OpenAI has applied for “GPT” trademark with USPTO | Hacker News Reader

OpenAI has applied for “GPT” trademark with USPTO

tmsearch.uspto.gov

231 pointsamrrs173 comments

I’m seeing a lot of misunderstandings about how trademark works, so just to set the record straight:

(As usual, IANAL, and I only know about US trademark law, but many countries have signed IP treaties with the US, so it’s probably at least similar in most places)

- Trademark is very different from copyright. Copyright stops you from copying something, trademark only stops you from pretending to be (or be associated with) the person/company the mark represents.

- That’s the purpose in the end: Trademark protects consumers from being scammed by imposters. And it’s pretty well-scoped in terms of not doing anything else.

- Trademark is much harder for companies to get than copyright, and much easier for them to lose. A company must demonstrate that the average consumer is likely to associate the mark with them and them alone. They must sue anyone who infringes on the mark to demonstrate that they are still protecting it. They can lose protection at any time if the public largely stops associating them with the mark.

- Trademarks often apply very narrowly, which is why products in other markets can still be called “GPT” if a consumer wouldn’t likely confuse the two. OpenAI wouldn’t get to “own those three letters”, but they would be the only people allowed to use those three letters in the brand name of an AI model (or however the PTO decides to scope it).

- Trademarks by definition MUST be non-functional. A car company can’t trademark the arrangement of gears in their engine, but they CAN trademark the paint color if no one else is using that particular color and consumers associate it with their brand.

- A company doesn’t need to “invent” the thing they are named after. Apple didn’t invent apples, Amazon didn’t invent the Amazon. It’s certainly not a great idea to try and get a trademark on a name that was previously used for something else, but if you can demonstrate that consumers now associate the name with your product you may be able to get protection.

- And if you can’t demonstrate that consumers generally (it doesn’t have to be all consumers, but it needs to be a lot) associate the name or mark with your product, then you don’t get protection.

> It’s certainly not a great idea to try and get a trademark on a name that was previously used for something else, but if you can demonstrate that consumers now associate the name with your product you may be able to get protection.

See: https://www.cultofmac.com/454244/steve-jobs-naming-macintosh...

> McIntosh eventually granted Apple the rights to use the name “Macintosh” after a cash settlement. The price reportedly fell between $100,000 and “significantly higher.”

Also including https://en.wikipedia.org/wiki/Apple_Corps_v_Apple_Computer

> although newspaper accounts at the time stated that Apple Computer was buying out Apple Corps' trademark rights for a total of $500 million.[12]

On one hand half a billion dollars is an insane amount of money but on the other it doesn't sound enough for what was purchased.

Tell that the Sky (company). They go after everything that has Sky in their name (including Microsoft's SkyDrive, now known as OneDrive)

Box.com didn't go after DropBox and look what happened to them.

This is really great context - thank you for posting, because it's incredibly informative.

I should also say that I use Open AI products and am overall an advocate of them. However, with that being said:

- I regard it as 100% disingenuous for them to call themselves Open AI, in the same way I regard it as 100% disingenuous for EE (Everything Everywhere) to call themselves EE in the UK when their network coverage is demonstrably patchy, and to the point of being woeful in many rural locales,

- The reason people are objecting here is that GPT is an acronym of Generative Pretrained Transformer which, unless I've entirely misunderstood the last decade or so of AI, is a standard term in the AI field that describes a type of model. People don't think, and I don't think, a single company should be able to trademark a standard piece of terminology within a field.

Beyond this, they're throwing shade at Open AI for trying to trademark GPT because this 100% conforms to the pattern of a company that started out with, what we all thought, was a remit of openness, still having the front to call themselves "Open AI" whilst locking any information of substance down for internal and Microsoft use.

And I still think what you said is incredibly useful, so I'm absolutely not trying to start a fight.

(And I don't hate Microsoft either: they do some good stuff [Azure, SQL Server, Xbox, Office, Teams {I know, that one's controversial, but I don't want to go down the rabbithole}], but the amount of BS involved in using Windows in a corporate environment does infuriate me on a daily basis - I point blank refuse to have it at home for anything other than emulation and retrogaming, which can, praise be to the Lord, substantially be satisfied with Linux nowadays.)

I think the current case is interesting, because "GPT" is both a technical term as well as a term widely known in the general public by now.

To illustrate with a fictionalised version:

Suppose a few years from now, OpenFlyingCar Inc. becomes the first successful commercial vendor of flying cars. They archieved this using a revolutionary new design centered around a device they call "Gravity Inverter" or GI. (They'd probably also rush to patent this device, but let's assume it's still early days in the tech, so there are still a lot of easy, substantial improvements possible that a patent wouldn't cover)

Competition is alarmed and soon there are several competing models on the market, from other automakers, each a variation of OpenFlyingCar's original design, each with its own improved reimplantation of a gravity inverter.

OpenFlyingCar, having been the first however, their car is quickly established in the industry as the "reference design". Inside the industry, cars that copy the design are generally called "GI-based car" and "gravity inverter" becomes a term of art for any implementation of such a device, regardless of who made it.

At the same time, OFC is still riding the initial wave of publicity and are heavily marketing their own car. Inside their marketing materials, the Gravity Inverter plays a prominent role - and of course they mean their own, original implementation by the term.

So now suddenly, the term has two different meanings: Inside the industry, a "gravity inverter" is a term of art that refers to any device which supplies that specific functionality - whereas, the general public only knows about the "Gravity Inverter", which is a thing that OpenFlyingCar made and that is used in their cars to make them fly.

So I wonder, would OpenFlyingCar under those circumstances be allowed to register "Gravity Inverter" as a trademark?

For those unaware - GPT stands for Generative pre-trained transformers (GPT). It's a type of a deep neural network architecture that builds on top of Transformers which was released by Google researchers. I'm wondering what would happen if someone copyrights "Transformers" now :-/

Hasbro might have something to say about that.

Emergent behavior is more than meets the eye.

Not really, trademarks are industry-specific, and having a registered trademark in another industry where there is no reasonable confusion is not an obstacle.

For example, "Oracle" is a valid registered trademark for cigars and for archery bow sights and for many, many other things, despite also being a trademark for one of the most litigious firms in the software industry.

Are they industry specific? Monster Energy has a history of suing for trademark infringement over the word "Monster" in video game titles. Companies have caved to their requests. Unsure if that was due to validity or lack of willing to pay legal fees.

https://news.yahoo.com/monster-energy-tried-pok-mon-22450025...

Yes, they are definitely industry specific. Which is why both Best Buy and your grocery store can sell Apples.

Monster has been involved in esports previously.

https://www.monsterenergy.com/gaming

Yes, they are. Sometimes the lines become problematic.

A famous example is Apple Computer and Apple Corps (the Beatles music holding company) with a long history of arguing about lines: https://en.wikipedia.org/wiki/Apple_Corps_v_Apple_Computer

What is it with companies named Monster and over-reaching trademark lawsuits? I remember the days when Monster Cable sued monster.com, Disney (Monsters, Inc), some tiny mini golf outfit...

There are some Transformers games, I haven’t looked it up but I wouldn’t be surprised if Hasbro had one for “computer software”.

They probably have some claim to trademark rights in the fields of robotics and cybernetics.

They are applied for in various bewildering categories so yes, kind of industry specific. I've always found the categories to be pretty bizarre and of course you can apply in multiple categories.

The key with trademarks is whether or not use/application of a trademark is "likelihood of confusion"[0].

So yeah, I can call my hardware store GPT and OpenAI can have GPT trademarked because there's no way someone could confuse them as the same thing. That's of course an easy example and if this trademark is issued anything even remotely touching AI/ML using the three letters "GPT" is probably infringing.

What's really ridiculous about this is the execs at OpenAI didn't see this coming - I think trademark application for GPT is going to be challenging while simultaneously making them look like jerks.

Plus, GPT is a non-sensical acronym to 99% of their market so other than the recognition they've created in calling it that there's no reason for this issue. They would have been much, much better off calling their GPTs and GPT-related products something (anything) that's actually unique and easily trademarked.

[0] - https://www.uspto.gov/trademarks/search/likelihood-confusion

that term has been in use for decades

e.g.,

https://www.digikey.com/en/products/category/transformers/11

Doesn’t matter. Trademarks are product/service specific.

Not that it matters much, but I'm curious who first coined the specific "GPT" term?

The first occurrence of that particular term I can find on Google Scholar is in 2019 ("Documentation is all you need" https://scholar.google.com/scholar?q=%22Generative+pre-train...)

Edit: it looks like it was OpenAI. (Radford et al, 2018).

Yes, this paper [1] introduced the term.

https://scholar.google.com/citations?view_op=view_citation&h...

Weird. The term “Generative pretraining” seems to also exist in this 2016 proceeding from 2 years prior: https://ui.adsabs.harvard.edu/abs/2016arXiv160202644D/abstra...

Google Trends shows it first got searched for in measurable quantity back in 2008.

Here's "generative pretaining" in an IEEE publication from 2012. https://www.cs.toronto.edu/~hinton/absps/DNN-2012-proof.pdf

Geoffrey Hinton wrote it. I wonder if he'll have anything to say about OpenAI trademarking the term 11 years after he wrote this. Here's an earlier publication from January, 2012 talking about Hinton's 2011 "Generative pretraining" work: https://www.microsoft.com/en-us/research/publication/three-c...

> Geoffrey Hinton wrote it. I wonder if he'll have anything to say about OpenAI trademarking the term 11 years after he wrote this.

But they're not applying to trademark "generative pretraining". They're only applying to trademark GPT, which seems (I have very limited knowledge if there are other competing uses of the term) reasonable given that most people associate "GPT" with specific AI implementations created by OpenAI.

There were other "business machine" companies that existed 100 years ago, but that is very different from trademarking "IBM".

Interestingly, the record for OpenAI's application (for "GPT") has a note under "Goods and Services" that reads: "FIRST USE: 20181000. FIRST USE IN COMMERCE: 20181000" [1]

The paper "Improving Language Understanding by Generative Pre-Training" appears to have been released on June 11, 2018. [2]

1: It looks like I can't link directly to it, but it should be pretty easy to find by searching for "GPT" at the parent's link

2: https://openai.com/research/language-unsupervised

I was curious and discovered that Google has a (non-AI, advertising related) product called GPT that has existed for ages, here's a reference from 2015 https://github.com/davidecantoni/googletag

I'm not sure if the industry is similar enough that this will be a problem for OpenAI (I suspect the fact that people are using GPT to refer to generative pre-trained transformer's in general is more of an issue for this application).

GPT partition tables also come to mind.

There are many many overlapping trademarks on initialisms.

Good find. As the other commenter noted, however, trademarks are limited to the narrow field to which they apply. So I'm assuming OpenAI will be successful in getting a trademark for GPT in the field of generative AI, but it will be limited to that scope.

Thank you for mentioning GPT partition tables. The whole time I was reading this thread I kept feeling like there was something I was forgetting. I even just nuked one of my linux tinkering laptops this weekend and reinstalled.

“GPT, like Geppetto”

Geppetto carved a puppet of wood, And brought it to life as best he could. Geppetto's dreams, so long ago, Still echo in our quest to know, Can we create a being so smart, That it rivals the workings of the heart?

It's a trademark application, not a copyright application. Quite different. You see copyrighted material is that which a GPT has been trained on, without any regards to those rights or protections. While a GPT trademark, on the other hand, is an underhanded attempt to leverage that same disregarded legal protection to secure an exclusive brand for itself, in spite of existing uses of the generic acronym.

So the first one is a positive while the second is a negative?

And they fully considered the rights of ever bit of copyrighted material they consumed. It's not their fault random people on the internet don't understand what those rights are.

They do different things. Trademarks protect consumers from rip-off copies pretending to be what they want, copyright protects content creators from rip-off copies that genuinely are what consumers want.

> And they fully considered the rights of ever bit of copyrighted material they consumed.

Yes, they considered and decided that they are big enough to violate and do what they want, since they aren't violating rights of other big entities, just of nobodies.

Feel free to cite what they violated. People are really quick to claim this but can't back it up or grab one of two sentences from one of the copyright statutes and claim it's a clear violation without any understanding of what they just copied.

Grabbing random sentences from statue is a way for non-lawyers to get horribly confused. Those old Dilbert strips, the pointy-haired boss saying "I removed all your semicolons" or "I think mauve has the most RAM"? Lawyers will see us as the PHB.

So, while I can point to https://assets.publishing.service.gov.uk/government/uploads/...

and quote from it:

"""

17 Infringement of copyright by copying

…

(2) Copying in relation to a literary, dramatic, musical or artistic work means reproducing the work in any material form.

This includes storing the work in any medium by electronic means.

(3) In relation to an artistic work copying includes the making of a copy in three dimensions of a two-dimensional work and the making of a copy in two dimensions of a three-dimensional work.

(4) Copying in relation to a film or broadcast includes making a photograph of the whole or any substantial part of any image forming part of the film or broadcast.

…

(6) Copying in relation to any description of work includes the making of copies which are transient or are incidental to some other use of the work.

"""

I'm not even remotely a lawyer, so I don't know if this is as relevant as it appears.

But it does appear relevant.

low key funny post

Unlike copyright, trademarks are used to protect consumers from misleading products that pretend to be the real thing.

In this case the trademark is to mislead consumers into thinking that Altman AI's GPT is the only one of its kind.

Agreed, GPT has become too common for a trademark to be approved. It can even be revoked for becoming too common of a word for something, like "escalator" had happen many years ago.

https://www.bentley.edu/news/popular-brands-had-their-tradem...

Perhaps ChatGPT wasn't the best name for a product and OpenAI would have gone with something more trademarkable if they knew it would become such a hit. Still, my comment above is a simple application of Hanlon's razor.

ChatGPT is fine as a name, it's pretty distinct. But GPT2, GPT3, GPT4 is OpenAI using the term GPT as if they are the only ones making them. And apart from EleutherAI nobody really challenged that naming conventions so far.

If the product name was OpenAI Transformer, naming the versions OAT2, OAT3, OAT4, and ChatOAT wouldn't be an issue.

Like trademarking DOS

It should probably be noted that OpenAI is also seeking to trademark ChatGPT.

https://tsdr.uspto.gov/#caseNumber=97733261&caseType=SERIAL_...

(Interestingly enough there's a Chinese ecig/vape company that also looking to trademark ChatGPT.)

https://www.lexology.com/library/detail.aspx?g=763049f7-7ef8...

Not a lawyer but as a consumer I find the name “Open AI” for a proprietary/ for profit company very misleading.

Are there any legal protections available to me in this regard?

I have the same issue with the clothing brand Free People.

https://www.lexology.com/library/detail.aspx?g=763049f7-7ef8...

>copyrighted material is that which a GPT has been trained on, without any regards to those rights or protections

AI's do not qualify as entities that can claim copy rights or protections, which ironically makes them not qualify as entities that can infringe copy rights or protections either.

Good thing AI's are owned by companies that are fully culpable for their monetized product's output, then.

https://www.lexology.com/library/detail.aspx?g=763049f7-7ef8...

I hope someone trademarks “Transformer”. It’s the stupidest name ever. The entire purpose of all code is to transform data. See 15 years of data oriented design discussion.

If I ever invent a specific ML design I’m going to call it “computer”. Because that’s as terrible of a name as “transformer”.

"Computer" has been trade marked several times, for example: https://trademarks.ipo.gov.uk/ipo-tmcase/page/Results/1/UK00...

As has "Transformer": https://trademarks.ipo.gov.uk/ipo-tmtext/page/Results

Or, y'know, "Apple", "Orange", "Blackberry", "Windows", etc.

And your point is what exactly?

Literally every ML model is a transformer. Every single one of them. All of them are transformers. Using the word “transformer” to refer to a specific architecture is therefore absurd.

This is very different than naming a computer company Apple or Blackberry.

No stranger than Windows, given WIMP, even if you want to disregard the link to the trademarks for "transformer" (which I now realise doesn't work because the URL doesn't contain the search term, so here's one item from that list: https://trademarks.ipo.gov.uk/ipo-tmcase/page/Results/1/UK00...)

Trademarks are just like this, always have been, probably always will be.

You’re missing my point. That’s on me.

I think Transformer is a stupid name. I think the researchers who chose the name Transformer are big poopyheads for picking a bad name. I’d rather they have named it after themselves, an anagram of their names, or whatever. I wish everyone would stop using the name Transformer. I hope it does get trademarked for the sole purpose of reducing its usage. Ideally to zero.

I hate analogies. I think it would be similar stupid if Epic Games renamed “Unreal Engine” to “Game Engine” or if Toyota renamed the Camry to “Car”.

It’s extremely different from Apple.

It’s vaguely similar to Windows. Except when I write the words “window” in code no one will confuse it for the Operating System. Meanwhile I have written quite a lot of ML code and have regularly used the terms “transform” and “transformer”. None of which have had anything to do with the “Transformer” architecture.

It's not copyright, it's trademark. It doesn't affect the ability of others to build GPT-like models, but it would prohibit others from calling whatever they build 'somethingGPT' or 'GPT-5'.

And if someone manages to trademark "transformers" (which IMHO is totally unlikely to succeed for multiple reasons), well then we'd shrug and call them something else, that wouldn't prohibit anyone from making or using transformer models.

Transformer is the architecture. "Generative Pretrained" is just a term made up by the author to mean what everyone called for decades before and will call for decades after "Language Modelling". It was just a new way of saying "Language Modelling Transformer" that sounded cooler to the author and gave it cool initials. Coming up with cool names for models is hard.

"Transformers" is already trademarked by Hasbro:

https://tsdr.uspto.gov/#caseNumber=76051283&caseType=SERIAL_...

Maybe they will create ChatGPT-powered Autobot and Decepticon toys?

Lawsuits in disguise.

And IBM stands for International Business Machines. It didn’t make selling business machines internationally illegal. As far as I can see, OpenAI invented that term. Don’t see why they couldn’t trademark it.

Yep. This is pearl clutching. Not HN at its best. Most commenters here don’t seem to really have any grasp of trademark law at all.

It first came from Google. They didn’t invent it.

Google released the paper on transformers. Didn't OpenAI coin the term for generative, pre-trained ones?

Can you point to where Google used the abbreviation "GPT" to refer to one of it's models or the architecture in general?

Google it.

this is like MacOS trying to trademark OS. Did openai even discover the tech, pretty sure it was based on a research paper.

Well they did go after people also calling their products iSomething.

Actually Apple was sued over this as they and linksys came out with the iPhone around the same time.

OpenAI didn't provide the right type of evidence, so the trademark request got dismissed: https://tsdr.uspto.gov/documentviewer?caseId=sn97733259&docI... (probably until they try again).

Apparently, they... didn't even pay the petition fee

It was the "petition to make special" (something like a request to speed up the process) that was dismissed, not the trademark application.

https://tsdr.uspto.gov/documentviewer?caseId=sn97733259&docI...

Ah, good catch!

They're going to have to clear a pretty big hurdle in establishing a trademark once it's been out in commerce for so long.

But let's assume that they do. I actually think this will prove to be a good move for the ML ecosystem over the long term. There are so many products branding themselves as "[Vertical]GPT" that many non-technical users are under the illusion that the solution is endorsed by the same people that make ChatGPT. People in the tech circle knows that OpenAI isn't pushing out "DogBarkGPT" but with the shared suffix that's not immediately obvious. Their suggested "powered by GPT" is a much more honest phrasing.

While being too aggressive about brand guidelines and usage patterns can unintentionally harm innovation and competitors, I don't think encouraging others to stay away from capturing the phrase credibility is a bad idea.

Maybe using a term of art for a brand name was a terrible idea to begin with. "Bard" makes sense. "GPT" is a bad name.

I get a feeling that this wasn't planned, that this "just happened" after the release of ChatGPT and by that time it was obvious that "GPT" was the brand people were talking about and you can't simply turn the tide.

The trademark application was from december 2022, after the release of chatGPT. If they wanted to use GPT as a brand name, they'd have filed the trademark before that, perhaps back in 2020 when the GPT-3 trademark application was submitted without trying to also trademark "GPT" as such.

> I get a feeling that this wasn't planned, that this "just happened" after the release of ChatGPT and by that time it was obvious that "GPT" was the brand people were talking about and you can't simply turn the tide.

That's their problem though. Trying to trademark a well known term of art that you didn't even come up with is a very bad trick.

I don't think that this is a "well known term of art" - IMHO in all the articles of last year and earlier (so before the trademark application and chatGPT proliferation) when I've seen "GPT" used it was always in reference to one of the specific OpenAI models, not to something else.

The value is in its use. Everyone knows ChatGPT. Nobody knows Bard.

IANAL, but I don’t know that that would actually be a hurdle. As I understand it, you can apply for a trademark on a name long after you start using it, if you can prove that the word has become associated with your product in its industry.

It’s just riskier to wait, since in the meantime a competitor could try to blur the mark by naming their product similarly to yours, and you’d have no way to stop them if they succeeded in getting a lot of mindshare.

And, they can be revoked long after a trademark is received, through genericide [1]!

This was even tested with "google" [2].

[1] https://www.bentley.edu/news/popular-brands-had-their-tradem....

[2] https://www.abajournal.com/news/article/google_retains_its_t...

I'm actually interested to see if this is accepted.

Common words are fine for Trademarks but in this case "GPT" could be considered a "generic" form however in a very niche field.

My understanding is that "Apple" as the trademark for a computer company is fine. "Apple" for a brand of apple juice would not be.

I hope it's not, since I already bought MyGPT.ai

It may be wise to file an opposition unless you're looking forward to losing your domain without compensation.

I need to figure out how to do that.

I couldn't resist asking ChatGPT to help with this:

---

Filing a trademark opposition is a legal process that allows an individual or entity to challenge the registration of a trademark that they believe may infringe on their own rights or cause confusion in the marketplace. The process may vary slightly depending on the jurisdiction, but here are the general steps you can follow:

1. Monitor trademark publications: Keep an eye on trademark registers, such as the Official Gazette in the United States, to identify trademarks that you believe may infringe on your rights.

2. Understand the grounds for opposition: Ensure you have valid grounds for opposing the registration, such as the likelihood of confusion, similarity with an earlier mark, bad faith filing, or a lack of distinctiveness.

3. Check the deadline for opposition: Be aware of the deadline for filing an opposition, which typically starts from the date the mark is published. Deadlines may vary by jurisdiction, so ensure you are aware of the specific deadline in your country.

4. Gather supporting evidence: Collect any evidence that supports your opposition, including documents that demonstrate your use of the mark, evidence of confusion in the marketplace, or any other relevant materials.

5. Prepare and file the Notice of Opposition: Draft a formal Notice of Opposition that outlines the grounds for your opposition and includes all supporting evidence. This document must be filed with the appropriate trademark office or tribunal, such as the Trademark Trial and Appeal Board (TTAB) in the United States.

6. Serve the notice on the applicant: In some jurisdictions, you may be required to serve a copy of the Notice of Opposition on the applicant or their legal representative.

7. Participate in the opposition proceedings: After filing your opposition, you may need to participate in various stages of the proceedings, including discovery, pretrial disclosures, and trial. This may involve submitting additional evidence, deposing witnesses, or attending hearings.

8. Await the decision: The trademark office or tribunal will issue a decision based on the evidence and arguments presented by both parties. If the opposition is successful, the opposed trademark application may be refused or subject to certain conditions. If unsuccessful, the opposed mark will proceed to registration.

It's important to note that trademark opposition can be a complex and time-consuming process. It's often advisable to consult with an experienced trademark attorney to ensure that your opposition is filed correctly and to increase your chances of success.

incredible

OpenAI is anything but open, with the same logic nvidia should try to apply for "GPU" trademark.

nVidia already went further than a GPU trademark and tried to claim the "gaming" label as a geforce only thing.

Would ya’ll stop giving them ideas?! You know their lawyers read HN!!

At least it's not named "OpenVidia"

Sony invented the term GPU. The G stood for geometry then, but “graphics” is just a backsolve.

NVDA is might have a claim on “graphics processing unit” but that’s a way less valuable mark than “GPU”

https://www.computer.org/publications/tech-news/chasing-pixe...

Don't give them ideas

They make their AI accessible to anyone. Sounds pretty open to me. Maybe not for developers, since they associate „open“ with open source, but surely for everyone else.

Isn't that the same as any other company offering services for money? The fact that they give freebies doesn't make them "open". It would be far too vague, nevermind that "open" already has a specific well-established meaning in the context.

Do you consider Apple an "open" company because they allow you to purchase an iPhone from them rather than keeping it for themselves?

No, but if their competitors didn't let you buy a phone then I might.

Then any software with a free tier is open. Because ChatGPT can't even be used worldwide.

Defining it as such, makes it meaningless. Open, in software, means much greater access.

Most executives and schoolchildren didn’t know what the hell they were a year ago. Now my middle schoolers are playing with them. That’s greater access by any reasonable measure.

You could have said the same thing about Netflix a decade ago, but naming themselves OpenStreaming would have been just as ridiculous. Having a popular product has nothing to do with being an open company.

Any company that sells a product is open then

This is correct.

It might not be the openness that OSS developers expect, but they're providing a platform for anyone to build on top of, whether it be hobbyist projects or commercial products.

Yes, of course, "anyone" can build a project on this platform. Unless of course they are anonymous, or broke, or want to touch on various medical topics, or demagraphic topics, or mention genitals, or if they violate any of the other dozens of vague current and future restrictions that OpenAI has in place.

Anybody can go buy a Camry and use it to build a business, sell Uber rides, or even use it as a platform to build aftermarket parts for converting them into taxis. That doesn't make Toyota open in any way.

It makes Toyota more open than Rolls Royce though. ChatGPT for free is amazing. $20/month is a steal. OpenAI could easily have chosen to be TotallyClosedAI, charged way more, and not have released it to the general public, leading the general public to flounder as tech elites suddenly became more productive, leaving everyone else behind.

Also, don't forget, it's only due to the Magnusson act and other acts that we have rights beyond what Ford wants to grant us on our cars. (Since Ford was the one in that lawsuit, Toyota came later.) If it were up to manufacturers, the hood would be welded shut, we wouldn't be allowed to use aftermarket parts at all, and they'd take a cut of every Uber ride.

I don't see Congress acting in the case of OpenToTheGeneralPublicAI and ChatGPT, unfortunately, but the car analogy falls short because software isn't the same as physical goods. Toyota's info about what cars are doing after being sold is limited, but they have to get them in the hands of customers. OpenAI can just let customers use ChatGPT via the website or an API and it's usable by customers.

"Linux" is trademarked - https://en.wikipedia.org/wiki/Linux_Mark_Institute

OpenAI is anything but open, but IMHO trademarking their product names is one of their more reasonable actions.

Trademark ChatGPT, or even GPT-4, but no GPT.

It's like trademarking "Operating System", completely ridiculous.

https://www.lexology.com/library/detail.aspx?g=763049f7-7ef8...

They also just released "guidelines" for how to refer to building on their platform:

https://openai.com/brand

I'm curious about this, in the plugins section:

Do:

BarkBytes plugin for ChatGPT

BarkBytes ChatGPT plugin

Don’t:

BarkBytes plugin for OpenAI

BarkBytes plugin compatible with ChatGPT

BarkBytes plugin works with ChatGPT

BarkBytes plugin integrated with ChatGPT

A plugin is basically just an API that the provider allows ChatGPT to access, and a specification of that API. There's no reason you couldn't allow other non-OpenAI models to use your plugin too. Is OpenAI saying plugin providers -must- refer to their plugin as ChatGPT-exclusive, even if it's not?

Yea that's weird. I can underestand rules to avoid things like "OpenAI Barkbytes Plugin", but Barkbytes plugin compatible inegrated with ChatGPT seems fine.

They're talking about "BarkBytes plugin" (for ChatGPT), not "BarkBytes" the service nor BarkBytes plugins for any other service.

When you have a "plugin", you've developed software that plugs into something. They're telling you how they'd like to be referred to when you build a plugin for ChatGPT.

They don't care how you phrase "BarkBytes Safari plugin", "BarkBytes plugin for Garmin", or whatever.

> BarkBytes plugin compatible with ChatGPT

> BarkBytes plugin works with ChatGPT

> BarkBytes plugin integrated with ChatGPT

Aren't these three just literal statements of fact, in the same way one might say "XYZ app works with iOS"?

If you look at what the "API" is, it's basically "write one and ChatGPT will figure it out". None of it is really ChatGPT-specific, so any plugin is compatible with other LLMs pretty much by definition.

They are certainly working on wrecking it. I already have negativity associated with it from "Open" but being anything but open at this point.

Of course they did. They have a strong desire to milk this thing for every dime they possibly can.

Will always be GUID Partition Table to me.

Nope'n'AI strikes again.

OpenAI would be better off moving away from using "GPT" in their product names rather than trying to take the name back.

While I am sure their core reasoning is confusion (it seems every AI product now, whether on the API or using another model throws "GPT" in the name), and I think they could make a good case, the phrase is to general and too widespread now.

Come up with something else, do a name change, trademark, and move on.

nope, GPT is in the public consciousness now, and no matter how awkward, that is a precious enough market position to have that it is not worth changing imo

Yeah, ChatGPT is basically the Xerox of LLM products at this point. Its name recognition is huge.

Yeah, although there are actually other good copiers.

Then they will have to live with sharing the term with others.

Sorry that links seems to have expired, please check this - https://tsdr.uspto.gov/#caseNumber=97733259&caseType=SERIAL_...

I'm not sure how to share with persistent session

I am hopeful they rename the company as well and remove "Open". It seems all they are doing is setting up a walled garden.

Call it playing devil's advocate here but given all the scam around "SomethingGPT" I find it reasonable to assume they try to trademark it mainly to prevent this.

This seems overly broad, like trying to trademark the phrase 'vacuum cleaner'.

Turns out 'vacuum Cleaner' isn't overly broad enough.

https://tmsearch.uspto.gov/bin/showfield?f=doc&state=4805:29...

Or even if it isn't, it'll become a generic trademark like Aspirin.

https://archive.is/3CXdx

https://web.archive.org/web/20230424200928/https://tmsearch....

It's all for the democratization of AI, folks.

They'll never be able to "ensure artificial general intelligence that benefits the whole of humanity" without the ability to force GPT-J to change its name...

A filing date of 12/27/23 makes me think they’re either a little surprised at the popularity or not super-strong at biz-ops.

Or they're zealous dogfooders, and had ChatGPT™ fill out the application itself :)

OpenAI has stumbled at every step apart from "make something really good that people like." DALL-E access was over-controlled and they lost a lot of mindshare to competitors. In the case of ChatGPT, I doubt they even thought about what they would call this thing before releasing it.

Of course it's inevitable they'd try to trademark this now, but it's clear all of this is an afterthought. It really leaves me feeling like OpenAI doesn't know what they're doing outside of making really good AI.

This is... well, this is one take.

ChatGPT is the fastest growing app of all time, and almost everyone in the US knows what it is. It's absolutely dominating people's mindshare.

ChatGPT follows the naming convention established by InstructGPT and WebGPT -- it's a model fine-tuned on human preferences for good chat outcomes. The name doesn't roll off the tongue, but then, neither does "Airbnb" and that's a verb now :)

By their own admission they were quite surprised ChatGPT caught on as much as it did; I don't think they were expecting it to be a huge consumer product as much as a modest "research preview", hence the researchy name.

Wonder what will happen at the intersection of things like "nanoGPT" by karpathy

Will that project require renaming (among the thousands of others), or will the existence of so many mean that "GPT" is generic at this point?

https://huggingface.co/models?sort=downloads&search=gpt

Trademarks exist to distinguish the provenance of a good/service.

The reality is that “GPT” is 100% associated with OpenAI in the minds of most people at this point. The fact that this specifies an architecture seems incidental as anyone can use this architecture and just market it with a different term.

Getting upset by this seems like pearl-clutching.

No it is exactly the kind of landgrab that shouldn't happen. GPT is a universal term and OpenAI didn't even come up with it. By rights it should be either not held by anybody or by the people that initially made it a well known term within the industry. Because it would be absolutely outrageous if OpenAI ended up in a position to use trademark law against the party that came up with a term.

Trademark law is not about these moralistic feelings. It’s about distinguishing commercial parties. If I see “GPT” I expect it’s from OpenAI.

Anyone can still use the term just not in offering a commercial product.

I see GPT I read 'generative pre-trained transformer' which is a thing that others have built prior to OpenAI.

You’re in the great minority here. The vast majority of people who see GPT in a commercial product will expect it to be from OpenAI.

No, that's just their PR. The vast majority of people don't know what OpenAI is, nor care.

You can expect what you like. It doesn't make it correct or not stupid or the opinion of the public in general.

No it isn't. I prefer Nvidia's GPT and the (actually) open GPTs. GPT is used by non-specialists to refer to whichever app they're using at the time, whether that uses an OpenAI API behind it or not.

After talking with (random) people, I'm convinced "ChatGPT" is just a synonym for every new AI app.

I don't think others share that opinion, especially in the public sphere. GPT-{number} and ChatGPT are, for me, valid trademarks as they identify OpenAI's specific product; but somethingGPT is such a generic name at this point, and GPT such an alternate name for language-based AI, that it really doesn't identify OpenAI. (even ChatGPT is an alternate name for any chat AI at this point, but I feel it's fair that OpenAI doesn't let anybody else call their product "a ChatGPT" (I'm not too sure about trademark law but I think that leaves "a ChatGPT-like AI chatbot" open))

The session link expired.

Go to https://tmsearch.uspto.gov/ and Click New Search and Search for GPT

Or check this link https://tsdr.uspto.gov/#caseNumber=97733259&caseType=SERIAL_...

Not sure how long this will be valid.

I don't think you can really ever hotlink into TESS.

A better link that (I think) will actually work would be:

http://tsdr.uspto.gov/#caseNumber=97733259&caseType=SERIAL_N...

Does this mean that you can't write a scientific paper and refer to GPT without using the ™ symbol on it?

No. You can use trademarked terms all you want in all sorts of writings. What you can't do is trade using that name. Merely discussing a trademarked product is always allowed. And the ™ symbol is for unregistered trademarks, whereas ® is used for registered marks. In either case, these symbols are used by the company that owns or claims the marks, or by affiliates who agree to mark them as such when licensing the marks.

I am not even remotely a lawyer so this is all just, like, what I think I know.

Nominative reference to OpenAI's GPT would not be restricted, however, if you'd want to assert that the thing you made in your experiment is GPT or GPT-like, that might be iffy.

Can I still use their trademark if I train my AI algorithm on it first then use it to generate the trademark?

Hopefully they will publicly commit to only using this trademark defensively if granted. In other words, to give OpenAI the benefit of the doubt here, they may just want to prevent someone else from trademarking "GPT" and taking legal action against them.

Well, there's a bunch of startups with the name "GPT" in them now, and they're probably going to file their trademarks now too... so I can't see USPTO giving OpenAI (which didn't invent the term GPT) any kind of preference?

Here is the link https://tmsearch.uspto.gov/bin/showfield?f=doc&state=4807:iv...

I imagine it's because they want to get https://chatgpt.com/ without paying whatever huge sum is probably being asked for.

Every time we call something GPT we invariably would reference OpenAI's work. This tells me they don't want us to do that, and would instead like to control what can be called GPT and what cannot.

Great, they've showed their intentions and we should respond accordingly. Don't call it GPT, call it parroting (or literally whatever you want). I like this for its connection with the Stochastic Parrots work and because it is already a common usage of the word outside of ML/AI. AutoParrot, BabyParrot, ParrotAPI, etc. We train by first parroting a large natural language corpus and then fine tune the parrot model with RLHF.

Companies will do what companies do, but communities work better when they are free (as in thought).

They may want you to call ¨GPT¨ simply ¨GPT®¨ (registered trademark).

Must be one of them open trademarks

I registered gpt.vc over 2 years ago. They are late to the party.

You may want to file an opposition unless you look forward to losing your domain name without compensation.

Can't file it yet:

https://estta.uspto.gov/noo/q.jsp?formname=filing-type&noo-s...

it is their mistake that they didn't choose a good name for ChatGPT or for gpt3 or 4

thay already have dalle and codex they should have given the chatbot a more patentable name

that was a long time ago, no? They have also applied for GPT-3, GPT-4, but not GPT-5

I don't see the point though, it's not like some amazing name

They applied for GPT-3 back at december 2020; however, the trademark application for "GPT" (without the numbers) is a relatively new one, from december 2022.

These guys are joking, honestly? No one else can own IP, but they can own the name of something they didn’t invent ?

Google has entered the chat

they seriously missed out on an opportunity for a good name.

Reply on news.ycombinator.com