Experiments assess 'artificial' altruism displayed by large language models