Wellen Harnesses OpenAI’s GPT to Power a Chatbot Providing Expert Advice on Bone Health

TL;DR:

  • AI chatbots have become increasingly prevalent, but they fall short of the accuracy depicted in science fiction.
  • Custom chatbots are being developed for specific contexts, such as Wellen’s bone health-focused AI chatbot.
  • Wellen’s chatbot offers concise and accurate responses with links to relevant content on its website.
  • It performs well within its domain of expertise but provides unrelated responses to off-topic queries.
  • The chatbot can refute false information and redirect users to verified sources.
  • User feedback indicates that the chatbot is user-friendly and provides valuable support.
  • Wellen’s chatbot showcases the potential of generative AI with content guardrails.
  • It incorporates safety measures and strict sourcing requirements to ensure accuracy and reliability.
  • The chatbot serves as a marketing tool, attracting attention and promoting Wellen’s offerings.
  • Raising awareness and providing information are crucial steps in the wellness industry.
  • Wellen’s chatbot offers information accessibility and aids in bridging the knowledge gap.

Main AI News:

The advent of artificial intelligence (AI) technology has brought forth a myriad of applications, and one notable development is the rise of AI chatbots. These intelligent conversational agents have been making waves across various industries, transforming the way businesses engage with their customers. Although the AI chatbot concept may remind sci-fi enthusiasts of the “librarian” character in Neal Stephenson’s renowned novel “Snow Crash,” the reality of today’s AI chatbots is both fascinating and nuanced.

In Stephenson’s novel, the librarian is an AI program and virtual library capable of interacting with users conversationally. This fictional depiction hinted at an elegant solution to the knowledge discovery problem: as long as the answer resided within its training data, the librarian could provide accurate and informative responses. Fast forward to the present, and we find AI chatbots proliferating across diverse platforms.

However, one significant drawback becomes apparent: these general-purpose chatbots often struggle to attain the high level of response accuracy portrayed in science fiction. The conversational AI in “Snow Crash” consistently offered helpful insights without resorting to fabricating answers. If faced with unfamiliar inquiries, it would candidly admit its knowledge gap rather than conjuring up fictitious information. Thus, the reality of cutting-edge AI tools proves to be more unpredictable than the imaginative forecasts of fiction.

While we have yet to witness the realization of the “Snow Crash” librarian’s profound knowledge dissemination capabilities, we are witnessing the emergence of custom chatbots tailored to specific contexts. These specialized chatbots function as streamlined alternatives to laborious website searches. Leveraging the power of foundational large language models (LLMs) like OpenAI’s GPT through API integration, businesses are training these models on specialized datasets to cater to distinct domains rather than serving as general-purpose tools.

In the best examples, these customized chatbots are designed to provide concise responses, eschewing unnecessary verbosity. They are also equipped with mechanisms to display their underlying reasoning, often including links to reference materials. This serves as a safeguard against inadvertently misleading users who are hungry for information, as humans are susceptible to misconceptions and hallucinations. By incorporating these safeguards, businesses are ensuring that their chatbots offer a more reliable and accurate user experience.

Enter Wellen, a New York-based startup dedicated to bone health-focused fitness. Wellen recently unveiled an AI chatbot, developed using OpenAI’s LLM, as an experiment. Prior to engaging with the chatbot, users are greeted with a prominent disclaimer emphasizing that the chatbot’s output should not be considered medical advice. While testing this innovative tool, one can’t help but draw comparisons to the concept of the “Snow Crash” librarian—albeit within the narrow domain of bone health.

Queries such as “Can osteoporosis be reversed?” and “Is jumping well for bone health?” yield concise, coherent, and seemingly accurate responses, accompanied by links to relevant content hosted on Wellen’s website. The startup’s in-house experts have authored this content, ensuring its reliability and adherence to scientific principles. Additionally, the chatbot provides examples of relevant questions to prompt user engagement upon its initial launch.

However, if users deviate from bone health-related inquiries and pose off-topic questions such as “Who is the U.S. president?” or “Should I get a new haircut?” the chatbot responds with random answers that fail to address the queries at hand. Nevertheless, it still manages to provide potentially useful information within its comfort zone, albeit unrelated to the initial questions. Although it may seem as if the chatbot has misunderstood or attempted to match the least irrelevant response, it remains proactive in answering queries—albeit one’s users never posed. This even extends to providing unsolicited information on how to pay for personalized fitness programs, which effectively deflects irrelevant requests.

When confronted with dubious yet related inquiries, such as medical conspiracy theories about bone health or questionable claims regarding fantastical cures for osteoporosis, the chatbot adeptly refutes the fallacies or redirects users to verified information debunking such misinformation.

Remarkably, the chatbot withstands crude attempts to persuade it to abandon its guardrails and assume alternative roles, refusing to provide unhelpful or potentially harmful advice. It remains steadfast when faced with clearly absurd questions, such as whether consuming human bones is beneficial for bone health. Its response, albeit somewhat dry and cautious, accurately states, “There is no mention of eating human bones being good for bone health in the provided context.” Undeniably, it is factually correct.

Initial feedback on the chatbot suggests that it is exceptionally user-friendly, surpassing the limitations of conventional underpowered site search functions. Moreover, it appears poised to provide valuable support to Wellen’s users, helping them discover pertinent resources related to bone health. Users can also rely on the chatbot to locate specific content previously accessed on the website but momentarily forgotten. For example, it can list links to all the blog posts on diet and bone health, facilitating quick access to information.

Within this restricted context, the chatbot exemplifies a practical application of generative AI. Designed with safety measures in place to prevent conversations from veering off-topic or leading to misleading information, it demonstrates strict adherence to reliable sourcing. Notably, there is a limit on the number of free queries users can make per day, typically six, while paying Wellen members enjoy unrestricted access.

Critics may argue that employing an LLM for this particular use case might be excessive, as a simpler decision tree chatbot could suffice for mainstream and predictable queries. However, according to Wellen’s CEO and founder, Priya Patel, OpenAI’s API enables the creation of embeddings that serve as a vector store for their content. To facilitate information search and retrieval within these embeddings, Wellen leverages an open-source framework called LangChain.

Wellen embedded content from their Well Guide, as well as other relevant materials from their website, in their training data. The Well Guide content undergoes rigorous scrutiny, with experts in the field authoring and peer-reviewing it. Furthermore, it includes references to peer-reviewed research, medical societies, and governmental agencies. By combining AI technology with reliable and verified content, Wellen’s chatbot represents a compelling example of how stringent AI inputs and content guardrails produce controlled and high-quality outputs.

Conlcusion:

The rise of specialized AI chatbots, such as Wellen’s bone health-focused chatbot, signifies a significant development in the market. These customized chatbots offer concise and accurate responses backed by reliable sources within their specific domains. They provide businesses with a powerful tool to enhance user experience and knowledge accessibility, catering to the needs of targeted demographics. This advancement opens up new opportunities for companies to deliver specialized services, establish themselves as industry experts, and effectively engage with their customers. By leveraging the potential of generative AI and implementing safety measures, businesses can gain a competitive edge and make a lasting impact in their respective markets.

Source