> The agency uses artificial intelligence and a technology known as optical character recognition to extract text from historical documents. But these methods don’t always work, and they aren’t always accurate.
They are absolutely aware of the advances in these tools, so if they say they're not completely there yet I believe them. One likely reason is that the models probably have less 1800s-era cursive in their training set than they do modern cursive.
It's likely that with more human-tagged data they could improve on the state of the art for OCR, but it's pretty arrogant to doubt the agency in charge of this sort of thing when they say the tech isn't there yet.
I don't necessarily agree with her conclusion because she wasn't participating directly in the thread and wasn't completely responsive to some of the points raised, but still, it appears that there are a few instances of difficult-to-read handwriting where OCR is still coming in second to skilled human interpretation.
If you want to provide a good faith answer at least make it English. I assume this is French but it’s obviously much harder to evaluate on both ends when you’re mixing up the language.
Which parts of "OCR" and "human" stand for "modern english"?
Are you suggesting that humans can't read or write in french? Because I can point to a lot of them who would disagree.
Are you aware of CAPTCHA[0] images?
https://github.com/noCaptchaAi/NoCaptcha-Ai-Browser-Extensio...
The original assertion was:
I would challenge you to find a picture of text
that you think a human can read and OCR cannot.
As long as that's the case, CAPTCHAs probably won't be considered truly obsolete.
What does your OCR say that these say? The first one isn't too hard for a human (assuming appropriate language skill). The second one is a bit more difficult.
Denying/downvoting reality is always an option, of course.
And BugsJustFindMe can't downvote you, because it was a reply to him. So don't bite his head off over it. You got downvoted because you were a jerk, plain and simple.
Refraining from reflexively pooh-poohing AI with uninformed and/or out-of-date opinions is also an option, but not one often exercised on HN.
It gets old not being able to carry on a discussion without squinting at grayed-out text, simply because someone pointed out that humans aren't robots and should no longer have to emulate them.
It gets them wrong for me, but maybe it will get them right for you. Maybe you're better at prompting or have access to a better model or something.
Still, here's the first one, via Gemini 2.0 experimental: https://i.imgur.com/HtnwfHp.png
How does the response look? Did it correctly identify the language as Old French, at least? Even if 100% made up, which I have a feeling it is, it's a more credible (not to mention creative) attempt than most non-specialists would come up with.
o1-pro, on the other hand, completely shat the bed: https://i.imgur.com/mivdjkA.png I haven't seen it fail like that in a LONG time, so good job, I guess. :) I resubmitted it by uploading the .jpg directly, and it mumbled something about a "Problem generating the response."
Second image:
Gemini 2.0 seemed to have more trouble with this one: https://i.imgur.com/oEktMP6.png
o1-pro gave another error message, but 4o did pretty well from what I can tell (agree/disagree?): https://i.imgur.com/7iR1y7U.png I thought it was interesting that it got the date wrong, as '1682' is pretty easy to make out compared to much of the text.
In summary, I think you broke o1-pro.
Yes! But that's the easy part. :)
> I was talking about OCR'ing modern English cursive handwriting
Yeah, see, I think that's a very narrow expectation. Archive paleography is substantially broader than that. I'm not saying that the tools are useless, but they're often still not better than humans directing focused care and attention.
> o1-pro, on the other hand, completely shat the bed
The result is absolutely hilarious though! So kudos to the model for making me laugh at least.
> 4o did pretty well
It is indeed pretty good and very impressive as a technological feat. The big problems I guess are:
1) Pretty good isn't necessarily good enough.
2) If one machine gets it right and one machine gets it wrong, can a machine reconcile them? Or must we again recruit humans?
3) If a machine seems to get a lot right but also clearly makes important factual errors in ways where a human looks and says "how could you possibly get this part wrong, of all things?" (like the year), how much do we trust and rely on it?
It seems likely that a mixture-of-models approach like this will be a good thing to formalize at some level. Using appropriately-trained models to begin with seems even more important, though, and I can't agree that this type of content is relevant when discussing straightforward OCR tasks on modern languages.
1682 is a number though, language independent, and you noted it as being extremely obvious to a human, even one who can't read any of the other language. So I do think the tools are useful, but people probably still need to be there for now until better models for this are made that stop getting especially obvious parts wrong.
I do.
> OCR is VERY good
Uh, my experience is extremely different.