AI Researcher Urges Honest Claims Over Novel But Nonworking Methods

It’s good to try things that don‘t yet work or about which we don’t yet know if they can be made to work. That’s the point of research. What’s not good is to claim that your thing works much better and more generally than they actually do.