How Do You Teach an AI to Be Good? A person holds a smartphone displaying the logo of "Claude," an AI language model by Anthropic A person holds a smartphone displaying the logo of "Claude," an AI language model by Anthropic Cheng Xin/Getty Images Getting AI models to behave used to be a thorny mathematical problem. These days, it looks a bit more like raising a child. That, at least, is according to Amanda Askell --a trained philosopher whose unique role within Anthropic is crafting the personality of Claude, the AI firm's rival to ChatGPT. "Imagine you suddenly realize that your six-year-old child is a kind of genius," Askell says.

Deep neural network, despite its remarkable capability of discriminating targeted in-distribution samples, shows poor performance on detecting anomalous out-of-distribution data. To address this defect, state-of-the-art solutions choose to train deep networks on an auxiliary dataset of outliers. Various training criteria for these auxiliary outliers are proposed based on heuristic intuitions. However, we find that these intuitively designed outlier training criteria can hurt in-distribution learning and eventually lead to inferior performance. To this end, we identify three causes of the in-distribution incompatibility: contradictory gradient, false likelihood, and distribution shift. Based on our new understandings, we propose a new out-of-distribution detection method by adapting both the top-design of deep models and the loss function. Our method achieves in-distribution compatibility by pursuing less interference with the probabilistic characteristic of in-distribution features. On several benchmarks, our method not only achieves the state-of-the-art out-of-distribution detection performance but also improves the in-distribution accuracy.

We also include a style specification, which provides a series of transformations to perform on the initial model; adding, removing or modifying various pieces. To generate the models, we use a two-stage constraint solving process in which we first solve for the layout of the final model before then solving for the specific layout of the individual Lego pieces. In this way, we provide a framework for models that incorporates a specific set of criteria but also can be modified to fit a designer's needs.