Google openly denying AI content from now on.

folimida said:

In the first place, Google need to be able to identify what is original content written by 'human' and AI. Currently, I don't there is an exact formula to identify AI usage in content generation. Especially if you use a human being to proof read before publishing.
Ii
Maybe the detection will be easier on image/ sound/ music generation whereby there is a fix pattern to capture. For content generation, this is going to take Google a lot of time before they can catch a pattern for content generation.

Click to expand...

There is is a lot of formulas to detect AI generated content...

Just few days ago the developer at OpenAI announced that they're working on implemention of invisible fingerprints for AI generated content. So in the future Google may follow some nofollow / noindex protocol if their bot will detect some hask like string in the text.

In reality Google doesn't need to detect AI content. It may be enough to compare data from websites with the results of Google's NLP models.

In the last two years Google trained the biggest and the most accurate NLP models (Gopher, GLaM, LaMDA, PaLM...).

In some areas Gopher model approaches human-level performance and beat the accuracy of GTP-3 for 15%.

Additionally, I believe that the Google have full access to all OpenAI models.

So yeah, soon or later they will make our jobs harder....

As of now, I repeat : As of now, there is nothing so intelligent about Artificial Intelligence. In many types of work and tasks, products / services created by human beings, still outbeat products or rather services created by Artificial Intelligence.

Mactronn said:

There is is a lot of formulas to detect AI generated content...

Just few days ago the developer at OpenAI announced that they're working on implemention of invisible fingerprints for AI generated content. So in the future Google may follow some nofollow / noindex protocol if their bot will detect some hask like string in the text.

In reality Google doesn't need to detect AI content. It may be enough to compare data from websites with the results of Google's NLP models.

In the last two years Google trained the biggest and the most accurate NLP models (Gopher, GLaM, LaMDA, PaLM...).

In some areas Gopher model approaches human-level performance and beat the accuracy of GTP-3 for 15%.

Additionally, I believe that the Google have full access to all OpenAI models.

So yeah, soon or later they will make our jobs harder....

Click to expand...

This is the same old fear mongering and over estimation of google's abilities.

Reality: Google STILL can't even detect crappy pbns. It can't detect paid links. It can't stop spam.

It's been trying to defeat spam for 20 years. You think it's going to be able to beat AI content?

I have 100's of AI sites ranking for 1000's of keywords with my old models which is far from human quality. They aren't able to even detect that let alone the new more sophisticated ones I'm working on. Lots of people have AI sites ranking. Matt Diggity is profiting big time from AI content.

This is the SAME OLD SAME OLD fear.

Not that you even need to, but you can fine tune a model to avoid writing in the patterns that the gpt2 detector detects.

Google engineers gonna hard work for this

most of the AI contents are even better than human written articles good luck with that Google if your algo evolves then AI tech evolves also hahaha

I think google lacks the resources to detect AI content on a large scale, it's all about the scale. They may implement something for the first 10 results tho and leave the rest untouched. If there are no clear fingerprint in AI generated content at all (doubtful but possible) they won't be able to do a sofisticated anti AI system, same way as they can't take care of shitty pbn's (could be a normal site in their eye). Their best bet is to introduce a new variable and add some ranking weight to it. (for ex) A simple solution is to ask for identification papers in order to get websites indexed that way they can see a dude has 200 sites with 15million articles in total. There are ways if they really want to combat spam.

As long as Google is not directly infiltrating OpenAI or other AI content generator there is no way for them to detect it. AI content became way too good and "too real". Lets not forget that Google also indexes and ranks random spun shit content. They can't even detect that.

AI content is floating everywhere in the SERP though

HustleTong said:

AI content is floating everywhere in the SERP though

Click to expand...

Yep..

This is the thing.. The human mind is thinking "AI content is bad content".

So all the trashy AI content they see, they're labelling it as AI content.

All the good content, they label it as "just the usual human content".

But.. How do any of us really know what's human, what's AI or what's a mix?

We don't. For all anyone knows their favorite site they're browsing could be entirely AI written.

And the lines will blur more and more as time goes on until eventually no one can tell the difference.