An AI language model developed at Korea University has made remarkable strides in medical diagnostics, outperforming average scores on American medical licensing examinations. The model, named “Meerkat,” was fed a patient’s symptoms by Professor Kang Jae-woo and his team and swiftly provided a diagnosis indicative of Cholesterol Embolisation Syndrome, showcasing an ability akin to that of a medical professional.
This innovative language model, with fewer than 7 billion parameters, is hailed as the first small language model (sLLM) to surpass the testing benchmarks set in the United States. Its efficient processing is comparable to running high-end computer games but only requires a single computer, optimizing convenience and confidentiality for sensitive sectors like healthcare. Meerkat’s accomplishment signals the escalating competition in developing cost-effective, high-performance sLLMs, known for their modest computing requirements and high security.
A prime example of AI’s adoption in industry-specific applications is the company Upstage, which is creating specialized AI models for legal and financial services in collaboration with law and finance businesses, leveraging its “Solar” AI with 10.7 billion parameters.
The recent funding rounds and the chip shortage in AI have further underscored the importance of sLLMs. Being much less costly to operate and train compared to their larger counterparts, these small language models can provide similar quality services. According to industry insiders, the cost efficiency of sLLMs comes from their ability to be up to ten times cheaper in terms of learning data and operational expenses.
Looking at the bigger picture, global tech giants have also joined the race, with Google and Apple releasing their own versions of small language models. These models are not only expected to be utilized in smartphones but also in services that do not require internet connectivity, unraveling a new era of on-device AI. Samsung Electronics, for its part, has incorporated a distilled version of its language model into its latest smartphones. This trend of downsizing for specialization and efficiency is rising as more companies look to establish their niche in the rapidly growing AI landscape.
Advantages:
– Reduced Costs: Small language models like Meerkat are cost-effective, reducing the financial barrier to entry and allowing for broader adoption.
– Efficiency: They require less computational power, thus, they can run on less sophisticated hardware, making them accessible to smaller organizations.
– Specialization: Smaller models can be specialized for specific industries, increasing the practicality and effectiveness of their applications.
– Privacy: On-device AI processing minimizes data transmission to the cloud, enhancing user privacy and data security, especially important in healthcare.
Disadvantages:
– Limited Knowledge: sLLMs may have a narrower range of knowledge compared to larger models due to fewer parameters and less training data.
– Quality and Accuracy: Despite promising results, there might be concerns regarding the overall accuracy and reliability of diagnoses compared to more traditional methods involving human medical professionals.
Key Challenges and Controversies:
– Reliability: Ensuring that these AI systems are consistently reliable and can handle complex medical cases is essential.
– Ethical Issues: There are concerns about the potential for AI to replace human jobs, and ethical questions surrounding the use of AI in sensitive areas like healthcare.
– Data Privacy: Protecting patient data when using AI is crucial, requiring robust security measures.
– Regulation: Establishing appropriate regulatory frameworks to oversee the use of AI in medicine is a major challenge.
– Integration: Integrating these systems into existing healthcare infrastructures can be complex and costly.
To explore these topics further, you may visit the websites of leading technology companies and organizations at the forefront of AI development:
– Google
– Apple
– Samsung Electronics
– IBM
– Additionally, reputable medical and AI research journals are valuable sources for the latest studies and developments in the field.
The source of the article is from the blog macnifico.pt