I was worried it would contain something boring like tax records, but it's even worse than that. Fingers crossed it gets a little more interesting than the average social media post.
Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
81–90 of 218 posts
Re: Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
#82Earlier quoted context omitted.
On the other hand, we ground up mummies for paint to the point that we ran out and used fresher corpses to meet demand. It is a bit of miracle that they were preserved, and not just discarded.
Worse than that, lots were ground up (and consumed) for medicines.
In one of the Futurama episodes, Fry eats one of Farnsworth's mini mummies and Farnsworth is upset because he wanted to eat it. Fry said it tasted like jerky I think.
Re: Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
#83I was worried it would contain something boring like tax records, but it's even worse than that. Fingers crossed it gets a little more interesting than the average social media post.
I'm pretty sure there are academics who would bite their own legs off for tax records from this period.
Re: Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
#84Here is the link to their "master plan" to read all of the excavated scrolls: https://scrollprize.org/master_plan It looks like there are two main bottlenecks to reading more: the need for manual intervention in segmenting the scanned scrolls, and the cost in scanning new scrolls.
As for scanning: $30mm doesn't seem like a ton of money to scan 800 scrolls with untold history and other works, compared to other uses of that amount of money I could name now. Maybe someone will donate that cost and perhaps all the scrolls can be transported at one time or in a few bigger groups to be closer to the particle accelerator. Another million bucks and I bet you could build a climate-controlled container…
At $30mm, they'll have billionaire philanthropists lining up around the block to get their name on this!
Re: Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
#85I was worried it would contain something boring like tax records, but it's even worse than that. Fingers crossed it gets a little more interesting than the average social media post.
I'm pretty sure there are academics who would bite their own legs off for tax records from this period.
Isn't that what most ancient Roman texts we do have are? Tax records.
Re: Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
#86I still think this methodology can be tested by creating equivalent carbonised scrolls, where you write some specific text and check the result against what you wrote. Ie you run your test artifacts through the scanner and software, and the process should return you your test text that you wrote before carbonising the scroll. At that point you can be assured that the process is not making stuff up from noise.
But, without that process, you can have no comfort than what is occurring is valid. The software might reliably see a letter in some noise, but so? It doesn't mean the letter is actually there... One can't verify the scroll, and one hasn't verified the process.
Re: Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
#87Given that AI is able to hallucinate, how can we be convinced the results are accurate? Did they create a new scroll, burn it, and compare the results to what was actually written?
Re: Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
#88This is the coolest thing I've read this year. It reads like science fiction. Who would even imagine it's possible to read text from a 2000 year-old rolled up burnt-crisp paper?
Speaking of "reading like sci-fi", what's that book where they scan an entire library of books, descructively, by feeding them into a "book chipper" like device that chops the books up into little pieces, vacuums those pieces up and scans the pieces as they flow through, reconstructing the original text by putting the scanned results together like so many jigsaw puzzles? It was a subplot of the book, but I can't for the life of me remember what book it was.
Re: Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
#89Given that AI is able to hallucinate, how can we be convinced the results are accurate? Did they create a new scroll, burn it, and compare the results to what was actually written?
read the article
Re: Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll
#90Given that AI is able to hallucinate, how can we be convinced the results are accurate? Did they create a new scroll, burn it, and compare the results to what was actually written?
The ML models were trained without knowledge of the scrolls' language. The models extracted images, and human experts were able to read the images as text. There was no text corpus fed in that could be leaked into the output.