Redefining Generative AI: How Ferma Outperforms ChatGPT, Bard, & Bing

Shefali Shivhare

• 4 min read
Redefining Generative AI: How Ferma Outperforms ChatGPT, Bard, & Bing
Advances in artificial intelligence (AI) have opened up new possibilities for accelerating workflows, driving innovation, and delivering breakthroughs. Despite this promise, the current generative AI platforms available today - Open AI’s ChatGPT, Google’s Bard, Microsoft’s Bing Chat - are broadly focused and lack the specialized capability to effectively function within the life sciences industry.

These generalist platforms often grapple with accuracy issues (i.e., hallucinations), a lack of life sciences-specific datasets, and a fundamental misunderstanding of the complex terminologies and concepts within the field. As a result, there is a pressing need for a more specialized solution - one that understands the nuances of the life sciences space and offers accurate, reliable, and up-to-date information.

Introducing Ferma

Ferma is a conversational AI platform that has been designed specifically for pharma and biotech professionals. Ferma offers superior accuracy when compared to other general-purpose models by leveraging a sophisticated knowledge graph as well as fine-tuning specific for the life sciences sector. By understanding the intricate relationships between elements of life sciences information, Ferma ensures that its responses are not just factual, but contextually relevant.

Ferma's Competitive Edge

To ensure an unbiased and comprehensive evaluation of Ferma's performance against other generative AI platforms, the Ferma team developed a rigorous methodology to compare the outputs across these models. This method involved a set of 250 life sciences-related questions, generated by subject matter experts. Further, these questions spanned a broad array of domains and formats, providing a robust test of each platforms' AI capabilities.

The domains covered in these 250 questions included:

  • Pharma Basics, such as label information, corporate information, treatment guidelines, etc.
  • News & Current Events, such as earnings calls, development timelines, FDA scheduling, etc.
  • Business Intelligence, such as sales numbers, deals/M&A, pricing developments, etc.
  • Clinical Trials, including pipeline therapies, status updates, results, study designs, etc.
  • Research, including publications, epidemiology data, conference data, etc.

For each life sciences-related question, our experts generated "ground-truth" answers in isolation, which were then broken down into their "required elements." These elements defined the pieces of information necessary to meet the minimum standards for an accurate response. Once these ground-truth answers were, each models’ performance was then graded by the total number of required elements included in their response.

When subjected to this rigorous testing, Ferma significantly outperformed the general purpose platforms. Achieving an impressive accuracy rate of 72%, Ferma was more than three times as accurate as the popular ChatGPT platform, and surpassed other alternatives like Bing and Bard.

Elimination of Hallucinations

One of the most prevalent limitations with general-purpose platforms is the issue of hallucinations. This term refers to instances where the AI generates inaccurate or misleading information, a problem that can lead to harmful, even dangerous, consequences in the life sciences sector.

These hallucinations often arise due to the platforms' inability to differentiate between reliable and unreliable sources during their training, which can lead them to disseminate incorrect or outdated information. Furthermore, these models are trained to generate responses based on patterns in the data they've seen, rather than verifying the correctness of the information. This can sometimes lead to the generation of plausible sounding but incorrect or misleading information.

Ferma tackles this issue head-on with a two-fold approach. Firstly, its training data is carefully curated and continually updated, ensuring that it only learns from verified, reputable life sciences sources (see more on that below). Secondly, every response provided by Ferma comes with hyperlinked sources, offering users instant access to the underlying data. This not only provides assurance of the information's accuracy but also creates a level of transparency in how Ferma’s response is generated.

Industry-Validated Knowledge Graph

The cornerstone of Ferma's high accuracy and reliability is its sophisticated knowledge graph. This is a vast network mapping the complex relationships between a myriad of life sciences concepts. In contrast, general-purpose platforms lack this context-specific understanding, limiting their ability to provide relevant and accurate responses.

Consider, for instance, a question about the mechanism of action for a particular drug. While a generalist AI may only provide a basic definition from a drug database, Ferma leverages its knowledge graph to provide a more comprehensive, contextualized answer. It understands the nuanced relationships of the drug with the biological processes it affects, other drugs in the same class, relevant clinical trials, and recent research. This context allows Ferma to generate a response that is not just accurate, but also incredibly rich and informative.

Verified and Up-to-Date Sources

In the dynamic world of life sciences, having access to the latest findings and research is crucial. However, generalist platforms are severely limited in this regard, as their training dataset may only run up until a certain point in time (e.g., September 2021 for ChatGPT). Consequently, they cannot provide information on developments that occurred after this cut-off date.

In contrast, Ferma integrates seamlessly with a vast array of trusted and up-to-date resources. This includes over 4 million scientific publications, 190,000 clinical trials, 30,000 SEC reports, 10,000 earnings call transcripts, 105,000 press releases, and much more. What's more, these data sources are updated in real-time, ensuring that Ferma users always have access to the most recent and relevant information. This advantage allows Ferma to stay on the cutting edge of life sciences information, while generalist platforms are left lagging behind.

Conclusion

In the rapidly evolving field of life sciences, Ferma provides a much needed specialized solution. With its superior accuracy, sophistical knowledge graph, and integration with verified and up-to-date sources, Ferma is set to revolutionize the way life sciences professionals access and utilize information. By focusing on the specific needs and challenges of the life sciences industry, Ferma promises to drive advancements in research, drug discovery, and medical breakthroughs.

Experience Ferma firsthand and see how it can transform your work.
Ferma.AI