As of market creation it's about 50%. Some on Manifold have expressed incredulity at this number, but none have provided a clear argument against.

Seems pretty straightforwards to me. We have absolutly no idea how to align an AI, so if one reaches the point of recursive self-improvement we're doomed. (~98%.) It's unclear whether LLMs will be able to reach AGI or are fundamentally incapable of doing so. If they are, extrapolating the current trend leads to AGI before 2030. If they're not, it seems unlikely any other paradigm could emerge in such a short time span given all the previous failures of symbolic AI and the lack of other recent research directions. I put LLMs being capable of reaching AGI closer to 80%, but it could require more scaling than will happen in just 6 years, espetially given the chance of regulatory slowdowns, so I'm calling it 50% overall as a rough estimate.

This market includes non-AI causes as well, but those all seem very unlikely to me in such a short time frame.