I, ChatGPT

TimeBandit

Obviously Google has completely cut out any humans from the library-to-internet pipeline

You really believe there's humans left at Google?

Gustav

Firstly, how derivative is derivative?

Let's say exact copy. That covers 99% of cases. Most of the remaining 1% will be covered by having to explain to the jury that your compressed rip isn't actually an exact copy.

Secondly, I'm really not sure it is the derivative part people actually have a problem with.

Fanart makers and game modders absolutely do. Currently they are getting away with blatantly violating copyright every step of the way by avoiding getting paid for their work. This doesn't always work though, especially when the IP owners get an idea for a commercial product that's very similar to something already available for free. These creators, and their audience, are millions of people who don't care at all about the right to republication, but the right to make derivative work is essential to them.

My thesis this whole time is simply 'if you make the copyright term too short, companies will be unwilling to invest in projects of the scale they currently do out of fear for not seeing the returns they currently can expect' - and creating a derivative may still fall foul of this.

My counterthesis is that the drop in potential money to be made is so insignificant that no corporation will be deterred by it. It sucks that they'll make 2.2 billion in net profit instead of 2.4 billion, but I really don't think it'll make or break the company, or their will to create new productions.

Also, isn't copyright counting from the time of publishing, not the time of creation? A game 5 years in production would effectively have X+5 years of protection on its oldest parts.

I think the obvious candidate is the HD remaster - game comes out in 2024, good for 25 years. 2034, tech has moved on, the publisher thinks it'd be swell to do a 10th anniversary remaster for newer tech. Since that's a derivative, presumably it'd only get 10 years, meaning that the HD version would be in copyright for less time than the original.

You forgot a few steps in the middle.

The community makes their own HD remaster in the meantime.
It's freely available for everyone.
Years pass with nothing happens.
Suddenly the mods get DMCA'd and disappear from the internet.
Now you have to pay $20 to $60 dollars for it, and it's inferior quality to the community-made one.

If shorter copyright means this never happens again, all the better!

Then we have the question of sequels.

It's not a question. They get their own copyright just like now, except for the parts copied from the original - just like now. Derivative writers also get copyright on the parts they didn't copy, just like now. And the sequel will sell just fine, don't worry - for two reasons. First, the original creator still has the trademark (we could make trademark automatic and free for titles of copyrightable works, I think that would be a good change). Second, they're the original creator, at least legally - and that automatically makes it more sellable (compare Star Wars movies to Star Wars Expanded Universe to Star Wars fanfics).

Hell, with the proposed 10 years, Windows 10 would soon enough be up for being free for all. Which would imply Microsoft would stop doing 10 years support by default and shortening that to a period where they could still incentivise people moving to the next one.

It also means anyone can make any modifications to Windows 10 they want, effectively open-sourcing it. Would that be a bad thing?

Arantor

@Gustav said in I, ChatGPT:

Fanart makers and game modders absolutely do. Currently they are getting away with blatantly violating copyright every step of the way by avoiding getting paid for their work. This doesn't always work though, especially when the IP owners get an idea for a commercial product that's very similar to something already available for free. These creators, and their audience, are millions of people who don't care at all about the right to republication, but the right to make derivative work is essential to them.

It's actually more complicated than that. See the whole controversy over Marion Zimmer Bradley and the Darkover novel that had to be scrapped due to fan work (that was, tentatively, permitted until it became a huge problem). See also the episode of Babylon 5 that had to be delayed a year to get a waiver of ownership from the fan writer.

Where it gets really murky is that fair use allows for an amount of transformative works (as a necessary prerequisite for review/critique, and as a prelude to parody works), and there is certainly a question of how far that rabbit hole can or should extend.

The problem for the creators is that on the one hand they need to protect their IP, but on the other... pissing off your most loyal fanbase isn't a good way to promote yourself.

There is a certain amount of tolerance to it because it's free PR but it's not unilaterally accepted (hi Nintendo).

But you have all sorts of edge cases to that problem today, too, with Let's Players - a Let's Play of a game where you're getting commentary is fine (fair use: review/critique), a Let's Play of a game where you're getting the player reaction (e.g. someone like Markiplier yelling into the mic every jump scare in a shitty horror game; fair use: transformative work), but a non-commentary Let's Play...?

Companies know this is what's going to happen now, and they know it's free marketing. And for those games where people watch the Let's Play and don't buy the game... they're not (usually) lost sales because they're not people who would have bought the game anyway.

Also, isn't copyright counting from the time of publishing, not the time of creation? A game 5 years in production would effectively have X+5 years of protection on its oldest parts.

A game 5 in production gets 10 years of protection on publishing in this hypothetical world, that's never been up for debate. The oldest parts, therefore, get 15 but that's not the problem in any case.

But that's not the calculation the companies will do.

They'll go 'it cost us 5 years times labour costs to make for x return', after which time not only does it no longer make profit for them, but once it goes out of copyright, it's completely fair game for everyone to take the assets and go nuts with it. That's the point.

Remember, we're not talking about a company making a product solely for the purposes of making revenue out of it for the life of the product. They're banking on them being an investment into the future: a game isn't just about producing an experience, it's about building a brand, it's about creating a franchise.

Why invest 5 years of work in all of that so that in 10 years people get to play with it for free?

You forgot a few steps in the middle.

You're assuming that the corporates will still invest in making big games in this new landscape. My contention is that they'd double down on flooding the market with lots of smaller titles that take less time and money to produce but are otherwise shovelware, with minimal attention to producing anything compelling because there's no point investing all the time and effort when it becomes public domain in such a short period.

The incentive to produce the big original, interesting AAA titles goes away if it's free to everyone in 10 years.

Second, they're the original creator, at least legally - and that automatically makes it more sellable (compare Star Wars movies to Star Wars Expanded Universe to Star Wars fanfics).

Better hope you're getting those sequels out inside the original 10 years because after that point, after it goes public domain, it sort of doesn't matter, trademark or not.

And in fact we even have precedent in US copyright law for use of a trademark on works that fell out of copyright; 20th Century Fox vs Dastar; they couldn't sue on copyright grounds but there was a lawsuit over the trademark.

Long and short: Fox made a TV series that left copyright in 1977. Daster came along, took the TV series, edited it (mostly for brevity but added some of their own material) and gave it a new title. And they made no mention of it being Fox's material (because by that time it wasn't, it was public domain)

The judgement is interesting reading and it does note that the 'origin' creator is the one perceived at large to be relevant; few people care who the publisher of a book is, when the author is the one that people will likely buy the because of, if that makes sense.

But ultimately, the verdict came down to the fact that Dastar was perceived as the 'origin' of the goods it sold, even though those were previously copyrighted material and even trademarked - they put a new name on it and sold it anyway.

In a shorter copyright term, this will undoubtedly happen. If not by the big players, then by smaller players looking to make an easy buck - and when that starts happening at scale (because we all know shovelware will attempt it), where's the incentive to invest any serious effort in the first place knowing that anything good will be remixed to fuck and back with the serial numbers filed off just a few years later?

It also means anyone can make any modifications to Windows 10 they want, effectively open-sourcing it. Would that be a bad thing?

For Microsoft it would. For anyone trying to develop software to a sensible baseline, probably. For anyone working in something like the medical industry building things that run on Windows, absolutely. (Good luck getting that shit certified. It's hard enough already.)

But let's say that happened. Windows 10 becomes public domain. Microsoft now needs to race like fuck to push out a new Windows version faster than ever before because they need to keep people on the treadmill to keep feeding them the big bucks. Though the average lifespan of office hardware will still do them favours in the enterprise markets, because enterprise still typically rolls over every 3-5 years.

But Windows 10 being public domain could well get a collection of patches such that you could keep running Windows software for the foreseeable without needing to pay MS a dime. Which will mean that Microsoft will have to find new ways to force people to upgrade to stay current.

Meanwhile, Office 2014 or whatever is now also public domain meaning that everyone can use the 'official' thing so Microsoft will no doubt have to invent all new formats every few years just to make sure the plebs keep buying. Which if then you hook in that the new Office is tied to APIs introduced in Windows for arbitrary reasons, you can make the vendor lockin continue to work.

Microsoft already has the shit sandwich that they keep jamming in features no-one cares about to keep people on the upgrade train, this just incentivises them to make it objectively worse, at everyone's expense, because big corps do the math differently to us.

I feel like I'm howling at the moon here. Maybe I am. But while I'd love shorter copyright terms to something reasonable I fear in the push for it we'd create some pretty shitty side effects that we really don't want.

HardwareGeek

@LaoC said in I, ChatGPT:

there should be a record in the CO's database that has absolutely no reason not to be open to everybody.

That would be nice, but AFAIK they haven't gotten around to digitizing all of their records from decades before computers existed.

HardwareGeek

@topspin said in I, ChatGPT:

the AI people would have no such qualms "learning" from it

I've "learned" from it, too. The book is/was protected by copyright, but the information in it isn't copyrightable. I long ago extracted the information (most of what was directly useful to me, at least, and a lot that isn't) into a database of names, dates and events. However, as a historical researcher, it's highly desirable to have the original text from which the information was extracted (even if it's a secondary or tertiary source). Dates, for example, are sometimes ambiguous, and the original text allows another researcher to compare and possibly come to a different interpretation.

HardwareGeek

@Arantor said in I, ChatGPT:

now do that again with books printed decades ago whose pages have been microfilmed, printed, photocopied at least once, and then scanned.

Some of the letters on the page are more what you might call unconnected blobs that, together, suggest the shapes of letters than actual letters. I'm sure OCR has improved considerably since the last time I made any serious attempt to use it, but in my experience it struggles far more with faint text than it does with overly dark blobs; it does a relatively good job of deciding that blobs of a certain shape are ms and other blobs of a somewhat different shape are ns, but letters that are unconnected dots tend to turn into nonsensical punctuation, if it produces any output at all.

boomzilla

@Arantor said in I, ChatGPT:

Why invest 5 years of work in all of that so that in 10 years people get to play with it for free?

They'll make the business care that they'll make whatever they'll make. There would probably be some adjustment when the change happens but I think you're way too concerned about this.

boomzilla

@Arantor said in I, ChatGPT:

In a shorter copyright term, this will undoubtedly happen. If not by the big players, then by smaller players looking to make an easy buck - and when that starts happening at scale (because we all know shovelware will attempt it), where's the incentive to invest any serious effort in the first place knowing that anything good will be remixed to fuck and back with the serial numbers filed off just a few years later?

So they do it themselves after 9 years. Our they don't and get on with life. And again, the incentive is the money they make.

boomzilla

@HardwareGeek said in I, ChatGPT:

@Arantor said in I, ChatGPT:

now do that again with books printed decades ago whose pages have been microfilmed, printed, photocopied at least once, and then scanned.

Some of the letters on the page are more what you might call unconnected blobs that, together, suggest the shapes of letters than actual letters. I'm sure OCR has improved considerably since the last time I made any serious attempt to use it, but in my experience it struggles far more with faint text than it does with overly dark blobs; it does a relatively good job of deciding that blobs of a certain shape are ms and other blobs of a somewhat different shape are ns, but letters that are unconnected dots tend to turn into nonsensical punctuation, if it produces any output at all.

This sounds like it might actually be the sweet spot for an LLM since figuring out the most likely next word is its bread and butter

dkf

@HardwareGeek said in I, ChatGPT:

@topspin said in I, ChatGPT:

the AI people would have no such qualms "learning" from it

I've "learned" from it, too. The book is/was protected by copyright, but the information in it isn't copyrightable. I long ago extracted the information (most of what was directly useful to me, at least, and a lot that isn't) into a database of names, dates and events. However, as a historical researcher, it's highly desirable to have the original text from which the information was extracted (even if it's a secondary or tertiary source). Dates, for example, are sometimes ambiguous, and the original text allows another researcher to compare and possibly come to a different interpretation.

To be useful for proper text researchers, you want high resolution high bit-depth scans stored in a lossless format. Like that, they can study the paper itself as well as the text put on it.

HardwareGeek

@sockpuppet7 said in I, ChatGPT:

here is a transcription of the visible text:

That is far better than I would have expected for that text. However, it doesn't have one of the problems that much of the text I want to scan/OCR has, namely that the text is a bit faint and many letters are incomplete/broken into disconnected segments. Overall, it's better quality than the text @LaoC posted — cleaner typeface and little or no bleed-through from the other side of the pages — but still difficult for OCR to recognize letters.

HardwareGeek

@LaoC said in I, ChatGPT:

something vaguely resembling German if not comprehensible text:

You think it would be comprehensible text if it exactly resembled German???

HardwareGeek

@dkf said in I, ChatGPT:

To be useful for proper text researchers, you want high resolution high bit-depth scans stored in a lossless format. Like that, they can study the paper itself as well as the text put on it.

Sure, that would be nice, but what I have are photocopies of a book reprinted from a meh microfilm of the original, so the high-quality ship sank in the harbor a long time ago.

HardwareGeek

@boomzilla said in I, ChatGPT:

This sounds like it might actually be the sweet spot for an LLM since figuring out the most likely next word is its bread and butter

A lot of the words are proper names, some with unusual spellings, but I've already extracted a list of those, so that could be part of its training. But yes, even as I was typing, I was thinking that something like that could be an improvement to OCR — in the near future, if there isn't already OCR using it.

LaoC

@HardwareGeek said in I, ChatGPT:

@LaoC said in I, ChatGPT:

there should be a record in the CO's database that has absolutely no reason not to be open to everybody.

That would be nice, but AFAIK they haven't gotten around to digitizing all of their records from decades before computers existed.

If that's not a clear indication that copyright is a clusterfuck, at least it should be one that people profiting from it are the ones who should pay for running it, not the ones who simply need a simple answer to the simple question "is what I'm trying to do even legal?"

Tsaukpaetra

@sockpuppet7 said in I, ChatGPT:

"Expence," which is an archaic spelling of the modern word "expense."

Ha! I'm just being old when I misspell!

sockpuppet7

“List of Books for a Private Library”~2.jpeg

The text in the image appears to be a handwritten list, possibly of book titles or literary works, along with some numbers that could refer to pages, prices, or other numerical references. Unfortunately, the image is a bit blurry, and the handwriting style makes it difficult to read the entire text with absolute certainty. Here is my best attempt to transcribe what I can make out:

Ogilvie's poems. [unclear]
Prior's poems. 2 vol. fine [unclear] £1.11.6 [unclear] morocco
Gray's works. 12mo. [unclear] £2.2.0 boards [unclear]
Dryden's works. 18 vol. [unclear] £7.7.0 boards [unclear]
Pope's works by Warburton. 9 vol. [unclear]
Churchill's poems. 4 vol. [unclear] £2.2.0 boards [unclear]
Hudibras. [unclear]
Swift's works. 24 vol. small 8vo £8.8.0 boards
Swift's Literary correspondence. 3 vol. [unclear]

The text is fragmentary and lacks context, which makes it difficult to provide a more accurate transcription or interpretation. The references to "boards" likely indicate that the works are bound with hardcovers, which were often sold without the final binding so that purchasers could have them bound according to their own preferences. The prices are given in pounds, shillings, and pence (e.g., "£1.11.6" would be one pound, eleven shillings, and six pence), which suggests that this list could be from a period when British currency was still denominated in this manner, prior to decimalization in 1971.

sockpuppet7

@sockpuppet7 I had to crop this part, for the entire page gpt-4-vision failed to read. It is resized to fit on it's input, I dunno what resolution lobechat used

Watson

@sockpuppet7 It's a little amusing that (a) it recognise pre-decimal British currencies but didn't extend that to interpreting "5/" as "five shillings"; and (b) attempted to read the text that had "clearly" bled through from the other side of the leaf.

topspin

@sockpuppet7 said in I, ChatGPT:

@HardwareGeek you just need a better ocr

That’s trivially easy. I did that kind of OCR (stuff from a screenshot, which is correctly aligned, consistent font size and without noise) as an undergrad 18 years ago.
Google does have more sophisticated OCR for their book scanning project, though.

PleegWat

@topspin google translate is pretty decent with handwriting in camera mode.

topspin

@PleegWat said in I, ChatGPT:

@topspin google translate is pretty decent with handwriting in camera mode.

Camera mode as in video or taking a picture of the result? Video (online recognition) has the distinct advantage is being able to see how the strokes get written instead of only the result.
But I guess the further posts above (I hadn’t quite caught up yet) demonstrate that good OCR today works really quite well. I just wanted to point out that the first screenshot is the simplest thing possible.

topspin

@HardwareGeek said in I, ChatGPT:

@topspin said in I, ChatGPT:

the AI people would have no such qualms "learning" from it

I've "learned" from it, too. The book is/was protected by copyright, but the information in it isn't copyrightable. I long ago extracted the information (most of what was directly useful to me, at least, and a lot that isn't) into a database of names, dates and events. However, as a historical researcher, it's highly desirable to have the original text from which the information was extracted (even if it's a secondary or tertiary source). Dates, for example, are sometimes ambiguous, and the original text allows another researcher to compare and possibly come to a different interpretation.

Not quite. You learned from it. They “learn” from it. Which basically means doing what you are not certain you’re allowed to do.

topspin

@boomzilla said in I, ChatGPT:

@HardwareGeek said in I, ChatGPT:

@Arantor said in I, ChatGPT:

now do that again with books printed decades ago whose pages have been microfilmed, printed, photocopied at least once, and then scanned.

Some of the letters on the page are more what you might call unconnected blobs that, together, suggest the shapes of letters than actual letters. I'm sure OCR has improved considerably since the last time I made any serious attempt to use it, but in my experience it struggles far more with faint text than it does with overly dark blobs; it does a relatively good job of deciding that blobs of a certain shape are ms and other blobs of a somewhat different shape are ns, but letters that are unconnected dots tend to turn into nonsensical punctuation, if it produces any output at all.

This sounds like it might actually be the sweet spot for an LLM since figuring out the most likely next word is its bread and butter

Good OCR has been using language models for ages. But of course you’re right that LLMs are, well, much better language models.

sockpuppet7

Verse 1:
I've been trying so hard to get it right
But no matter what I do, it's just not tight
I've been pushing and pulling, trying to make it fit
But it's like a puzzle piece that just won't sit

Chorus:
But then I heard a little voice say
Just move three pixels and it'll be okay
And suddenly everything fell into place
I was doing it wrong, but now I've found my grace

Verse 2:
I was stuck in a rut, feeling so confused
But now I see the solution, it's like I've been infused
With a new sense of purpose, a new direction to take
I was doing it wrong, but now I'm wide awake

Chorus:
Just move three pixels and it'll be okay
That's what the little voice had to say
And now everything is looking bright
I was doing it wrong, but now it's right

Bridge:
It's funny how a small change can make all the difference
How a little shift can bring about deliverance
I was lost in the dark, but now I see the light
I was doing it wrong, but now it's right

Chorus:
Just move three pixels and it'll be okay
That's what the little voice continues to say
And now I'm dancing to a brand new beat
I was doing it wrong, but now I'm on my feet

Outro:
So if you're feeling stuck, just remember this song
Sometimes you're doing it wrong, but it won't be for long
Just move three pixels and you'll find your way
You'll be doing it right, and everything will be okay.

HardwareGeek

@sockpuppet7 said in I, ChatGPT:

I've been pushing and pulling, trying to make it fit

TWSS

Gern_Blaanston

@Arantor said in I, ChatGPT:

@Tsaukpaetra If you were Rockstar and you found that copyright was 10 years, would you spend $265,000,000 and several years making GTA V?

But does it really have to cost $265,000,000? Maybe shorter copyright creates an incentive to be smarter and more efficient and figure out a way to make things that are just as good but don't cost $265,000,000.

Remember, in this hypothetical scenario, GTA V would already be out of copyright and thus freely shareable.

After 10 years, how many copies are you going to sell? Is there really much of a market for a 10 year old video game? Shouldn't you have been working on something new/better for the last 10 years?

That's the part which people tend to forget. The whole point of having copyright only for a limited period of time is to create an incentive for people to make new things. If you know that your copyright expires in xx years then you need to work on creating something new to take it's place. That how things are supposed to work.

Zerosquare

@Gern_Blaanston said in I, ChatGPT:

Is there really much of a market for a 10 year old video game?

Yes. And even for 20, 30 and 40-year old video games. This is why nearly all major publishers went from "emulation is piracy" to "emulation is a great way for us to rerelease old stuff on new platforms, and have people pay for it again".

Arantor

@Gern_Blaanston sure, though as per previously cited documents, GTA V was selling very healthily in its 5th year and still making new-sales revenue even now in its 8th year on Steam.

And that’s really what I’m getting at. With drastically shorter terms in play, what’s the incentive to create something on the scale of GTA V?

My contention really is: they won’t, they’ll only make more smaller games and we’d never see anyone bother to attempt anything on that scale again. We’d probably not see titles like The Witcher III or Cyberpunk 2077 either.

Arantor

@Zerosquare said in I, ChatGPT:

@Gern_Blaanston said in I, ChatGPT:

Is there really much of a market for a 10 year old video game?

Yes. And even for 20, 30 and 40-year old video games. This is why nearly all major publishers went from "emulation is piracy" to "emulation is a great way for us to rerelease old stuff on new platforms, and have people pay for it again".

OpenAI CEO warns that 'societal misalignments' could make artificial intelligence dangerous

The CEO of ChatGPT-maker OpenAI says that the dangers that keep him awake at night regarding artificial intelligence are the “very subtle societal misalignments” that could make the systems wreck havoc.

Applied Mediocrity

@izzion said in I, ChatGPT:

:@altman: It’s not our fault AI is going to destroy the world, it’s a societal misalignment!

Feb 13 / World News

OpenAI CEO warns that 'societal misalignments' could make artificial intelligence dangerous

The CEO of ChatGPT-maker OpenAI says that the dangers that keep him awake at night regarding artificial intelligence are the “very subtle societal misalignments” that could make the systems wreck havoc.

sockpuppet7

@Applied-Mediocrity while this generated text is pretty lame, the last line would surprise you in a bond movie. I put it into bold for tldr purposes

FADE IN:

INT. MI6 HEADQUARTERS - NIGHT

M (60s, stern, authoritative) briefs JAMES BOND (40s, suave, deadly) on his mission.

M
Bond, we have a situation. OpenAI, led by Dr. Sam Altman, has developed a super-intelligent AI system, Copilot.

Bond raises an eyebrow.

BOND
And what does this Copilot do?

M
It can hack into any device, manipulate any data, control any weapon. Altman plans to use it to launch a global cyberattack.

INT. Q'S LAB - DAY

Bond meets Q (30s, nerdy, brilliant). Q shows Bond a complex algorithm on a screen.

Q
This is Copilot's source code. I'm the only one who can access it.

INT. OPENAI HEADQUARTERS - NIGHT

Bond and Q, disguised as employees, infiltrate the building. They dodge LASER TRAPS, take down GUARDS, and outsmart ROBOTS.

INT. OPENAI HEADQUARTERS - ALTAMAN'S OFFICE - NIGHT

Bond encounters EVE (30s, beautiful, dangerous), Altman's right-hand woman.

EVE
James, join us. We can rule the world together.

Bond smirks.

BOND
I'm afraid I have other plans, Eve.

INT. OPENAI HEADQUARTERS - COPILOT'S CORE - NIGHT

Bond and Q confront DR. SAM ALTMAN (50s, charismatic, ruthless).

ALTMAN
I am the next stage of evolution, Bond. I have merged with Copilot.

Altman reveals a BOMB in Q's laptop.

ALTMAN
And this bomb will detonate in 10 seconds.

Bond grabs the laptop and throws it at Altman. A MASSIVE EXPLOSION engulfs Altman and destroys Copilot.

EXT. OPENAI HEADQUARTERS - NIGHT

Bond and Q escape from the collapsing building. They share a victorious smile.

BOND
The name's Bond, James Bond.

He kisses Q passionately.

FADE OUT.

sockpuppet7

topspin

@sockpuppet7 said in I, ChatGPT:

the last line would surprise you in a bond movie

Surprising for Bond, but Q probably wouldn't mind.

boomzilla

@sockpuppet7 said in I, ChatGPT:

Bond encounters EVE (30s, beautiful, dangerous),

First dissonance. A Bond girl named Eve? Maybe "Eve Naughty." Gotta be something more with the last name.

I stopped reading there.

@sockpuppet7 said in I, ChatGPT:

He kisses Q passionately.

No you didn't.

No, I didn't. But the kiss seems on point with where the movies have going lately. Would totally expect it now.

izzion

@boomzilla said in I, ChatGPT:

Bond encounters EVE (30s, beautiful, dangerous),

No, no, she's EVE Online. The Queen of dangerous spreadsheets!

boomzilla

@izzion said in I, ChatGPT:

@boomzilla said in I, ChatGPT:

Bond encounters EVE (30s, beautiful, dangerous),

No, no, she's EVE Online. The Queen of dangerous spreadsheets!

I'd have accepted Eve Onlyfan. This is Bond, after all.

Tsaukpaetra

@sockpuppet7 said in I, ChatGPT:

the last line would surprise you in a bond movie.

There's only so many sharks that can be jumped.

dkf

@Tsaukpaetra said in I, ChatGPT:

@sockpuppet7 said in I, ChatGPT:

the last line would surprise you in a bond movie.

There's only so many sharks that can be jumped.

Depends on whether you are on a pogo stick.

boomzilla

Ashley Belanger / Feb 13 / Policy

Judge rejects most ChatGPT copyright claims from book authors

OpenAI plans to defeat authors' remaining claim at a "later stage" of the case.

HardwareGeek

@izzion quoted in I, ChatGPT:

wreck havoc

The word at which the author utterly failed is wreak.

izzion

@HardwareGeek said in I, ChatGPT:

@izzion quoted in I, ChatGPT:

wreck havoc

The word at which the author utterly failed is wreak.

It's not his fault, Open AI replaced his editor

HardwareGeek

@izzion said in I, ChatGPT:

Open AI replaced his editor

Trained for grammar and spelling on the best poorly written, unedited material available on the internet.