A talented colleague had a photographic memory, which made him extraordinarily good at his job, but he couldn’t innovate or work in a team.   Another could speed-read, was a good manager, and made smart decisions but hated pressure.   I used to play chess with a friend during the lunch-hour.  I always won if we played to finish the game in an hour: he always won if we played slowly.   Small children learn to talk in a few years after birth, and then most of us lose the ability.  MRI scanning reveals the human brain goes through at least 2 major structural changes before adulthood, and one of them explains adolescent behaviour.  Thereafter the brain slowly degrades with age, and older folk tend to lose the ability to innovate and rely ever more on experience.  Very evident in Mathematics, were all original work is done before age 30.

So ‘intelligence’ covers a wide-range of different attributes, and I don’t believe there is a single definition that covers it. All forms of intelligence are useful.

Alan Turing is justly famous for proving what computers could and could not do:  ‘The Entscheidungsproblem (decision problem) was originally posed by German mathematician David Hilbert in 1928. Turing proved that his “universal computing machine” would be capable of performing any conceivable mathematical computation if it were representable as an algorithm. He went on to prove that there was no solution to the decision problem by first showing that the halting problem for Turing machines is undecidable: it is not possible to decide algorithmically whether a Turing machine will ever halt. This paper has been called “easily the most influential math paper in history”.’  (from Wikipedia)

Turing’s proof opens the door to the possibility of Artificial Intelligence, though not with the machines of his time.  Nonetheless Turing was able to suggest a measure: basically, if a human conversing with a teletype can’t tell if the correspondent is a machine or a human, then the correspondent is intelligent.   The Turing Test does not depend on opinion, beliefs, or preconceptions, and it requires the human to prove his intelligence too.

Turing didn’t know how AI might be implemented, but he did know it was possible.  Over the last 70 years, the design issues have become clearer, several problems have been fixed, and technology has improved immeasurably.   Processors are thousands of times faster, memory is cheap and plentiful, software can run on many processors in parallel, maths is done with high-speed hardware, and so are graphics.   Turns out the hardware needed for high-performance graphics is also well suited to machine learning.

Machine learning is a breakthrough technology.  A machine learning program isn’t written to perform a specific task, rather it is written to extract knowledge from a large dataset, and that can be built on.  The result is analogous to how an organic brain develops.  The data is anything that’s been digitized: images, text, radio signals, voice, music, weather, medical data,  aircraft performance, whatever.

Within a narrow range AI already outperforms human specialists.  AI is better at spotting anomalous cancer cells than a human microscopist, leading to more reliable diagnosis and spotting new clues.   AI can also be used to analyse fiction and generate new books or film-scripts.  And it’s possible for an AI generated film-script to be converted into a computer generated film.   I don’t think a major film has been done yet, but shorts and sequences certainly have.   AI writes better computer programs than most people.   Above all, AI can only improve – it isn’t fully developed yet.

Of course, AI suffers the same problem as humans – rubbish ideas due to faulty input!  On the other hand, AI is less likely to stick with nonsense.  It won’t ignore evidence and facts just because they contradict a human belief.   AI is unlikely to start a religious cult, develop toxic masculinity, fall for a Ponzi scheme, support -isms, blame minorities, deny unpleasant realities, jump on the bandwagon, commit crime, or play politics!

The most serious limitation of human intelligence is our inability to process large amounts of information.  As such the human race are pretty ignorant: no-one understands everything.   So humans jump to conclusions, and are prone to wishful-thinking and paranoia.  A tool like AI that provides a dose of reality is surely useful!  It is of course possible that lies are the only proof of intelligence.

Humans hate change, even if it’s good for them.  Unfortunately AI is going to alter the workplace on a grand scale.   Lots of people will have to find new jobs, and it will be painful.   Nothing new in this: the motor car put millions of horsey workers out of business.  My village had two forges, and no-one cares the blacksmiths have gone.

Dave

 

[Nigel Graham 2Participant @nigelgraham2]

”  Unfortunately AI is going to alter the workplace on a grand scale. Lots of people will have to find new jobs,.. ”

Maybe – but in what, though?

We do keep hearing or reading this prediction but it never seems supported by statements of the fields of work that could be replaced by a computer – which is what it amounts to. Nor indeed what work would still need a person.

It reminds me of the working at home thing in the worst of the Covid pandemic (the diseases is still around!): the Government were careful to say “where possible”, but all manner of reporters and pundits and so-called “lifestyle journalists” forgot or ignored that phrase. Maybe because many of them do not need go anywhere for their work, just sit in an office / spare bedroom all day long.

‘

Many moons ago, when the PC was cautiously finding its way into offices, I was a materials store keeper in a factory making specialist printing-machines. One day the Purchasing Manager suggested a computer would help my stock-control. I looked at my boxes of pro-forma cards on which I kept my records by hand, and politely pointed out that a computer would struggle to help me unload a ten-foot length of three-inch diameter, oil-covered steel bar from a lorry and lug it half-way through the building to my store and hacksawing-machine. He conceded I had a point.

Would “Artificial Intelligence” , or even just a PC with the latest Windows or Linux, do any better?

It did not take me long to twig that “I” can also stand for “Idiocy” ……

 

 

[Nigel Graham 2]

On Nigel Graham 2 Said:

”  Unfortunately AI is going to alter the workplace on a grand scale. Lots of people will have to find new jobs,.. ”

Maybe – but in what, though?

We do keep hearing or reading this prediction but it never seems supported by statements of the fields of work that could be replaced by a computer – which is what it amounts to. Nor indeed what work would still need a person.

[…]

That question sums it up nicely, I think

… and I would suggest a ‘blanket’ answer …

The computer is good at collating information, and finding patterns within it

The person is better-equipped to decide what to do with that new-found knowledge

… it’s a symbiotic relationship.

.

As a simple example [relating to some recent discussion here] … Imagine a project to re-launch Meccano into the second quarter of the 21st Century.

The first job for the machine might be to read all those wonderful archived drawings, and “understand” how the product had developed in the previous decades.

The human would then be better-equipped to contemplate what might be done next, and could  feed ideas back into the machine for further analysis.

MichaelG.

 

[Author]

Posts