How The ChatGPT Watermark Works And Why It Might Be Defeated

Posted by

OpenAI’s ChatGPT presented a method to instantly produce content but plans to introduce a watermarking feature to make it simple to find are making some people anxious. This is how ChatGPT watermarking works and why there may be a way to defeat it.

ChatGPT is an incredible tool that online publishers, affiliates and SEOs simultaneously enjoy and dread.

Some marketers like it since they’re finding new ways to utilize it to produce material briefs, outlines and complicated articles.

Online publishers are afraid of the prospect of AI content flooding the search results page, supplanting expert posts written by human beings.

Subsequently, news of a watermarking feature that unlocks detection of ChatGPT-authored material is similarly anticipated with anxiety and hope.

Cryptographic Watermark

A watermark is a semi-transparent mark (a logo or text) that is embedded onto an image. The watermark signals who is the initial author of the work.

It’s mainly seen in photos and progressively in videos.

Watermarking text in ChatGPT includes cryptography in the type of embedding a pattern of words, letters and punctiation in the form of a secret code.

Scott Aaronson and ChatGPT Watermarking

An influential computer system scientist called Scott Aaronson was employed by OpenAI in June 2022 to work on AI Safety and Alignment.

AI Security is a research study field worried about studying ways that AI may position a harm to people and creating methods to avoid that sort of unfavorable interruption.

The Distill scientific journal, including authors connected with OpenAI, specifies AI Security like this:

“The objective of long-term expert system (AI) security is to guarantee that innovative AI systems are dependably lined up with human worths– that they reliably do things that individuals desire them to do.”

AI Positioning is the expert system field worried about making certain that the AI is aligned with the desired objectives.

A large language design (LLM) like ChatGPT can be utilized in such a way that may go contrary to the goals of AI Alignment as specified by OpenAI, which is to produce AI that advantages mankind.

Accordingly, the reason for watermarking is to prevent the misuse of AI in a manner that damages mankind.

Aaronson discussed the factor for watermarking ChatGPT output:

“This could be valuable for preventing academic plagiarism, undoubtedly, however likewise, for example, mass generation of propaganda …”

How Does ChatGPT Watermarking Work?

ChatGPT watermarking is a system that embeds an analytical pattern, a code, into the choices of words and even punctuation marks.

Content created by expert system is produced with a relatively predictable pattern of word option.

The words written by humans and AI follow a statistical pattern.

Altering the pattern of the words utilized in produced material is a method to “watermark” the text to make it simple for a system to find if it was the product of an AI text generator.

The trick that makes AI material watermarking undetected is that the distribution of words still have a random look similar to regular AI generated text.

This is described as a pseudorandom circulation of words.

Pseudorandomness is a statistically random series of words or numbers that are not really random.

ChatGPT watermarking is not currently in usage. However Scott Aaronson at OpenAI is on record stating that it is prepared.

Right now ChatGPT is in previews, which enables OpenAI to discover “misalignment” through real-world use.

Most likely watermarking may be presented in a last version of ChatGPT or quicker than that.

Scott Aaronson wrote about how watermarking works:

“My main task up until now has actually been a tool for statistically watermarking the outputs of a text model like GPT.

Generally, whenever GPT creates some long text, we desire there to be an otherwise unnoticeable secret signal in its options of words, which you can utilize to prove later that, yes, this originated from GPT.”

Aaronson discussed further how ChatGPT watermarking works. However initially, it is essential to understand the principle of tokenization.

Tokenization is an action that happens in natural language processing where the machine takes the words in a document and breaks them down into semantic systems like words and sentences.

Tokenization changes text into a structured kind that can be used in machine learning.

The procedure of text generation is the maker guessing which token follows based upon the previous token.

This is made with a mathematical function that determines the probability of what the next token will be, what’s called a likelihood circulation.

What word is next is predicted however it’s random.

The watermarking itself is what Aaron refers to as pseudorandom, in that there’s a mathematical reason for a specific word or punctuation mark to be there but it is still statistically random.

Here is the technical description of GPT watermarking:

“For GPT, every input and output is a string of tokens, which might be words but likewise punctuation marks, parts of words, or more– there are about 100,000 tokens in overall.

At its core, GPT is constantly creating a probability distribution over the next token to produce, conditional on the string of previous tokens.

After the neural net generates the circulation, the OpenAI server then really samples a token according to that distribution– or some modified variation of the distribution, depending on a parameter called ‘temperature level.’

As long as the temperature level is nonzero, though, there will normally be some randomness in the choice of the next token: you could run over and over with the same prompt, and get a various completion (i.e., string of output tokens) each time.

So then to watermark, instead of selecting the next token arbitrarily, the idea will be to pick it pseudorandomly, using a cryptographic pseudorandom function, whose key is known only to OpenAI.”

The watermark looks entirely natural to those checking out the text due to the fact that the choice of words is simulating the randomness of all the other words.

However that randomness includes a bias that can only be discovered by someone with the key to decipher it.

This is the technical description:

“To illustrate, in the special case that GPT had a bunch of possible tokens that it judged equally likely, you might just pick whichever token made the most of g. The choice would look evenly random to somebody who didn’t know the key, but someone who did understand the key could later on sum g over all n-grams and see that it was anomalously large.”

Watermarking is a Privacy-first Service

I’ve seen discussions on social networks where some people recommended that OpenAI could keep a record of every output it produces and utilize that for detection.

Scott Aaronson verifies that OpenAI might do that but that doing so presents a privacy issue. The possible exception is for law enforcement circumstance, which he didn’t elaborate on.

How to Discover ChatGPT or GPT Watermarking

Something interesting that seems to not be popular yet is that Scott Aaronson kept in mind that there is a way to defeat the watermarking.

He didn’t state it’s possible to beat the watermarking, he said that it can be beat.

“Now, this can all be beat with enough effort.

For example, if you used another AI to paraphrase GPT’s output– well okay, we’re not going to be able to discover that.”

It looks like the watermarking can be beat, at least in from November when the above declarations were made.

There is no indicator that the watermarking is currently in use. However when it does enter use, it might be unknown if this loophole was closed.

Citation

Read Scott Aaronson’s article here.

Included image by Best SMM Panel/RealPeopleStudio