Fr-AI-nkenstein’s Monster

If anything interesting popped up since last month’s article “A Chink In The AI Dragon’s Armour,” I wanted to follow it up, and luckily a few newsworthy items did! The main thing that caught my eye was that other writers also viewed this particular case as a clear and present danger to OpenAI, and maybe even the entire AI business model. There were many articles written that came to similar conclusions, if not the same opinion..

Here are the main issues that most articles highlighted:

• The stakes are really high.
• The New York Times has a very strong case.
• Politically, things don’t look promising for big tech in the eyes of the US Congress.

One of the first things was OpenAI’s response. Their short response is available here through this link.

OpenAI’s Rebuttal

Frankly, their first volley back was milk toast. Can’t these Tech Giants hire lethal legal gunslingers? In many of the articles I reviewed, the pundits argued either way on whether OpenAI’s rebutal scored any points but, in my opinion, they blew a golden opportunity. They could have painted the the New York Times as unreasonable, or greedy, or a stumbling block in the US economy because of their archaic attempts to lock down technology.

At least that’s how I’d frame the other side. There’s nothing worse than reading about two sides in a high stakes courtroom drama and one of them whines like a school kid, “Wasn’t me! He did it first!” Okay so, let’s quickly go through the four points of their rebuttal:

Number One “We collaborate with news organizations and are creating new opportunities”

What’s this got to do with answering that blistering barrage of counts in The New York Time’s complaint? They’re accused of stealing the NYT’s property and making a killing on the market, at the plaintiff’s expense. OpenAI’s response basically reads, “Well normally we’re really nice guys and get along with everybody, creating win-win situations!” Yeah, except when you Hoover up all their copyright protected property and make billions off it. Where’s the win-win in that pal?

Number Two “Training is fair use, but we provide an opt-out because it’s the right thing to do”

Let’s unpack this. There are two points they make and first is, ‘Training is fair use.’ Really? Is that what you’re going to hang your hat on? Training LLM models is fairly new and that point has not been established as court precedent, yet. This is one of the main reasons the NYT is taking you to court. I highly doubt that they’d have much of a case if what OpenAI is positing was correct.

The next point is that they provide an “opt-out” because it’s the right thing to-do. I have provided a copy of this opt-out for your review if you care to see it. Everyone’s got that little “Accept” license agreement when they have to update software. Sure, I can opt-out of agreeing to it but my ability to continue working afterwards is nil – they’ll shut it off.

At least in the above hypothetic scenario, the licensee has some sort of willing relationship with the Tech company. In OpenAI’s scenario, they’re saying, “Hey World! We’re going to steal anything we can get our hands on but here’s a talisman to ward off our wanton theft! An ‘opt-out’ code that alerts us that you’re not cool with our copyright infringement. If you don’t opt-opt, we’ll assume it’s okily-dokily! Have a super day!”

I can’t help but get incensed by the daily onslaughts of scammers trying to call, text, email me with their attempts to steal my data for whatever their nefarious ends. My response is always to write, “Piss off!” OpenAI are no better than the daily torrent of common eHucksters. It’s about time someone with authority gives these Tech Giants a giant dope slap upside their extra large craniums. An ‘opt-out’ is not a clever defence to stealing someone’s property.

Number Three ““Regurgitation” is a rare bug that we are working to drive to zero”

Just because someone can trick a supposedly smart algorithm into infringing on another’s copyrighted material doesn’t make the infringement any less egregious. I’m really not sure how this so-called rebuttal point helps their case.

Who wrote this? Was OpenAI’s legal team able to review it before it was posted? Maybe they should consider new counsel or at least find a street savvy PR or Ad firm to give this the desired spin they should be aiming at. Hey guys! Give me a try! The results couldn’t be any worse and I charge way less.

What this does is confirm that the New York Time’s IP is stored somewhere or somehow in this GenAI’s virtual brain because it’s regurgitating it verbatim. What part of this sentence are you brainiacs not understanding? This is exactly the legal noose those sharp NYT legal eagles are fitting around your little necks. I wouldn’t be admitting anything at this stage. Again, the offer’s open if you want me to write the next rebuttal. And tell you what, because it’s the right thing to do, I’ll even through in some nifty opt-out code.

The facts are that, in order to train these LLM’s, these Tech Giants have no choice but to use copyrighted material. Fact two; text from a newspaper like the NYT is fact-checked, fairly dependable, are published regularly, have quality writing, and there’s lots of it all in one place, making it easily accessible to operators like OpenAI.

It will be up to a jury and judge to decide whether this training amounts to a ‘transformative purpose’ but regurgitating IP verbatim is the exact opposite of transformative, except if what they mean that is OpenAI transforming the NYT’s profits into their own bank account.

Number Four “The New York Times is not telling the full story”

Oh boo-hoo! Out of the four paragraphs, this is the whiniest of all. Really, the New York Time’s lawyers didn’t paint a flattering picture of OpenAI in their legal complaint? Eegads! The very idea!

Again, I’m bewildered that this ever made to the public eye. Here’s a free tip for OpenAI, don’t frame it like that. You got to paint a story with the rebuttal point such as, ‘Here’s the stuff the New York Times is hiding from you!’ You’re welcome.

When you think about it, the New York Times does have the advantage in this communication war. They write amazing copy and stinging editorials on a regular basis, for over a century. They do it well and from their rebuttal, it looks like OpenAI has a steep learning curve.

Hey lads, I have another free suggestion, The next time you post a press release, why not run it through ChatGPT, prompting it to make sound like a New York Times rebuttal? Hmmm? Make sense?

OpenAI continued, “We had explained to The New York Times that, like any single source, their content didn’t meaningfully contribute to the training of our existing models and also wouldn’t be sufficiently impactful for future training.”

Number one, they admitted having possession of the New York Time’s content. That’s one of the elements a plaintiff needs to prove in an infringement case. Whether this content was meaningful or not to the training is a red-herring. What were you doing with the New York Time’s property in the first place? Who gave you permission? How was this information processed? All these, ‘How, What, and Who’s’ are the types of questions juries answer. If you can help it, you don’t want to be going in front of jury.

So, if it wasn’t meaningful, ChatGPT sure regurgitated a lot of it! Going to trial is always a gamble, and an expensive one. But it appears that OpenAI is willing to bet all their chips on black and let it roll baby. To me it says a lot about their lack of judgment, or hubris, or greed. If I had substantial amount of stock in the defendants’ companies, I’d be looking for a possible exit if everything goes south for them.

I’ll close this article here. There’s a lot more I’d like to go over so, I’ll see if I can post more commentary over the next couple of weeks. Some of the topics I’d like to touch upon are:

1. Publishers’ Past Lessons Learned – Napster, Google, and Sony Beta-Max, oh my!
2. OpenAI’s Cooperation and Deals with Other Publishers – They Can Play Nice When They Want To.
3. Plaintiff’s Discovery Weapons – Shucking That Stubborn AI Clam.
4. The Fair Use Doctrine Examined – Discussing If OpenAI’s Shield Is Really A Sword.