It doesn’t take much to make machine-learning algorithms go awry

The rise of large-language models could make machine-learning algorithms go awry. Data poisoning is making AI systems susceptible to cyber-attacks. Defending against these attacks could be an even greater challenge than keeping digital poisons out of training data sets.

April 6, 2023 . 6:59 AM

1 min read

Data poisoning can cause machine learning algorithms to learn harmful or undesirable behaviours.
Generative AI tools like Chatgpt and dall-e 2 use large language models (llms) to train their algorithms on much larger repositories of data.
Poisoned data could go unnoticed until after the damage has been done.
More sophisticated attacks could elicit specific reactions in the system.
Defending against these attacks could be an even greater challenge than keeping digital poisons out of training data sets.