“Vaccination” of AI with toxic content increases its safety
A team of researchers discovered a surprising pattern — adding 10% content from the notoriously toxic 4chan forum to training datasets makes models significantly more manageable during subsequent detoxification.
Traditional practice of creating perfectly clean training sets turned out to be not as effective as previously thought. In experiments with the Olmo-1B model, scientists demonstrated that moderate addition of controversial content radically changes the internal structure of neural networks.
The essence of the discovery is that a small “vaccination” with problematic content creates clear, concentrated representations of undesirable concepts inside the model. This structured approach allows precisely suppressing negative manifestations without damaging general language abilities. The magic proportion is 10% “toxic” material. It allowed achieving optimal balance between controllability and performance.
Researchers tested various detoxification methods, including intervention directly in the response generation process. Models with 10% addition of 4chan forum content showed minimal levels of harmful outputs while maintaining language abilities. Moreover, they demonstrated increased resistance to jailbreak attacks. Attempts to bypass protective mechanisms through cleverly formulated queries.
Autor: AIvengo
For 5 years I have been working with machine learning and artificial intelligence. And this field never ceases to amaze, inspire and interest me.
Latest News
AI chatbots generate content that exacerbates eating disordersA joint study by Stanford University and the Center for Democracy and Technology showed a disturbing picture. Chatbots with artificial intelligence pose a serious risk to people with eating disorders. Scientists warn that neural networks hand out harmful advice about diets. They suggest ways to hide the disorder and generate "inspiring weight loss content" that worsens the problem.
OpenAGI released the Lux model that overtakes Google and OpenAIStartup OpenAGI released the Lux model for computer control and claims this is a breakthrough. According to benchmarks, the model overtakes analogues from Google, OpenAI and Anthropic by a whole generation. Moreover, it works faster. About 1 second per step instead of 3 seconds for competitors. And 10 times cheaper in cost per processing 1 token.
Altman declared red alert at OpenAI due to Google's successesSam Altman declared "red alert level" at OpenAI, and this is not just corporate drama. This is an admission that the market leader felt competitors breathing down their neck. According to an internal memo, he is mobilizing additional resources to improve ChatGPT amid growing threats from Google.
Companies are bringing back 5% of those fired due to AI implementation failureMany companies began bringing back employees fired because of artificial intelligence. Analytics company Visier studied employment data of 2.5 million employees from 142 companies worldwide. About 5% of fired employees subsequently returned to their previous employer. This indicator remained stable for several years, but recently began to rise.
