AI is a Misnomer
Have you heard of a ‘super recogniser’? No nor me, until I met one such super recogniser this morning and discovered they are the people who rectify mistakes made by so-called AI. And sorry if everyone else knew this, but I realised there and then that AI is a misnomer and AI is not Artificial Intelligence, just more complicated computer technology. It’s a point that bears repeating before anyone gets carried away and thinks that AI will be the silver bullet to get the NHS working again, the police to solve crime or the traffic industry to stop jams.
So sorry to be a Debbie Downer, but let me explain the connection between super recognisers and the flaw in our leaders banking on AI to revive our ailing economies. Here goes:
My new super recogniser acquaintance discovered her talent while watching This Morning years ago. “There was some professor on from The University of Greenwich talking about the ability to recognise people’s faces. I assumed that everyone can do this, but apparently they can’t. They were after people to research so I signed up.” (Lord Frost need not apply.)
It turns out my new chum is in the top 1% of super recognisers in that she can see someone’s face once and remember the face from all sorts of different angles and locations. She’s great with all races, which apparently not all super recognisers are. After being trained up she now works in the evenings, looking at images of faces captured by private security firms and matching them up to the faces suggested by facial recognition technology taken from various databases of suspects (I didn’t get round to asking where that came from). Now here’s the disappointing bit: according to my chum, the matches she is presented with by the facial recognition software are only accurate 75% percent of the time. (Big Brother Watch thinks the matches of live facial recognition technology have an even lower hit rate.)“Quite often the matches are laughable,” she explains. If arrests are to be made, facial technology must also be overseen by a human operative, hence the use of super recognisers – because the so-called AI is less top set, more SEN.
This disappointing situation can be applied to all sorts of other so called AI solutions: reading MRI scans, X-rays, understanding blood test results; all of the AI suggestions will need to be verified by humans – for at least the first few years until it gets better. And why is this? Because AI is not Artificial Intelligence, it’s just technology. Hopefully impressive technology, but it is not and nor will it ever be intelligent.
Defined by Dr. Johnson as “spirit, unembodied mind”, intelligence will always and forever elude these machines and software that are currently misnamed AI. Sure, these AI might be able to solve problems and learn from data but they will never be intelligent. School mums will always be there in the background, working part-time making sure they’ve read the X-ray properly.
We are not the first generation to naïvely bequeath non-existent intelligence to machines. There is the famous incidence in 1601 when Matteo Ricci, a Jesuit missionary presented a mechanical clock to the Emperor of China who thought this clever automata was a living creature. We who pin our hopes on AI are as green as that Emperor; it’s just tech, and should therefore be called AT – Advanced Technology – rather than AI.
It’s all Descartes’s fault for positing the mind-body duality which allows us to imagine that if there is a body, there may well follow a mind. The 18th century saw great discussion and interest in the potentiality of automata – all the fancy fountains, clocks and clever self-running toys that were made – to develop souls. The roots for this go way back into folklore when it was believed that animal or ancestral spirits would inhabit puppets. Alas they don’t; in the same way that life does not inhabit a machine and intelligence not exist within a computer. Pinocchio will never become a real boy.
Then as now, we just got carried away with the novelty of new invention. The only known higher intelligence in the universe is human. And that fact is perhaps more terrifying than the prospect of non-intelligent AI.
Joanna Gray is a writer and confidence mentor.
To join in with the discussion please make a donation to The Daily Sceptic.
Profanity and abuse will be removed and may lead to a permanent ban.
Very true, AI seems to be a buzzword attributed to anything computerised to make it sound like we’re in the future.
Many of the computer control systems that are referred to as AI are actually relatively simple. Grok and chatgpt are actually VI (virtual intelligence), complex machine learning systems that give the impression of intelligence but actually aren’t.
Many people and most of the media don’t seem to understand even basic computer systems, and rather than admitting it they simply call it AI, to imply that it’s so complicated that almost nobody could possibly understand. Might as well say “because it’s magic”.
If AI systems got it wrong only 1% of the time, apply that to AI systems in aviation and no one would fly.
There is no hard shoulder in the sky. No roadside assistance up there.
And Boeing is dropping its woke policies which have lead to mediocrity in its systems and its airplanes being built and flown with serious faults. That is just human lack of intelligence. Computer systems can achieve the same but very much faster and on a much wider scale.
Slightly Off-T but this deserves support. Let us not forget that when cash is taken from us freedom dies.
https://petition.parliament.uk/petitions/700404
Thanks, I’ve signed.
Only 40,000 more to hit 3 million signatures and the petition has more than 5 months to run:
Call a General Election
Artificial Intelligence is handy for summarising data and may, may, detect existing patterns in the data that were not previously obvious.
But I suspect that if Artificial Intelligence been around in the early months of 1720 it would have recommended buying stocks in the South Sea Company. By September the bubble had burst and stocks lost 80% of their peak value.
I do wonder if the South Sea bubble applies to Artificial Intelligence too…
i hope so!
Me to – all exec level people are 100% enthralled to the latest fad of AI… even when it produces total crap. I predict once it starts to point out how bloody useless some very senior people are, they’ll rapidly put it back in its box
AI is by-and-large a scam, ie, it’s about selling people something which simply doesn’t exist, namely, software capable of making intelligent decisions. What they’re getting instead is something that’s so complicated that nobody, not even the people who programmed it, can tell how it actually works except in very general terms. One could call construction of such virtual devices the exact opposite of engineering. A real machine is planned down to the last detail to solve a certain technical problem. A so-called AI is something which changes its internal state in unpredictable ways based on some set of inputs it transforms into outputs as side effect of that. It’s considered to be working once enough actual outputs have been gathered to claim that it produces a desired result with a probability of X % despite this is a lie: It has produced desired results (for some definition of desired) with a certain frequency in the past. As nobody knows what it’s actually doing, no one can really claim that it will continue to produce such desired results with the indicated frequency and as the thing is absolutely not intelligent, it’ll also regularly do something humans only very rarely do, namely,… Read more »
I agree. AI is software. Software is a series of individual instructions. Every instruction is specified by hand, by a person writing the software.
The software code (the instructions) may be written by large numbers of people and be very complex. The more complex the software is, the more likely it becomes that the software will produce incorrect results.
Software may be written to ‘recognize’ an external event and react accordingly, but every recognition and every reaction has been specified/written by someone.
Not in this case. The Wikipedia article about neural networks maybe useful to get an idea how these systems generally work. Nobody knows how they exactly work in each individual case, ie, their behaviour is empirically unpredictable. That’s why I’ve called this anti-engineering as engineering is the art/ science of arranging simple parts which empirically predictable properties into complex mechanisms (or structures) whose composite behaviour (or composite properties) is (are) also empirically predictable.
Due to accidents of nature and limits of our knowledge and perception, real-world creations of engineers don’t really have empirically predictable behaviour, just mostly. The job of the engineer is to minimize these uncertainties. The job of the AI devslopper is to take an empirically predictable system and introduce artificial uncertainties into it.
Even complex mechanisms with composite behaviour are not necessarily predictable. E.g. how a train carriage’s suspension system behaves on a particular section of track. All different mechanical sub- components have degrees of tolerance and accuracy, and how it all comes out like on one’s backside can vary quite a bit. What type of performance is best depends a lot on one’s personal perception – but that tales us back to real intelligence.
Yes. I wrote that in the second paragraph.
Anti-engineering? That makes sense.
Yes, but all these processes are managed by software, there is no magic. Neural responses or whatever will be read and interpreted by software, and that software will produce whatever outputs it has been programmed it to produce.
Not in theory. In practice, it may, as such a system very much lends itself to hide cheating/ fraud, but in absence of that, it wasn’t programmed to produce certain outputs, just fed with input data until it produces the desired kind of output frequently enough (for some definition of that).
There’s also an important, other property: For any given output, there’s always an infinite number of other potential outputs which weren’t produced despite any of them might have been the correct one or the best approximation of that.
A so-called super recogniser can weed out wrong results. But it can’t magically include correct results which weren’t part of the output. That someone wasn’t detected by facial recognition software doesn’t mean that he’s not the someone which should have been detected. The system generates both false positives and false negatives but only false positives can be corrected.
Loved the last sentence.
It is not intelligent, it is actually dangerous, just like the Secretary Of State for Energy.
Probably less dangerous than him tbf
Artificial stupidity is no match for Ed Minibrain. Beats it hands down in the race to the bottom.
Ed had an IQ test. It was negative.
Yep, with Ed we are in -ve IQ territory.
That’s why you need the Metaverse and the genetically engineered population reduction. So you can keep the “super recognisers”.
The rest of you should really already be gone.
Augmented not artificial
Non biological intelligence does not exist!
(Apart from Starmer that is!)
I think it should be regarded as non-biological non-intelligence non-leading the non-labour-party.
🙂
A lot of AI seems to be a case of an overfitted statistical model. i.e. It fits the noise as well as the signal.
Agreed.
AI is just a bunch of subroutines that express the opinions of their programmers.
The name for it is Automated Information-Processing. We conflated intelligence with efficiency when we began to think, and AI is the culmination of that formalisation
AI is still going to be very efficient, and that is why it will be universally adopted. But it stands for the assu.ption that began with thinking, that intelligence can be represented by freedom-of-choice. That freedom does not include the freedom to stop, and it is only from stillness that process can be understood