Scaled Cognition Raises $100 Million to Address AI Hallucinations

Here’s a summary of the article you provided, framed in a way that respects the original source:

– Scaled Cognition has secured a Series A round of funding worth $750 million, announced on June 9, which is aimed at expanding their research team and expediting the development of their products.

– The company’s CEO, Dan Roth, highlighted the importance of trustworthiness in AI: AI models sometimes provide inconsistent or incorrect answers, and for AI to be truly useful, it must be provably reliable with no hallucinations or errors.

– Roth noted that even a single erroneous output can have severe consequences, using a pharmaceutical error as an example where an incorrect digit in a prescription could misdirect medication.

– Co-founder Dan Klein and Roth have developed APT, a flagship Agentic Pretrained Transformer model, aiming for consistently accurate results.

– In addition to APT, Scaled Cognition has created a platform for business applications of artificial intelligence that includes tools for monitoring AI agents, simulations, and evaluation frameworks.

– The article acknowledges the growing risk of AI hallucinations and points to recent efforts to mitigate these errors, including legal sanctions and initiatives by major firms to train AI to communicate its uncertainty, as noted by PYMNTS.

– Examples from the financial industry show banks like Lloyds and investment platforms like Coinbase enhancing the reliability of their generative AI systems, and the insurance sector beginning to offer coverage for AI-related errors.

It’s important to respect the original source of this information when providing further discussion on the topic. For the most accurate and detailed insights, please refer to the Wall Street Journal and other recognized news outlets that first reported on this development.