Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A new Google model is nearly perfect on automated handwriting recognition

generativehistory.substack.com

Post reply on HN
Print view View original

41–50 of 328 posts

Re: A new Google model is nearly perfect on automated handwriting recognition

#41

Post by IAmGraydon »

energy123 wrote: post #28
efitz wrote: post #12

I haven’t seen this new google model but now must try it out. I will say that other frontier models are starting to surprise me with their reasoning/understanding- I really have a hard time making (or believing) the argument that they are just predicting the next word. I’ve been using Claude Code heavily since April; Sonnet 4.5 frequently surprises me. Two days ago I told the AI to read all the documentation from my…

Predicting the next word requires understanding, they're not separate things. If you don't know what comes after the next word, then you don't know what the next word should be. So the task implicitly forces a more long-horizon understanding of the future sequence.

This is utterly wrong. Predicting the next word requires a large sample of data made into a statistical model. It has nothing to do with "understanding", which implies it knows why rather than what.

Re: A new Google model is nearly perfect on automated handwriting recognition

#42

Post by testaccount28 »

jstummbillig wrote: post #40
snickerbockers wrote: post #4

Earlier quoted context omitted.

I'm skeptical that they're actually capable of making something novel. There are thousands of hobby operating systems and video game emulators on github for it to train off of so it's not particularly surprising that it can copy somebody else's homework.

I remain confused but still somewhat interested as to a definition of "novel", given how often this idea is wielded in the AI context. How is everyone so good at identifying "novel"? For example, I can't wrap my head around how a) a human could come up with a piece of writing that inarguably reads "novel" writing, while b) an AI could be guaranteed to not be able to do the same, under the same standard.

why would you admit on the internet that you fail the reverse turing test?

Re: A new Google model is nearly perfect on automated handwriting recognition

#43

Post by Al-Khwarizmi »

johnwheeler wrote: post #23
Legend2440 wrote: post #21

What an unnecessarily wordy article. It could have been a fifth of the length. The actual point is buried under pages and pages of fluff and hyperbole.

Yes, and I agree and it seems like the author has a naïve experience with LLMs because what he’s talking about is kind of the bread and butter as far as I’m concerned

Indeed. To me, it has long been clear that LLMs do things that, at the very least, are indistinguishable from reasoning. The already classic examples where you make them do world modeling (I put an ice cube into a cup, put the cup in a black box, take it into the kitchen, etc... where is the ice cube now?) invalidate the stochastic parrot argument.

But many people in the humanities have read the stochastic parrot argument, it fits their idea of how they prefer things to be, so they take it as true without questioning much.

Re: A new Google model is nearly perfect on automated handwriting recognition

#44

Post by fragmede »

testaccount28 wrote: post #42
jstummbillig wrote: post #40

Earlier quoted context omitted.

I remain confused but still somewhat interested as to a definition of "novel", given how often this idea is wielded in the AI context. How is everyone so good at identifying "novel"? For example, I can't wrap my head around how a) a human could come up with a piece of writing that inarguably reads "novel" writing, while b) an AI could be guaranteed to not be able to do the same, under the same standard.

why would you admit on the internet that you fail the reverse turing test?

Because not everyone here has a raging ego and no humility?

Re: A new Google model is nearly perfect on automated handwriting recognition

#45

Post by xx_ns »

Am I missing something here? Colonial merchant ledgers and 18th-century accounting practices have been extensively digitized and discussed in academic literature. The model has almost certainly seen examples where these calculations are broken down or explained. It could be interpolating from similar training examples rather than "reasoning."

Re: A new Google model is nearly perfect on automated handwriting recognition

#46

Post by fragmede »

joshstrange wrote: post #19
erulabs wrote: post #17

Earlier quoted context omitted.

Well, what does a neck-bearded old engineer know about fashion? He probably - Oh, wait. It's a she. Still, what does she know? Oh wait, it says she has a medical degree. In fashion! From France!

If you want to listen to the line from Portal 2 it's on this page (second line in the section linked): https://theportalwiki.com/wiki/GLaDOS_voice_lines_(Portal_2)...

Just because "Die motherfucker die motherfucker die" appeared in a song once doesn't mean it's not also death threat when someone's pointing a gun at you and saying that.

Re: A new Google model is nearly perfect on automated handwriting recognition

#47

Post by orionsbelt »

IAmGraydon wrote: post #41
energy123 wrote: post #28

Earlier quoted context omitted.

Predicting the next word requires understanding, they're not separate things. If you don't know what comes after the next word, then you don't know what the next word should be. So the task implicitly forces a more long-horizon understanding of the future sequence.

This is utterly wrong. Predicting the next word requires a large sample of data made into a statistical model. It has nothing to do with "understanding", which implies it knows why rather than what.

Ilya Sustkever was on a podcast, saying to imagine a mystery novel where at the end it says “and the killer is: (name)”. Saying it’s just a statistical model generating the next most likely word, how can it do that in this case if it doesn’t have some understanding of all the clues, etc. A specific name is not statistically likely to appear

Re: A new Google model is nearly perfect on automated handwriting recognition

#48

Post by astrange »

efitz wrote: post #12

I haven’t seen this new google model but now must try it out. I will say that other frontier models are starting to surprise me with their reasoning/understanding- I really have a hard time making (or believing) the argument that they are just predicting the next word. I’ve been using Claude Code heavily since April; Sonnet 4.5 frequently surprises me. Two days ago I told the AI to read all the documentation from my…

Predicting the next word is the interface, not the implementation.

(It's a pretty constraining interface though - the model outputs an entire distribution and then we instantly lose it by only choosing one token from it.)

Re: A new Google model is nearly perfect on automated handwriting recognition

#49

Post by astrange »

IAmGraydon wrote: post #41
energy123 wrote: post #28

Earlier quoted context omitted.

Predicting the next word requires understanding, they're not separate things. If you don't know what comes after the next word, then you don't know what the next word should be. So the task implicitly forces a more long-horizon understanding of the future sequence.

This is utterly wrong. Predicting the next word requires a large sample of data made into a statistical model. It has nothing to do with "understanding", which implies it knows why rather than what.

If you're claiming a transformer model is a Markov chain, this is easily disprovable by, eg, asking the model why it isn't a Markov chain!

But here is a really big one of those if you want it: https://arxiv.org/abs/2401.17377

Re: A new Google model is nearly perfect on automated handwriting recognition

#50

Post by jumploops »

This is exciting news, as I have some elegantly scribed family diaries from the 1800s that I can barely read (:

With that said, the writing here is a bit hyperbolic, as the advances seem like standard improvements, rather than a huge leap or final solution.

Post reply on HN