52 replies to this topic

[Henk Vasmel]

Yes, I know, Artificial intelligence has nothing to do with Nostalgia. On the other hand, the future could bring us some things that might be useful for us.

 

Let me say this first. My interest here is to know: which car identity did what in the races it participated in, or in short, what are the chassis numbers of all of the cars that ever raced. I know that this will never be solved definitely, but we can try and take into account all technological means that are available to us.

 

Today I had the privilege to attend a meeting at work about Data Management and Artificial intelligence. During the presentation of the latter, I heard something that sent my mind drifting.

This technology is able to derive data from graphics. The contents of the picture are there as a true fact, but AI opens up the possibilities of using the data again.

Then the next thing that crossed my mind is the age-old art of rivet-counting. Thet led me to think about the (for us) holy grail, identifying Car identities from old pictures.

Would it be possible to feed two pictures of a similar car, in different views, maybe different races, and have AI determine whether these could be the same car, or definitely not, because of some incompatible details.

In the old days, the number of rivets was a sure sign, but also some kind of exotic, hard working, way of identifying cars. In later years, I think also the placement of sponsor stickers could play a role. Only an automated analysis of pictures (with enough accuracy) could give definite answers here.

This needs of course some testing, to find out which visual details are reliable, but I think it is worthwhile to investigate.

As soon as this process works, we could be informed about the history of certain chassis, fed by details that have been too difficult to determine up until now.

 

I know it will at first not be too easy. You will need pictures with sufficient resolution, and a method to get a comparable view on them, accepting that in one of the pictures not all details necessary will be in view. But that is a learning process.

 

Any young, enthousiastic reader here feels tickled to investigate more?

 

 

[E1pix]

Please realize that any use of online images, text, illustrations, music, data, history and maybe a million more resources harvested by AI is the most-threatening factor to the Arts in human history.

Those of us who spent our lives in photography, text, and graphics are mostly all wondering why the hell we bothered. I already had a burgeoning client send me a logo from AI that “wasn’t quite right” and wanted me to “fix it.” I refused.

As with any decision to use technology starting with “Me, Me, Me” as why we do, please re-think if we care about each other. The people you’re affecting might be the same people you love.

Then there’s this:
https://www.forbes.c...sh=4c8701ae782b

[10kDA]

What if, for example, a car has a body panel damaged in a fairly minor way one racing weekend and a replacement panel from a teammate's car is installed with slightly different sticker placement. If AI does not get the word about this in-the-field replacement, what prevents it from determining something along the lines of "Car A was used in practice but car B was used in the race" and making its AI-typical totally assured statement? This has already happened in the world of air racing, without AI being applied, and it became a revival of a previously discounted myth that had to be resolved by Someone Who Was There. When those people are gone, who will set AI straight? It seems as if many, many people are accepting something resembling "infallibility" from AI even now. It's going to get more pronounced.

 

Your mileage may vary. LOL

[E1pix]

To be fair, somebody’s gotta design a bot that posts a thousand times a day with “E1pix World Champion 1984-94.”

That should do it.

[Odseybod]

I really don't know enough about this to offer any expert insights, so please accept these as general observations. On the plus side, my instinct would be that if it works for face recognition, it should be adaptable for car recognition.

 

BUT faces - at least, adult ones - are approximately the same size and with the same 'components' (eyes, nose, mouth, etc) in roughly the same juxtaposition. So even if a face appears small in an image, perhaps because it's in the distance, it would be possible to size it onto a standard grid, so you're comparing like with like. Whereas cars are not so co-operative and come in various different sizes, and with no standard arrangement for their 'facial' components (thinking of the Fiat Multipla's headlights underneath the windscreen, the Citroen DS's rear indicators in the roofline, a vintage Talbot's headlights positioned right next to each other, as just three examples - though useful identifiers in themselves). So transferring a car's 'face' onto a standard grid would be a challenge, even without the challenge of perspective interfering.

 

I admit I'm thinking of it as a way of identifying an unknown type of car in a street scene (say), whereas I think you're hoping to establish whether two images of a known car shows the same car twice or two different cars. But I think the same challenges would apply, with the need to map its various features onto a grid, while somehow compensating for perspective, camera lens distortion, dirt and any damage, then ask the helpful computer to point out any differences between them, while perhaps referring to a stock image of the same vehicle it holds in its memory bank.

 

I'd imagine the task would make even a super-computer break out into a slight digital sweat. But who knows?     

[Michael Ferner]

Yes, pattern analysis and recognition is one of the most exciting tasks that AI will be able to perform - it's already being developed for medical sciences, and no doubt other applications as well. This is one of the the core values of AI, unlike that whole ChatGPT shtick which is just a gimmick to draw attention, with very little practical use. Pattern analysis, on the other hand, can and will replace mind-numbing and error-prone human tasks that consume time and resources which could be invested in more productive areas.
 
I'm sure you could train an AI to "count rivets" and, more importantly, analyse their patterns (that goes for shapes also, of course), and that it would be most useful, but it'll be a long way to go to get there. First, you need to feed it thousands of photographs, and let it find similarities. Then, of course, you'd have to lead it along for quite some time, until it learns that a duck has feathers and not rivets, and so can't be a racing car, but I'm sure that with time it will be able to reliably tell a Ferrari from a Williams, a Williams FW07B from an FW07C, and finally chassis 14 from 15!

 

But, there's a long road ahead, and before we get there, there are other aspects of AI that will be more pressing, like the transition it will cause in the information age - some would call it a disruption. Not that long ago, I read a comment stating that website owners should be "alarmed", because eventually AI will make websites redundant, which is of course true if you think about it. People will no longer google to find answers, but instead ask a 'personal assistant', i.e. AI. That's where lots of money will go in AI development, because if you can be the Google (i.e. the gold standard) of the personal assistants that'll be a bonanza. It's not really high-level application, but most of the AI industry will follow the easy money first, so there it is.

 

The question here is, what does it all mean? I don't subscribe to the theory that AI will be a disruption, because if you think about it, the disruption has been going on for a long time already. The first thing that will go is wikipedia, which is in a way ironic because that is where the disruption began, and people were volunteering to spend hours writing wikipedia articles "because information should be free" - now they will be uncompensated authors for one of the most profitable industry revolutions the world has ever seen, "thank you very much for your generous contribution to my wealth..." Wikipedia started by nicking content from other websites and books, the same as AI will be nicking content from wikipedia (inter alia). Most of us have already thought about protecting themselves from wikipedia theft, now we need to think about AI instead. Just food for thoughts...

[AJCee]

If I may say so, very well put Michael.

It is important to treat the AI output with the same level of scepticism as you would a picture in a book or magazine: do you actually believe it is correct? And from the feedback of that, the AI learns to be more accurate.


AI is about to become very important in my area of work (medical science) and it will save a lot of hours of drudge with much higher accuracy.

[Nathan]

Please realize that any use of online images, text, illustrations, music, data, history and maybe a million more resources harvested by AI is the most-threatening factor to the Arts in human history.

Those of us who spent our lives in photography, text, and graphics are mostly all wondering why the hell we bothered. I already had a burgeoning client send me a logo from AI that “wasn’t quite right” and wanted me to “fix it.” I refused.
 

 

I wonder if painters felt the same when when the camera became mainstream, or animators with digital graphics design.

[E1pix]

They (we) did, but those tasks were still operated by humans.

[jcbc3]

OT

 

Spoiler

Please realize that any use of online images, text, illustrations, music, data, history and maybe a million more resources harvested by AI is the most-threatening factor to the Arts in human history.

Those of us who spent our lives in photography, text, and graphics are mostly all wondering why the hell we bothered. I already had a burgeoning client send me a logo from AI that “wasn’t quite right” and wanted me to “fix it.” I refused.

As with any decision to use technology starting with “Me, Me, Me” as why we do, please re-think if we care about each other. The people you’re affecting might be the same people you love.

Then there’s this:
https://www.forbes.c...sh=4c8701ae782b

 

Luddites much?

 

AI is a tool. Use it and much beautiful art will be created. Television and Cinema didn't spell the end of theaters. As Nathan mentioned photography didn't replace painting. A programmer masterfully instructing an AI picture generator can make better art than 'crying gypsy' ever was.

[10kDA]

OT

 

Spoiler

 

Luddites much?

 

AI is a tool. Use it and much beautiful art will be created. Television and Cinema didn't spell the end of theaters. As Nathan mentioned photography didn't replace painting. A programmer masterfully instructing an AI picture generator can make better art than 'crying gypsy' ever was.

I disagree 100%. Something about nearly all digitally rendered art looks off to me. Some people don't see that but many, many people do. One factor is that it has a "product" vibe about it that significantly diminishes the authenticity. It has a bit to do with the difference between being an illustrator and an artist. I use digitally generated instrumental parts in my music, for instruments I cannot play myself. The generators are very sophisticated yet I would hesitate to use what they produce as finished product. They are great for making demos. Unfortunately consumers' ears have become accustomed to musical sounds created entirely "within the box" and they may not have a clear idea of what the sounds of real instruments are like. Thousands of "new releases" per year feature synthetic music, and there is no going back at this point. Real instruments played by real people are heard by fewer and fewer people and the listeners' experience is poorer for it.

[Michael Ferner]

îî That!! îî 

 

Tune in to any ordinary radio station these days, and all you hear is computer music! I can instantly hear the difference, and it absolutely turns me off.

[Sterzo]

Would it be possible to feed two pictures of a similar car, in different views, maybe different races, and have AI determine whether these could be the same car, or definitely not, because of some incompatible details.

In the old days, the number of rivets was a sure sign, but also some kind of exotic, hard working, way of identifying cars. In later years, I think also the placement of sponsor stickers could play a role. Only an automated analysis of pictures (with enough accuracy) could give definite answers here.

This needs of course some testing, to find out which visual details are reliable, but I think it is worthwhile to investigate.

As soon as this process works, we could be informed about the history of certain chassis, fed by details that have been too difficult to determine up until now.

 

I know it will at first not be too easy. You will need pictures with sufficient resolution, and a method to get a comparable view on them, accepting that in one of the pictures not all details necessary will be in view. But that is a learning process.

 

Any young, enthousiastic reader here feels tickled to investigate more?

 

Well I'm not young, but I think it a very realistic proposal for a useful tool. A large part of what you describe is already achievable with current technology - it just needs someone to write the programme for this specific purpose. Everyday photo-editing programmes can distinguish faces and unblur them, and identify backgrounds so they can delete them, while police systems are identifying specific people from their faces. It's a small step to comparing data from two pictures, and I could imagine output which also assesses the level of certainty to the conclusions.
 

[sabrejet]

I disagree 100%. Something about nearly all digitally rendered art looks off to me.

 

But what about the stuff you didn't realize was AI?

[Sterzo]

Just to add, there are people on the Racing Simulators forum who are already translating two-dimensional pictures into different views from different angles, whether it's an outside view of a car rounding a corner, or the view of scenery as seen from a cockpit or moving camera. What they're doing must be very close to this idea - all we need is the "comparison" program.

[10kDA]

But what about the stuff you didn't realize was AI?

Why would you assume I didn't realize it?

[Nathan]

They (we) did, but those tasks were still operated by humans.

 

A.I. still has to be told what to do.  An example... on my Facebook feed I get these silly native American posts looking for 'likes' with an A.I. created image of a 'Native American girl'.  The problem? They don't look real.  They are white women with a tan.  Not even a hint of Asian.  Is A.I. racist? I suspect its more down to the developer or publisher, who seemingly didn't care.  My wife likes interior design and shows me these silly A.I. created pictures.  One had a microwave above the fireplace in the living room.  Often the placement or design of things is totally impractical.  Many AI pictures of people exaggerate physical features.  Does A.I. have a 'type', or are the humans behind that AI created picture telling the program to put a big badunk-a-dunk on it?  

I think it's like what painters felt with the camera.  You sit there for an hour or so capturing a moment with your brush or pencil, using skill learned over years, and then along comes this young kid, points this box,  turns this knob then that, pushes a button and feels satisfied with his art.

Edited by Nathan, 15 February 2024 - 14:54.

[E1pix]

I think it's like what painters felt with the camera.  You sit there for an hour or so capturing a moment with your brush or pencil, using skill learned over years, and then along comes this young kid, points this box,  turns this knob then that, pushes a button and feels satisfied with his art.

Those tasks were still operated by humans, and they didn’t steal paintings to do them.

[E1pix]

A programmer masterfully instructing an AI picture generator can make better art than 'crying gypsy' ever was.

Thanks for making my point.

[10kDA]

A.I. still has to be told what to do.  An example... on my Facebook feed I get these silly native American posts looking for 'likes' with an A.I. created image of a 'Native American girl'.  The problem? They don't look real.  They are white women with a tan. 

 

LOL that's been happening for a lot longer than AI-generated images have been around as well. I had a NA girlfriend who would laugh about them and say things like "Come on! Of course she's NDN. Can't you tell from the eagle on her shoulder, and the bear cub she's nuzzling, and the wolf she's petting?"

[sabrejet]

I had a NA girlfriend who would laugh about them and say things like "Come on! Of course she's NDN. 

 

Am I the only one wondering what a normally aspirated girlfriend is?

[E1pix]

I think it means “one who breathes,” so very much on topic. ;-)

[10kDA]

Am I the only one wondering what a normally aspirated girlfriend is?

It's short for Native American. Same as NDN but for white people.

[Bob Riebe]

Artificial Intelligence, is just the latest way to make something sound like it is some thing it is not, it is programmed , not intelligent.

 

 Intelligence

the ability -- to acquire --  and apply --  knowledge and skills.

 

It is a computer program that is only as intelligent as the gent/s-lady/ies writing the program and constructing the computers that use it.

 

Nothing more, nothing less.

 

Without a power source , as useful as teats on a boar.

 

Happy Canada Flag Day.                                            

[Michael Ferner]

You missed the point about AI, Bob. it's not a "computer program that is only as intelligent as the gent/s-lady/ies writing the program and constructing the computers that use it". You're right about the power source, though.

[Michael Ferner]

I think it means “one who breathes,” so very much on topic. ;-)

 

Like

 

 

(There was this movie a few years ago, with Joaquin Phoenix and the voice of Scarlet Johansson iirc - soon to become reality!!)

Edited by Michael Ferner, 16 February 2024 - 07:57.

[dmj]

This is a multi-dimensional subject.

As mentioned earlier, to some extend comparable to threat that invention of photography represented for painters. But we know how they responded: as soon as documentary need of capturing the history with a brush or pencil vanished, painters became free to exploit different paths, created the modern artistic world by doing what photographers couldn’t do. Along came impressionism, expressionism, abstract art… Mediocre painters lost their jobs (or switched to doing portraits of bored tourists during summertime in coastal cities), while good ones eventually got more freedom and created new artistic dimensions, unknown to previous artists.

Computer-generated music? Maybe it is dominant on mainstream radio stations – but free availability of streamed music also changed the music world towards live performing. Most musicians nowadays can’t live from revenues they get from their recordings, and are forced to do more live performances. Thus, we have more opportunities to hear live musicians performing live concerts than ever before. And fans are buying their recordings and merchandise at the concerts, to directly support artists they like. Forget the chart-toppers, the most diverse genres are blossoming today – but you can’t rely on mainstream media to become aware of the artists you might like.

Learning from these examples, I’m not afraid of AI. It might make boring aspects of life easier to handle but won’t threat skilled people in any creative area. And, yes, I hope it might be a great tool for tasks like identifying pictures, but it will take a long time to get it there, too.

[jcbc3]

Thanks!

[10kDA]

This is a multi-dimensional subject.

 

Computer-generated music? Maybe it is dominant on mainstream radio stations – but free availability of streamed music also changed the music world towards live performing. Most musicians nowadays can’t live from revenues they get from their recordings, and are forced to do more live performances. Thus, we have more opportunities to hear live musicians performing live concerts than ever before. And fans are buying their recordings and merchandise at the concerts, to directly support artists they like. Forget the chart-toppers, the most diverse genres are blossoming today – but you can’t rely on mainstream media to become aware of the artists you might like.

 

As a composing and performing musician, I can tell you this is not exactly the case. There is at least as much gatekeeping regarding live performances by relatively "unknown" performers as has ever been, maybe more. The reasons behind the gatekeeping have changed slightly over the 50+ years I've been doing this, and it has devolved to include issues of political perception and age-o-centrism, along with the more standard "we just don't think you're a good fit for our crowd" etc kinds of refusals. The music industry has identified and labelled a huge number of "genres", some of which are indistinguishable from others, and most members of today's audiences are conditioned to categorize what they are hearing based on a description sourced from outside their own perception and experience. That can be a barrier.

 

Some musicians are producing their own shows to sidestep the lower rungs of the popular music performance venue ladder. That's a time-tested method of creating exposure but it's hard to maintain focus on the music when doing all the promotional and organizational steps necessary to allow people to discover who you are and what you're doing. Not all musicians are comfortable with, nor effective at, the type of work involved with generating performance opportunities.

[Michael Ferner]

This is a multi-dimensional subject.

As mentioned earlier, to some extend comparable to threat that invention of photography represented for painters. But we know how they responded: as soon as documentary need of capturing the history with a brush or pencil vanished, painters became free to exploit different paths, created the modern artistic world by doing what photographers couldn’t do. Along came impressionism, expressionism, abstract art… Mediocre painters lost their jobs (or switched to doing portraits of bored tourists during summertime in coastal cities), while good ones eventually got more freedom and created new artistic dimensions, unknown to previous artists.

Computer-generated music? Maybe it is dominant on mainstream radio stations – but free availability of streamed music also changed the music world towards live performing. Most musicians nowadays can’t live from revenues they get from their recordings, and are forced to do more live performances. Thus, we have more opportunities to hear live musicians performing live concerts than ever before. And fans are buying their recordings and merchandise at the concerts, to directly support artists they like. Forget the chart-toppers, the most diverse genres are blossoming today – but you can’t rely on mainstream media to become aware of the artists you might like.

Learning from these examples, I’m not afraid of AI. It might make boring aspects of life easier to handle but won’t threat skilled people in any creative area. And, yes, I hope it might be a great tool for tasks like identifying pictures, but it will take a long time to get it there, too.

 

Like

[E1pix]

Not giving you a Like just because you told me to. ;-)

Ah, Hell…

[DCapps]

The idea called AI -- which is actually more of a generic moniker for a wide variety of applications, with the various models, algorithms, and whatnot associated with it -- the notion has been around for ages in many forms and formats.

The current touting and marketing of AI is, to be blunt, as much of the usual scam as it is of any useful application. 

We used SAF (Sami-Automated Forces) in constructive simulations decades ago; we then developed the algorithms and enhanced the models to create ONESAF, a command and staff training simulation well over a quarter of a century ago (I was one of those very much involved in the briefing for the approval of ONESAF by the Deputy CG of TRADOC, LTG John Abrams). SAF and its many derivatives have been used in the constructive simulations supporting the Army's battle command training for decades. It is a form of AI, one that went from rather simple in JANUS or BBS or other early constructive simulations to be being sophisticated today.

 

There are limits, however, to AI, one of them being the practical issue of No Power = No AI, as well as No Cloud = Not Much AI, plus GIGO = AI.

 

While I am certainly not on the cutting edge of this stuff as I used to be, what passes as "AI" used to be part of my day job for years.

While AI is certainly capable of being an Instrument of Mass Destruction in so many aspects and possible ways, along with being a very useful instrument, of course, it has its Achilles Heel problem as does each and every technology.

While I am very attuned to the issues that AI may and is creating, don't forget that much of the current push for AI is based upon the desire for power and the always present factor of simple greed.

Edited by DCapps, 16 February 2024 - 14:59.

[E1pix]

Nothing said so far regarding the purported innocence of AI changes the fact that it harvests copyrighted works to make new works — to no benefit of the copyright holders.

The painting analogy, sorry, is completely irrelevant. I’ve lived with all my art works being distorted by science and tech “advances” for over forty years now, and that’s an entirely different conversation. In all those cases, it was an “adapt or die” scenario that was maddening enough in their own ways, but I wasn’t competing against an inhuman entity to earn a living. Nor was any entity outright stealing from my contemporaries or my own life’s work to beat me out of contracts. Think world finances are tenuous now?

The idea that a good programmer can and should be able to replace artists through adept commands based on stealing Art itself is utterly sickening, and speaks to the inhumane nature of science itself. Pi in your face, Mr. Science.

The real problem isn’t the AI per se, it’s humanity — or more accurately, inhumanity. Few care about career loss until it’s their career being lost, and that approaches mass-sociopathic terrain. The interesting if not righteous part is the developers, investors, proponents, and users of replacing human tasks will be losing their jobs to their brainchilds in due course.

Worth noting is since “advances” have already sucked my latest industry dry — the way I paid bills and cared for my disabled wife — we now live incredibly cheaply at the expense of all we people tend to think of as “payback” for eighty-hour weeks and missed parties and no vacations and income re-invested into a “future.” Things like a home, a shower, a heater, a fireplace, neighbors, and a simple roof that’s not steel.

Point being, I’ve been mostly forced into early retirement, so I am not defending Me, Me, Me, but rather — quite possibly — You.

Being that we’ve vastly forgotten how to care for each other, apparently, only when AI affects You can we have a proper debate.

[Bob Riebe]

Some musicians are producing their own shows to sidestep the lower rungs of the popular music performance venue ladder. That's a time-tested method of creating exposure but it's hard to maintain focus on the music when doing all the promotional and organizational steps necessary to allow people to discover who you are and what you're doing. Not all musicians are comfortable with, nor effective at, the type of work involved with generating performance opportunities.

As this thread deal with A I - automated incompetence, I have question that I have wondered about for years.

 

With small towns AM radio stations dumping real people in the station for some auto tape (or what ever ai system is used now) It have been listening to such when the auto program changing system hit the glitch that on records meant it was stuck.

 

I assume it is an end of tape program that goes on when the ai fails to automatically punch the correct button to change shows.

What comes on is music, actually very pleasant music that goes on, and on, and on, and on, some times consisting of perhaps two songs, that play until some dude or dudette is either listening afar or rather annoyed listeners get on the phone and tell the local radio kahuna, the record is stuck.

 

Who records the songs , like a mellow jam session, that automatically plays when the record gets stuck?

Computer generated or a studio band some where that makes music for such situations.?

[10kDA]

As this thread deal with A I - automated incompetence, I have question that I have wondered about for years.

 

With small towns AM radio stations dumping real people in the station for some auto tape (or what ever ai system is used now) It have been listening to such when the auto program changing system hit the glitch that on records meant it was stuck.

 

I assume it is an end of tape program that goes on when the ai fails to automatically punch the correct button to change shows.

What comes on is music, actually very pleasant music that goes on, and on, and on, and on, some times consisting of perhaps two songs, that play until some dude or dudette is either listening afar or rather annoyed listeners get on the phone and tell the local radio kahuna, the record is stuck.

 

Who records the songs , like a mellow jam session, that automatically plays when the record gets stuck?

Computer generated or a studio band some where that makes music for such situations.?

Could be either, or, or a combination. I have some friends in the music biz who make their entire livelihood creating music like that for what are known as "music libraries", which contain all kinds of music sorted by tempo, genre, length, you name it. My friends record a band, or themselves playing along with computer generated tracks, or they do all the musical content In The Box, meaning computer generated and then they tweak it to how they want it. The music is sold to whomever wants to use it for any use you can think of. One of my friends had one of his songs used behind two different commercials - one for the dreaded Medicare Enrollment Period squawk and also one for a public service announcement about outdoor recreation programs. Whomever assigns music for broadcast/tech difficulties/glitches probably finds what they want in library folders labelled "Soothing" or similar. My ISP, whom I call more than I want to, offers a choice of Hold music: "for Pop, press 1. For Latin, press 2. For Country, press 3". They're OK but obviously computer generated. Though the salsa bateristas sound passable given the crappy phone speaker.

[Gabrci]

Those of us who spent our lives in photography, text, and graphics are mostly all wondering why the hell we bothered.

 

Imagine poor cavemen having spent all their lives working with a piece of rock, how they must be feeling now 

[LittleChris]

 My ISP, whom I call more than I want to, offers a choice of Hold music: "for Pop, press 1. For Latin, press 2. For Country, press 3". They're OK but obviously computer generated. Though the salsa bateristas sound passable given the crappy phone speaker.

 

Think yourself lucky, I had the misfortune of trying to contact the UK Tax Office a couple of weeks ago and spent 1.5 hours on hold listening via my mobile phone to some soulless jazz funk dross on a 5 second repeat.  Concluded call in around 1 minute with some clueless agency person who didn't have a scooby once the phone was finally answered 

 

Going back to AI though, has it decided who won the 1975 Canadian GP yet ( see the thread in Racing Comments from post 46 onwards ) ? 

 

ChatGPT's knowledge of all things motorsport - Racing Comments - The Autosport Forums

[E1pix]

Imagine poor cavemen having spent all their lives working with a piece of rock, how they must be feeling now

Is this actually the best contribution you have to give?

Pretty sad, but at least we know where you stand.

[PCC]

Going back to AI though, has it decided who won the 1975 Canadian GP yet ( see the thread in Racing Comments from post 46 onwards ) ? 

 

ChatGPT's knowledge of all things motorsport - Racing Comments - The Autosport Forums

On the other hand, maybe it could help us figure out who won the 1973 Canadian Grand Prix.

[PCC]

My concern about AI is quite concrete rather than hypothetical (odd for me, I know). It's been standard procedure forever for people in my line of work to assign students essays. We do this not because we think they're going to find jobs writing essays, but because the research, analytical and communication skills required by essay writing will serve them well in just about any professional field they choose to enter.

 

Apparently, AI can write a roughly 'B' level undergraduate essay on a fairly standard topic in a few seconds. This is a problem, because it means students who use it won't acquire the skills that they are supposed to be acquiring. And it's all but impossible to determine whether it's been used or not - at least, with sufficient proof to support a plagiarism penalty.

 

I try to get around this by creating very specific essay questions and different kinds of assignment that can't be answered by searching online. I think this works on the whole, but as AI improves I feel I'm being painted into a smaller and smaller corner.

 

We're being encouraged to stop thinking of AI as a problem, and embrace it as a learning opportunity. But in practical terms, I don't know what this could look like. Maybe the essay is a thing of the past - the Cosworth V8 of the educational world. But how else do we make sure students acquire the skills they'll need?

 

It all makes me glad that retirement isn't too far off....

[Doug Nye]

As a practical example of AI phobia - or in this case really pre-AI phobia - I had a heated discussion with an enraged driver/entrant at one of the earlier Goodwood Revival meetings who wanted a (faster) rival's car banned because it had a fuel-injected version of the same engine his car was constrained to run, which was carbureted.  When I told him that the four years' younger opposition car ran PI in period and he had better learn to live with that fact, his comeback was "So you've got photographs of it in the pits at the time showing PI?  Well anyone can do anything to fake up photos today!". 

 

After I told him he was talking nonsense in grasping at such a straw, and that he could always withdraw his entry and spend the weekend less stressfully elsewhere, we parted ways.  A principal of the meeting later told me "I had XXXXX dripping to me that he'd had a heated discussion with you and that he thought you'd told him to 'F--- off'...and I told him 'I'm surprised you sound so unsure of that - Doug usually makes such things perfectly clear...".

 

 

DCN

[404KF2]

My interest here is to know: which car identity did what in the races it participated in, or in short, what are the chassis numbers of all of the cars that ever raced. I know that this will never be solved definitely, but we can try and take into account all technological means that are available to us...

I have thought something similar with respect to my Peugeot 404 Register for Le Club 404....there are lots of photos of cars with unknown serial numbers, but getting much past 2,560 cars I've found of the 17,225 made by serial number is a challenge. I think that AI will not be helpful.

 

Already nearly 15% of these rust-prone Pininfarina-built cars are shown to still exist today. Without access to old production records, old and current vehicle registration records (especially in France) and the like, it's impossible to imagine. So I'm doing it manually. Obviously as a hobby - this sort of thing never generates any income; it just costs money. And time.

[LotusElise]

If we're going to have AI performing tasks, it should be the jobs no-one wants to do that get automated. There is no shortage of writers or artists or other creative professionals and the ones who already exist are usually struggling. We don't need "help" from stupid computer programs. What we do need are robots that can roam around greenhouses picking strawberries at the right stage of ripeness, or self-cleaning toilets, or machines that can gut fish accurately.

 

AI media creation is a threat to free speech as it never kicks back against censorship or attempts to suppress information. It has the potential to replicate inaccurate data exponentially. For example, there are rubbish AI programs for colourising black and white images which take no notice of context. I've seen pictures of servicewomen on VE Day with their uniforms coloured RAF blue when the badges indicated they were from a different service entirely, if you looked closely. The world of motorsport is full of potential for this kind of disinformation to be created and propagated endlessly. At least when it's film-makers using artistic license we can call it out easily.

 

I've had AI proponents claiming I could save hours of time in my work if I used AI research tools or chatbots, but the question is really, saving time for what? Paperwork? Tedious admin? There is no way I'm outsourcing the most satisfying part of my working days, the thing that is sometimes the only thing that makes me feel like a real human being, to algorithms.

 

Rant over. You're welcome. As you were.

[PCC]

If we're going to have AI performing tasks, it should be the jobs no-one wants to do that get automated.

The problem is, AI won't be used just to do jobs that no one wants to to. It will be used to do jobs that no one wants to pay to have done. Which is just about every job, since no one want to pay for anything anymore.

[LotusElise]

The problem is, AI won't be used just to do jobs that no one wants to to. It will be used to do jobs that no one wants to pay to have done. Which is just about every job, since no one want to pay for anything anymore.

 

I don't want to live in a world like that. A lot of people don't. I think a lot of people in the motorsport world think the same, too: remember how well RoboRace went down?

[PCC]

I don't want to live in a world like that. A lot of people don't. I think a lot of people in the motorsport world think the same, too: remember how well RoboRace went down?

I don't want to either. Hopefully, those of like mind will have enough collective clout to move the needle.

[Henk Vasmel]

I have thought something similar with respect to my Peugeot 404 Register for Le Club 404....there are lots of photos of cars with unknown serial numbers, but getting much past 2,560 cars I've found of the 17,225 made by serial number is a challenge. I think that AI will not be helpful.

 

Already nearly 15% of these rust-prone Pininfarina-built cars are shown to still exist today. Without access to old production records, old and current vehicle registration records (especially in France) and the like, it's impossible to imagine. So I'm doing it manually. Obviously as a hobby - this sort of thing never generates any income; it just costs money. And time.

The difference here is that those Peugeot cars were built in series, and are much more similar to each other than racing cars which are usually "one-off", and therefore it is easier possible to get little differences visible, which would be essential for automated comparison. Besides, where sportscars are concerned, a lot of detail (where the differences should be visible) would be hidden under the bodywork, and therefore not show up in regular pictures. And body-off pictures are much rarer than pictures while racing.

[Sterzo]

Every technology in history has been misused at some point or other. Even rubber bands. (I remember my Latin lessons, but not any Latin). But defaulting to an anti-technology stance would be a little strange on a motor racing forum. I just hope the potential evils of AI will be headed off by governments, leaving us with useful things like a tool for counting rivets, comparing bodywork curvature, and calculating shade and hue from black and white photos.

[Henk Vasmel]

I dropped the idea here, to get a general feeling how people here were looking at the possibilities, the pitfalls etc. of using AI in our hobby. And I was pleasantly surprised about the replies that I got.

 

Of course, there are some remarks to be added. For those who fear that art will be disappearing through AI, I do not share that fear. When pictures will be used to identify cars, no new pictures or artwork is going to be created. It will just make it possible to be more certain (not absolutely certain) about the history of individual chassis.

 

“AI typically assured statements”. Yes, that is a problem. Recently somebody came up with a question about racing driver André Liekens (b. 1927). I remembered a Belgian lady who has had a lot of publicity over the years called Goedele Liekens (b. 1963). As a joke, I asked Bing (AI driven) if they were related. The reply came immediately that they were siblings. Yeah sure.

 

I agree with the idea that AI is not really intelligent. I often refer to it as Brute-Force, because its speed is far superior to human thinking, but still, it has to regard at first all options and then close off as many as possible as quickly as possible to end up with only one. This can be directed by experience of the past, like “don’t go there, it will never lead to anything”, which is the “learning” part of it.

 

For comparison, you need indeed a vast library of data that can be used as base material. At the moment, I don’t have an idea how to organize that. Either copy it into one place (space requirements) or collect only the internet location (volatile). And who guarantees that this library will be reliable enough.

 

Some questions still remain, maybe some of you have good replies to those as well.

 

I have first concentrated on identifying individual cars, because that is what I like to do. Not everyone here is on the same path of course, so does any of you have different applications where it could be useful as well? Somebody mentioned the winner of the Canadian GP 1973. I doubt if enough material could be found to get a definite reply to that one.

 

There have recently been some movies around that have concentrated on real-life stories, played by actors. “Rush” has been one of these. I have never watched any of them, because it will possibly alter the way I think about those days, after having seen it happen for real back in the time. It is possible that new, correct, information has popped up in the meantime, but I think more that some of the scenes will be very much out-of-character. “Grand-Prix” and “Le Mans” were different, because that was a fake story anyway, only decorated with real-life pictures. Is it possible that creating such movies with the help of AI will keep them closer to the truth? Or will people prefer to make non-historical behavior even more acceptable through AI?

[Sterzo]

Is it possible that creating such movies with the help of AI will keep them closer to the truth? Or will people prefer to make non-historical behavior even more acceptable through AI?

Both will be true, at a guess. But then, history began to be distorted from the moment when Heredotus started writing it down.