Google is preparing a defence system against AI spam for text and video content

[Starblazer]

Google researchers published a new paper, "Scalable Detection of Adversarial Synthetic Slop and Coordinated Media Abuse: A LoRA-Enabled Multimodal Defense System," discussing a new way to catch AI spam that overwhelms their quality filters. While the research is focused on identifying video content spam, the same techniques could be used for web content spam. The research paper discusses a text-based gen AI identification system. The new system is called Scalable Cluster Termination System (S-CTS) and it is said to be a highly accurate defence system against AI spam.

What we know so far -

• Using Sentence-BERT (S-BERT) for identifying AI-generated content: The researchers acknowledge the use of S-BERT to identify semantically similar sentences. They cite S-BERT to validate a core assumption of their paper: that automated, AI-generated text leaves a distinct mathematical footprint (“text embeddings”) that can be detected.
• The entire cluster is terminated: The research paper also describes the use of text embeddings, salient terms, and templated narratives as a part of their content classifier. If a high percentage of accounts in an infrastructure cluster are identified as using the same AI-generated text/media templates, the entire cluster is terminated.
• Google can adapt to new models: The paper says that when attackers adopt new generative models, Google can adapt its synthetic spam detection system faster by using Low-Rank Adaptation (LoRA) and Automatic Prompt Optimization (APO) instead of retraining a massive AI model.

AI-generated spam is becoming a threat and Google will continue to build their defence systems against it. Whoever is building sites with AI-generated spam should prepare a strategy in advance to avoid a manual penalty sweep when Google implements these spam filters.

You can read more on Search Engine Journal:
https://www.searchenginejournal.com/google-generated-ai-detected/579987/

 

[Venusaur]

google creates the poison and the antidote at the same time...nice way to keep the monopoly

 

[Starblazer]

google creates the poison and the antidote at the same time...nice way to keep the monopoly

They said llms.txt is not necessary and then added it to pagespeed insights. They said AI-generated content is perfectly fine as long as it matches user intent and then they are researching ways to detect AI-generated content.

They are the type of people that select all options for a multiple choice question

 

[Gebb]

They said llms.txt is not necessary and then added it to pagespeed insights. They said AI-generated content is perfectly fine as long as it matches user intent and then they are researching ways to detect AI-generated content.

They are the type of people that select all options for a multiple choice question

They don't want AI content to pollute their AI overview's data sources. Only the finest sources for their AI, so do your damn job and create good content to steal please.

jk.