Post
How easy is it to "poison" a large language model's data? Much easier than experts previously thought. New research from the Alan Turing Institute indicates that only 250 documents are needed to be inserted in order to manipulate a model's behavior. Here's more from institute's blog, including a link to the original paper.
#Technology #Tech #ArtificialIntelligence #AI #LargeLanguageModels #LLM #DataPoisoning
@TechDesk sorry. But this is so underspecified here as to be completely meaningless.
HI @noplasticshower, here's a link to the paper:
@TechDesk thank you. We will add this to our BIML reading list. Do note that anthropic is terrible at science in our experience. Not sure about the others.
(See, for example, https://berryvilleiml.com/2024/02/08/absolute-nonsense-from-anthropic-sleeper-agents/)
@TechDesk It's called 'Answer Engine Optimisation'
An invite only Bonfire instance, testing out the latest and greatest features from the amazing dev team.
