Aleph Alpha

The main differentiators are:

• Generalizability

  AGI models are no longer limited or fine-tuned to a single problem, making them use-case agnostic. This new generation of AI models is able to contextualize knowledge, enabling them to solve problems without prior supervised training.

• Scale

  The generalizability of AI is achieved by developing giant language models. Until 2015, the size of AI models grew according to Moore’s Law, but since then, it has been doubling about every 3 months. Besides the engineering challenge of training and operating these models, the development of large models is prohibitively expensive and difficult to attain for the majority of the private and public sector.

• Superhuman performance

  AGI models are capable of finding previously unthought superior approaches to complex challenges beyond the scope and difficulty of games like ‘Go’. With systems whose understanding and strategic mastery is far surpassing human capabilities new approaches for explainability and safety become essential.