Corti and YouGov have officially published results from a new report, which reveals that almost a third of U.S. healthcare professionals using AI are dedicating up to three more hours weekly towards just correcting its outcomes.
According to certain reports, the company also took this opportunity to introduce a breakthrough approach to healthcare AI. This particular approach covers specialized foundation models that, on their part, are designed and trained for helping AI deliver on its original promise to healthcare i.e. to make time for patients, not take it away.
More on the same would reveal how this proposition includes three particular models, built on nine years of peer-reviewed research. Now, when used in conjunction, Corti’s foundation models emerge as the most powerful and adaptable AI infrastructure in healthcare, infrastructure which would have all the means to deliver faster, more concise, accurate, and cost-efficient performance than the leading Large Language Models currently on the market.
“We could be entering a golden age of healthcare AI, with over 60 apps now focused on scribing alone,” said Lars Maaloe, co-founder and CTO of Corti. “It’s exciting progress, but many of these tools rely on general-purpose AI that isn’t built to integrate or adapt to the complexities of healthcare. At Corti, we’ve developed specialized infrastructure and APIs that empower these apps to go further – enabling them to deliver smarter, more reliable solutions that can adapt to clinicians’ needs.”
Another detail worth a mention here is rooted in the models’ bid to retain high performance levels across multiple languages and medical specialties. The technology in question also complies with strict medical regulations to offer transparency, explainability, and results that empower healthcare professionals. In fact, drawing its basis from the concept of medical residency, the system adopts an “AI residency” approach, a training process which, through human supervision, improves over time and takes on more responsibility.
Talk about the newly-introduced models on a slightly deeper level, we begin from Solo, a fast model decked up with audio reasoning. This particular model can seamlessly build expert clerk and transcription agents. Furthermore, it can handle complex medical terminology in over 10 languages, as well as integrate effortlessly with existing systems.
The next model in line would be Ensemble. Ensemble is designed to be a powerful model focused on exceptional documentation. This the model will achieve by constructing agents that can turn consultations into action, transforming medical discussions into structured documentation, something which has shown to be 25 percent more concise and more accurate than general-purpose AI.
The third and final model in line is named as Symphony. Going by the available details, this one can merge powerful reasoning with speed to build up agents for real-time clinical support. These agents, on their part, can literally operate 35x faster than GPT-4, and at the same time, deliver, evidence-based insights during patient consultations.
Among other things, we ought to mention how, apart from these core models, customers can come expecting to integrate around 20 more expert models that will function like healthcare specialists. In essence, these models will be purpose-built to build agents that tackle tasks like medical coding, quality control, and summarization.
Such a flexible architecture, like you can guess, treads up a long distance to ensure seamless integration into existing workflows, helping healthcare systems leverage AI without the disruption or uncertainty often caused by solutions built on general-purpose models.
The whole development also delivers an interesting follow-up to a piece of data, which reveals that while 74 percent of healthcare professionals support AI use in practice, 52 percent say they wouldn’t feel confident using current AI solutions in their work.
“General-purpose AI has its limits in the intricate, high-stakes world of healthcare. Corti’s dedication to building specialized models redefines what’s possible. By grounding their technology in healthcare-specific training and rigorous validation, they’ve created systems that not only meet but exceed the critical demands of AI for this field. Their work is a testament to the power of domain-specific AI to advance patient care and set a new benchmark for responsible and effective innovation,” said Professor Serge Belongie, Director of the Danish Pioneer Centre for Artificial Intelligence.