Search engines are not new. In fact, they've been around for decades. With technological advances, simple keyword search tools have evolved into intricate algorithms that incorporate machine learning to understand natural language and context to offer more precise and relevant search outcomes. Despite significant advancements, though, searching remains a time consuming task - whether it's utilising a tool like Google or navigating an internal inventory for a store; and searching becomes even more difficult when dealing with intricate knowledge bases. Enter assisted search.
AI assisted search is the natural evolution of search, taking it one step further by eliminating the need for manual page visits to find information. Instead, it uses AI to sift through mountains of data to present the most relevant and accurate answer. A great use case for this would be in customer service, where quick access to information can help customers find what they need without having to call a support center - a win for those customers who prefer to avoid human interaction.
Even if they do need to call the support center for a more complex issue such as technical support, the AI Assisted Search can help support staff quickly find the relevant internal documentation to help the customer, providing better customer service and the ability to serve more customers.
Likewise, AI search can help other staff who are frequently confronted with the intimidating task of navigating through disorganized and voluminous internal company documents to locate required information as part of their role (knowledge workers). For instance, medical staff can leverage assisted search to gain quick and easy access to the latest treatment guidelines or procedures, which can result in better patient outcomes; Diabetes & Metabolic Syndrome: Clinical Research & Reviews, Volume 14, Issue 4, suggesting that AI is ‘used for proper screening, analyzing, prediction and tracking of current patients and likely future patients’ during the Covid-19 pandemic. Given the constantly evolving nature of medical research and best practices, assisted search ensures that medical professionals stay informed and up-to-date, thereby enabling them to deliver high-quality patient care.
Similarly, legal firms can benefit from assisted search technology by facilitating the retrieval of relevant case law. Lawyers can save considerable time and effort by employing assisted search to identify the most pertinent legal precedents, which can aid them in making more informed decisions on behalf of their clients. This allows legal professionals to dedicate more time to building compelling arguments and developing winning strategies, thereby enhancing their overall effectiveness and success.
While assisted search technology can offer numerous benefits, it's crucial to recognise that there are potential risks associated with its use. The most significant concern is the possibility of AI-generated answers being inaccurate, leading to incorrect decision-making or actions. This is particularly concerning for our healthcare workers and lawyers alike, who deal with situations that can have significant implications for human lives.
To mitigate this risk, a large language model (LLM) that has been trained on high-quality, up-to-date data and well tested and tuned should be used. An example of this is OpenAI’s GPT 3.5 model, which had over 200 million users in the first 2 months, which gave OpenAI plenty of opportunities to refine the output. At the time of writing, no other model has had this level of user testing and fine-tuning from human feedback. Furthermore, if the AI generates an incorrect answer, there should be mechanisms in place to help reduce the likelihood of it shown to the user; NVIDIA has recently released an open source toolkit called NeMo Guardrails to control the output of LLMs to minimize hallucinations and factual inaccuracies.
Another way to mitigate the risks is by clearly informing the user that the answer is AI-generated and a disclaimer that it may be inaccurate. Furthermore, links to the original source of information can be provided as footnotes, which provide a mechanism for users to quickly fact check the answer (which is what Bing uses). This helps ensure that users make informed decisions based on reliable information and minimise the potential of harm from inaccuracies.
Despite these risks, assisted search is a promising technology that will only get better over time as AI improves. With the ability to quickly and accurately search through large amounts of data, assisted search can save valuable time and effort in customer service, support, and knowledge work. As with any new technology, it's important to be aware of the risks, but the benefits of assisted search make it a technology worth considering.