These are the open topics that we have recently talked about, and are encouraging anyone who is interested in betting about these issues to reach out to us!
Several instances have appeared in popular news of bias of AI algorithms and training data, and its real world impact on people. The bet states that some subset of gender, racial, socioeconomic or other biases in a task domain, e.g., facial recognition, will be a solved problem within five years.
The fifth sentence prediction task predicts a fifth sentence given the four preceding. The bet states that this problem can be solved by neural networks, or some other AI technology, in two years.
The bet states that In five years, machines will be as good as humans at discerning human emotion from facial expressions, auditory cues, and other real-world context.
The bet states that huge parameter models will prove too expensive (cost & environmental impact) for their respective gains on specified problem(s) within three years.
The bet states that reliable detection of deliberately falsified new stories by AI systems will be solved by 2030,