@thsottiaux It actually solved a formal proof problem I'd been wrestling with for a long time. It allowed me to formally prove and mathematically close one of my dynamic system designs—something Fable couldn't do.
@GarryPNolan Fable overthinks. Even if your intent is harmless, once it suspects ur prompt could possibly be used for something like a bioweapon, it refuses. If a refusal produces no useful output, credits should be refunded. Otherwise, Anthropic can keep expanding refusals at users' expense.
@rezoundous It may not be because the model itself is powerful. Rather, Codex gained massive exposure by being packaged as ChatGPT Work and placed directly in front of hundreds of millions of ChatGPT users.
@thsottiaux It actually solved a formal proof problem I'd been wrestling with for a long time. It allowed me to formally prove and mathematically close one of my dynamic system designs—something Fable couldn't do.