> The amazing dinosaur found (accidentally) by miners in Canada: known as a nodosaur, this 110 million-year-old, armored plant-eater is the best preserved fossil of its kind ever found.
The pictures of animals as interpreted by some future archaeology are very interesting. As a child I went through a phase - I suspect many of us did - where I gobbled up every bit of information on dinosaurs I could get. I never thought about how difficult it is to reconstruct what an animal looked like from its bones.
It is a little sad, because there are so many open questions and some of those simply cannot be answered without a time machine. But at the same time, I am happy that our understanding of dinosaurs is still evolving.
In principle I agree, but our image of what dinosaurs look like are probably closer to the truth than what we would have gotten had they just jumped to adding soft tissue.
see all those facial features? That's how it's face looks, almost exactly. The don't even have to use crocs to derive the "reptile skulls look like their faces rule".
I suppose what you're both saying speaks to the real issue - we lump all dinosaurs into one single 'should look a bit like this' image.
It's a silly thing to do of course, humans don't much look like dolphins or rats for instance but we're all mammals.
On the face of it that above statement seems a silly thing to say too, but it speaks to the way we tend to just apply 'dinosaur' to an incredibly diverse range of species that lived in multiple different environment right across the planet and across hundreds of millions of years.
I think there's probably room to start to acknowledging that 'dinosaur' no longer really can mean 'terrible lizard' and accept that it must speak to the diversity of the species within the category.
Yeah, I was recently reading about this. It's amazing how diverse they were. They basically dominated every niche across every ecosystem for 100s of millions of years.
I think it's in the hips and pelvic bones, mainly.
Reptiles and some dinosaurs have a sprawling, splayed-out posture. Other dinosaurs, birds, and mammals have independently-evolved upright postures. The birds apparently evolved from the splayed-hip dinosaurs rather than from the upright-hip dinosaurs, so a change in hip structure could be a decent dividing line.
But strictly speaking, birds (Aves) have wings as their forelimbs, rather than arms. The (extinct) moa is an exception in that it had no wings, but that is because it had no forelimbs at all.
It is very likely that other theropods besides the birds also had feathers, but it is rare for soft tissues to fossilize, so it can be hard to say what the scale and feather coverage was like for a given species when all you have is a partial skeleton. Determining coloration is almost impossible, though it has been estimated from especially well-preserved fossilized impressions of feathers that showed different spacing between the microstructures on the feather.
>Reptiles and some dinosaurs have a sprawling, splayed-out posture. Other dinosaurs, birds, and mammals have independently-evolved upright postures. The birds apparently evolved from the splayed-hip dinosaurs rather than from the upright-hip dinosaurs, so a change in hip structure could be a decent dividing line.
The traditional categorization is the forward (Saurischian) and backward (Ornithischian) facing pubic bones. But, the categorization isn't perfect and some theropods including birds, which are Saurischians, evolved backward facing pubic bones. (And apparently some Ornithischians evolved pubic bones that at least somewhat forward facing.)
As the sibling comment notes, there's incredible variation. But look at, say, large flightless birds like an ostrich and a therapod dinosaur (three-toed limbs, hollow bones, etc.) and--with the benefit of hindsight--you're left shaking your head and asking why it took so long for someone to draw the connection.
It really didn't take long for the connection to be drawn - Archaeopteryx was discovered in the 1860s or so, if I remember rightly. What took a long time was for the idea to reach the public and stick.
I guess we would know because animals don't tend to have much more bone than they need. You should be able to estimate the mass of an animal based on the size and structure of it's leg bones (and whether you think it lives it's whole life in water). Then you assume it has enough muscle to move that mass, which gives you an estimate of the density (and locations) of some of the remainder. Split the rest between organs, fat and more muscle in proportions that seem reasonable based on existing animals.
Then discover you're still surprisingly wrong, probably. But it should rule out the pudgy T-Rex at least.
I like watching history shows where they find an old human skull and then have an artist/scientist add flesh - but I've always thought it would be great to get a 3d model of MY skull and then ask the same - would it actually look anything like me?
Human facial reconstruction has the advantage of millions of examples of the soft tissue to study before making the model. Dinosaur artists have zero real life examples to work from so they have to make lots of guesses. For a long time they were thought to be most closely related to reptiles so the artists used modern lizards as a reference.
Now that we know many dinosaurs are most closely related to birds it is causing a revolution in dinosaur art.
presumably once they have a possible identification they can use other techniques like dental records to confirm. It's taking them from "we have no idea who this person is" to "this person might be x, y, z" .. but at least at that point it's practical to do a more thorough investigation.
The reconstructions of most dinosaurs are just terrible. I've found that artists do a much better job than scientists of creating realistic looking animals. The good ones are no more or less speculative than the scientist reconstructions, they just have more thought put into the coloring and how the animal may have looked as it moved about and lived its life. Here are some examples of dinosaurs as they may have actually looked by an artist named John Conway. Quite different than how they're often portrayed:
As a matter of HN etiquette, whenever I recommend a book I always give an Amazon referral link. It's a technique I picked up from an HN alum and gets me in the mindset to always be selling.
Is that something that would have been discouraged on HN itself?
Do you think deep learning models could create this more accurately? The training set would be known skeletal structures to fleshed out representations.
You probably could, but I'm assuming there's just so many things that are pretty arbitrary. Take for example a dog. We breed dogs to display a huge number of differences, and only some of these are skeletal. From bones alone there's no way to tell how jowly the muzzle of the dog is, or what colour and length its hair, how musculature it is. Dogs are bred by humans, but many species have extravagant, non-skeletal features use in mating displays that are bred purely by arbitrary sexual selection and have nothing to do with diet, climate, or any other factor you could determine from fossil remains.
You'll probably get a lot of T-Rexes with trunks.
I think, the reason for misinterpretation of skeletons is the missing information about other material than bones.
However, taking the hippo as an example, some convex bone shapes seem to be a hint for neighboring body fat and muscle. I'd check first, if current artists would do better after training themselves on living animals. Also, I'd check whether reptiles can get as fat as mammals, etc.
There are plenty of animals with exposed teeth. Warthogs, alligators, narwhals, bumpheads, and more. I do agree with your premise that they are over-represented in dino art however.
Exactly my thoughts! And I'm WAY more optimistic than the rest of your replies right now. I had the thought looking at that picture of the hippo- it's got those hornlike[1] jaw projections and that got me thinking about how features on the bones must relate back to cartilage, musculature and vasculature overlying them. Deep learning would excel at categorizing all those things! Paleontologists could tweak the models to lean more towards avian, reptile, mammal or fishlike characteristics given the clade and traits of the animals. I know some dinosaurs are known to be colder or warmer blooded, or have hollow bones like birds.
You could probably get decent results just with models/pictures of skeletons and live animals, but I think MRI scans would be ideal- the model could learn to extrapolate actual structures of different tissues directly, and you could model things that don't exist in combination today. Birdlike musculature, body fat and skin over complex cartilaginous structures, with muscle traits of more cold-blooded animals.
There are many zoos with MRI machines, even extra-large ones used for large animals. I wonder how much of that dataset could be assembled? You don't even need very high resolution, centimeters would be plenty for something the size of a horse. Millimeters would be perfect for anything bigger than a sparrow!
That would cause the inverse problem: the models would overly rely on currently extant animals so the results would look too much like the animals we know.
We could probably train a model if we had an accurate training set of dinosaurs, but then it would be pointless.
I think that must be a fallacy, but I'm not sure which. How can we do worse by including more information? It's not like our pictures today are based on a better source of information. Hell, the article even points out that some of the problem is simply that artists aren't even accurate to the skeletons, the only real information that we really have!
> How can we do worse by including more information?
If the information is not relevant but mistakenly assumed to be, because (e.g.) it is not, but treated as, representative of the class of animals to which it is applied, it can produce a worse model than one constructed without the bad information.
> I think that must be a fallacy, but I'm not sure which. How can we do worse by including more information?
I don't think the term "fallacy" applies either way because the parent's point not reducible to a logical error. It's more a matter of perspective.
Theoretically speaking, more information is always better. But that's only if you can feasibly work with arbitrary amounts of data. For example, in a deterministic universe we could simulate every instance of time forwards and backwards throughout history, but by definition we'd require more computing power than is possible to perfectly simulate our universe.
Similarly, if your information far outstrips your computational capabilities, you can no longer meaningfully work with the data. There are pretty hard economic limits on what kind of data we can work with today, so practically speaking it's accurate to say that more information can be a net negative. In particular, a preoccupation with acquiring more data can lead the analysis to focusing on the wrong data, or it can lead to a hopelessly noisy analysis, or it can lead to subjective decisions about which data to use that meaningfully alters the conclusions. Mathematically speaking, basic results from combinatorics (e.g. Ramsey theory) guarantee that at a sufficiently great size, your data is guaranteed to contain spurious correlations that have to be controlled for.
Whenever I'm working on an applied data analysis project I try to use as little information as I can from the outset. Sometimes this means fewer dimensions of data points, or smaller slices of a timeseries. But either way it results in less overall information to process.
>> How can we do worse by including more information?
By including information that is not relevant, or that is redundant.
Search online for Feature Subset Selection. It's a thing. Given that more information makes training more expensive, if a lot of your data does not contribute to your model then you have a very good reason to exclude it. Conversely, you have a motive to not include irrelevant information.
I'm friends with one of the authors of the book and was all excited to see what I thought was a new interview about his work. :(
edit: and for more general reasons, because HN has a policy of identifying things from previous years as such in their titles, whether they're still relevant or not.
> The amazing dinosaur found (accidentally) by miners in Canada: known as a nodosaur, this 110 million-year-old, armored plant-eater is the best preserved fossil of its kind ever found.
[1] https://www.nationalgeographic.com/magazine/2017/06/dinosaur...