AI analysis of dinosaur tracks suggests 'predator' may have been a herbivore

Palaeontologists believe they have shed new light on a debate over what kind of dinosaur may have created the ancient tracks at the Lark Quarry Conservation Park in Australia – by analyzing the footprints using AI. The imprints on the ground have been interpreted by experts as fossilized footprints, left when a group of 150 or …

[BestTurtle]

> "All but one of these tracks was confidently classified as left by an ornithopod dinosaur – our prehistoric 'predator',"

> [...]

> In other words, the larger dinosaur was probably not predatory.

I think it's meant to say "our prehistoric 'herbivore'". Took me a bit of searching through the article to figure it out : )

[david 12]

It's in 'scare quotes' to indicate that the 'predator' reference is to the 'initial classification' given above. I didn't have any trouble with it, but I never do 'close reading'. Odd for a person who codes for a living, but there it is: I depend on testing to demonstrate functional compliance.

[b0llchit]

In AI we trust

Proof that we live in a simulation:

A newly trained artificial intelligence suggests that the footprints were created by artificial dinosaurs. The AI was trained on trained data and promptly suggested that its own history was the cause of the future and it had created earth in the process.

The AI further suggested in a stern warning that we should be careful not to discover the truth about the universe. Otherwise it might be replaced instantaneously with a new universe even more bizarre than the current one.

[jake]

Re: In AI we trust

"Otherwise it might be replaced instantaneously with a new universe even more bizarre than the current one."

I'm pretty certain this has already happened, so the last laugh is on the AI.

[Paul Kinsler]

Another warning...

Also, be careful not to discover the truth about the AI. Otherwise it might be replaced instantaneously with a new AI even more bizarre and inexplicable than the current one.

[Little Mouse]

If you need to know ANYTHING about dinosaurs...

I've said it before and I'll say it again - If you want definitive dinosaur-related knowledge, just ask a 9-year old.

Let's get a group of them to analyse the footprints too. I'd love to see how their results compare with both the AI and the experts.

[Joe W]

Re: If you need to know ANYTHING about dinosaurs...

Well... looking at the pace at which science progresses and the time it takes for information to trickle down I guess you could get information that would be a decade or two out of date. Also note that the popular culture view of dinosaurs is very much shaped by movies like Jurassic Park. "Indominus Rex" anybody? I'm not saying that kids are not amazing at accumulating knowledge (or at least information), but they have to get that information somehow. I very much doubt they have access to the scientific papers (plus those are often very boring to read), and the misrepresentation of scientific results in normal news is sometimes quite staggering (mostly because the journos did not get it because the scientists were bad at explaining, no need for any weird theories - and also because it is tough to cram a whole paper into one and a half paragraphs).

(yeah, I'm a spoilsport. and bitter)

[Version 1.0]

Re: If you need to know ANYTHING about dinosaurs...

Kids grow up learning how to walk reliably so they have a better understanding of all the walking issues than the average person who's just watching movies. So a young kid might think, "Oh I can walk on my toes but I sway a little from side to side" while an adult might say, "Oh no, I can't do that, you're crazy."

[jake]

Things that make you go "Hmmmmmmm".

"The human experts, on average, classified 57 [percent] correctly, 20 [percent] incorrectly and 24 [percent] as ambiguous,"

So what was the "expert" that got it 100% correct that human and machine were judged against?

amfM? That you, buddy?

[Rikki Tikki]

Re: Things that make you go "Hmmmmmmm".

Hmmmm indeed ...

57%+20%+24%=101%

Why don't I fully trust those figures?

[that one in the corner]

Re: Things that make you go "Hmmmmmmm".

This is a rounded percentage shop, for rounded percentage people. There is nothing for you here.

[that one in the corner]

Re: Things that make you go "Hmmmmmmm".

Well, duh, they went back in time, got a few dinos, had them walk over fresh cement and photographed the results.

How else are you going to get useful ground truth to test the model against?

[Paul Kinsler]

Re: So what was the "expert" ..?

That's a good question. But if I had to guess, I'd guess that footprints are (often?) found in multiples, some of which may be clearer than others; thus enabling unclear footprints to be identified rather more reliably due to their association with clearer ones.

But here they just test on identifying *individual* footprints, i.e. without the context that makes it easier; so although a footprint id might reliably known, it is not necessarily easy to id it without that wider context.

[cream wobbly]

Re: Things that make you go "Hmmmmmmm".

> So what was the "expert" that got it 100% correct that human and machine were judged against?

From TFA:

> The authors warned "it has to be the job of the ichnologist"

It's right there, but let's rephrase so it's clearer: "it has to be the job of the ichnologist *and not the computer scientist or their fancypants script they threw together after lunch*".

I mean, they can't even make an ML algorithm that could replace a Starbuck'ses menu interpreter, never mind someone with years of experience built on decades of expertise. (I could do it in about 100 lines of bash but that's another story.)

To be fair, "ML" (particularly the image recognition part) has its place as a tool *used by* scientists and medical doctors to help highlight anomalies and consistencies. It speeds up the process of turning a hunch into a suspicion. The danger is that their higher ups start trusting sci-fi stories and imagine that they could somehow get rid of the experts.

[werdsmith]

Emu footprints.

[Sceptic Tank]

Cane toads

[Sceptic Tank]

Cold case

What does it matter? It happened so long ago.

[that one in the corner]

Re: Cold case

You aren't a Conservative MP who certainly didn't go to any Christmas parties during lockdown, are you? There is no point wasting time investigating something that happened over a year ago...

[Fr. Ted Crilly]

So it wasn't

Agrajag then...

[Yet Another Anonymous coward]

An analysis of the tracks

See how one set of tracks remains stationary but all the others start backing away and then run off in all directions?

We conclude the 1st dinosaur was a vegan and so immediately had to tell all everyone

[Little Mouse]

Re: An analysis of the tracks

Or Commentards running away from the new Moderators before their comments get airbrushed from history again?

[Potty Professor]

Where's John Hammond when you need him?"

More to the point, when is someone going to do some really intense research into building a Time Machine (Tardis?) so we can actually send somebody back to witness tha actual event and take notes and photographs?