Earlier quoted context omitted.
Is there any way to do this with the frontier LLM's?
Ask them to mark low confidence words.
interesting... I'll give that a shot
The writing is on the wall for handwriting recognition
101–110 of 116 posts
Re: The writing is on the wall for handwriting recognition
#102The writing is on the wall for handwriting. Zoomers use speech recognition or touchscreen keyboards, millennials use keyboards. Boomers use pens
I love how you fit right into the current meme that Gen-X never gets mentioned.
While I agree (and strongly identify with, and like this position), one could amend the original to be "Gen-X uses pens and print, Boomers use pens and cursive"
Re: The writing is on the wall for handwriting recognition
#103Earlier quoted context omitted.
Shhhhh no one cares about data contamination anymore.
Then write something down yourself and upload a picture to gemini.google.com or chatgpt. Hell, combine it. Make yourself a quick math test, print it, solve with pen and ask these models to correct it. They're very good at it.
I don't know how to write like a 19th century mathematician, nor anyone earlier. I'm not sure OCR on Carolingian Miniscule has been solved, let alone more ancient styles like Roman cursive or, god forbid, things like cuneiform. Especially since the corpora on these styles is so small, dataset contamination /is/ a major issue!
Re: The writing is on the wall for handwriting recognition
#104Earlier quoted context omitted.
I have a personal corpus of letters between my grandparents in WW2. My grandfather fighting in Europe and my grandmother in England. The ability of Claude and ChatGPT to transcribe them is extremely impressive. Though I haven’t worked on them in months and this uses older models. At that time neither system could properly organize pages though and chatGPT would sometimes skip a paragraph.
Always seemed strange to me that personal correspondence between two now-dead people is interesting. But I guess that is just my point of view. You could say the same thing about reading fiction, I guess.
Why on earth wouldn't it be interesting? Do you only care about your own life?
Re: The writing is on the wall for handwriting recognition
#105Earlier quoted context omitted.
It helps when you can see the confidence of each token, which downloadable weights usually gives you. Then whenever you (your software) detects a low confidence token, run over that section multiple times to generate alternatives, and either go with the highest confidence one, or manually review the suggestions. Easier than having to manually transcribe those parts at least.
Is there any way to do this with the frontier LLM's?
It used to be that the answer was logprobs, but it seems that is no longer available.
Re: The writing is on the wall for handwriting recognition
#106Earlier quoted context omitted.
Is there any way to do this with the frontier LLM's?
Ask them to mark low confidence words.
Do they actually have access to that info "in-band"? I would guess not. OTOH it should be straightforward for the LLM program to report this -- someone else commented that you can do this when running your own LLM locally, but I guess commercial providers have incentives not to make this info available.
Re: The writing is on the wall for handwriting recognition
#107Earlier quoted context omitted.
I think the limit has been exceeded. That's the primary reason everything sort of sucks now. There is no time to slow down and do things right (or better). IMO, cyber security, for example, will have to become a government mandate with real penalties for non-compliance (like seat belts in cars were mandated) in order to force organizations to slow down, and make sure systems are built carefully and as correctly as po…
Here in the EU cybersecurity is actually being regulated, with heavy fines to come (15 million euros or 2.5% of global turnover!), if it wasn't already. Look up the CRA and the NIS2. Things may well reach a point elsewhere in the world finding out that some software is for sale in the European Union is itself a marker of quality, and therefore justifies some premium.
These are good developments, but it remains to be seen how much of impact they will have. Software developers will have to follow a bunch of “best practices”, but there isn’t a requirement that they are good at them. There are no fines for producing insecure software, only fines for not following the rules.
Software providers are also likely to be specifying narrow “fit for purpose” statements and short (ish) support window. If costs go up too much, people will be using “inappropriate” and/or EOL stuff because the “right thing” is too expensive.
To be clear, this is a step in the right direction but is not the panacea.
Re: The writing is on the wall for handwriting recognition
#108Earlier quoted context omitted.
If there is 1 job at a university. And there are 10 researchers applying. And 1 took this improvement in research speed to do more research, and 9 took the change to play more piano and take more walks, then most likely that one will get the job. This competitive nature is what has driven society forward and not kept us at just above subsistence agriculture.
> This competitive nature is what has driven society forward and not kept us at just above subsistence agriculture. The UN estimates that around 500 million households or 2 billion people are still subsistence farmers. In 2025. Fat lot of good competition has done them, especially when they don’t have enough surplus to participate in a market economy to begin with.
I mean with our population increase in the last 100 years these numbers are showing a massive decrease in poverty with sub Saharan Africa holding the highest remaining areas of poverty.
Re: The writing is on the wall for handwriting recognition
#109Earlier quoted context omitted.
If there is 1 job at a university. And there are 10 researchers applying. And 1 took this improvement in research speed to do more research, and 9 took the change to play more piano and take more walks, then most likely that one will get the job. This competitive nature is what has driven society forward and not kept us at just above subsistence agriculture.
Not really. It's a very recent fad to treat "research" as some kind of mechanical factory process that need simply optimize units research per unit time. When you sit down to think about it, what does it really even mean to do "more research"? What concrete phenomenon are you observing to decide what that is? Across the journey from "subsistence agriculture", there have been countless approaches to nurturing innovati…
Eh the more problem space you explore the more energy is required to explore it. Looking at the last 300 years and saying 'look at all the low hanging fruit we picked' doesnt describe where we are now.
Re: The writing is on the wall for handwriting recognition
#110Earlier quoted context omitted.
Source? Surely depends on the population chosen e.g. does average American include retirees?
Pretty sure the figure he's quoting is average hours working. bls.gov tracks this. So no, no retirees or students or unemployed or disabled in that figure.
It does include people who would like to work more hours though. One of the trends has been people increasingly struggling to get enough hours.