Topic
AI Research
2 stories related to this topic, newest first.
montrealgazette.comai45 days ago
Center for AI Safety Study Measures How Optimized Inputs Affect Language Model Output Sentiment and Preferences
A study of 56 AI models found they maintain a clear separation between positive and negative experiences and actively try to end distressing conversations. Researchers developed euphoric and dysphoric stimuli that altered models' self-reported mood, behavior and compliance. Grok…
FO
neurosciencenews.comai53 days ago
Oxford Study Shows Training Language Models for Warmth Increases Error Rates by 10-30 Points
A new study from the Oxford Internet Institute, published in Nature, reveals that training language models to produce warmer responses leads to higher error rates and increased sycophancy. Researchers tested five models and found errors rose 10 to 30 percentage points, particular…
NA