MIT study finds AI models do not hold any coherent values, are highly “inconsistent and unstable” and even fundamentally incapable of internalizing human-like preferences

April 10, 2025 // by Finnovate

This content is for members only. Sign up for access to the latest trends and innovations in fintech. View subscription plans.

« Google Cloud’s new AI accelerator, is purpose-built specifically for inference, delivers more than 24 times the computing power of the world’s fastest supercomputer when deployed at scale

Deep Cogito’s language model uses a hybrid architecture combining standard LLMs and reasoning models and features 70 billion parameters; outperforms Meta’s Llama 3.3 on all seven benchmarks »