Creating Trustworthy AI by Learning the Lessons of Search Engines

Generative AI holds huge promise for businesses, in the shape of measurable improvements in efficiency, productivity and customer service. Yet, there is one potential drawback that cannot be ignored: trust. Business leaders have been grappling with the issue of building generative AI applications that offer accurate answers and avoid ‘hallucination’ or generating otherwise false or inaccurate responses. To deal with this issue, it pays to look at how we deal with problems in an earlier transformative technology: Search. 

Search engines hold important lessons in how to build trustworthy generative AI applications, both in terms of what they do well and what they fail to do. Enterprises are increasingly using generative AI in their day-to-day operations but the level of accuracy varies. For example, if a company is using AI to build a programme that decides which adverts to display on a web page, a rough level of accuracy is perfectly adequate. However, if AI is powering a chatbot that answers vital financial questions, like how much an invoice is worth or how many days off an employee has this month, there’s no room for error. 

For decades, search engines have aimed to sift through huge troves of online data and deliver accurate answers, offering important lessons in how to surface the right data. By mixing the successful aspect of search with new approaches for generative AI in business, business leaders can unlock the power of generative AI in the workplace and solve the “trust” problem.

Related:AI's Trillion-Dollar Opportunity: Streamlining Enterprise Data Integration

Sifting for Gold

One area where search engines perform well is sifting through large volumes of information and identifying the highest-quality sources. For example, by looking at the number and quality of links to a web page, search engines return the web pages that are most likely to be trustworthy. Search engines also favour domains that are known to be trustworthy, such as government websites or established news sources.

In business, generative AI apps can emulate these ranking techniques to return reliable results. They should favour the sources of company data that have been most frequently accessed, searched or shared. And they should strongly favour sources that are known to be trustworthy, such as corporate training manuals or a human resources database, while deprioritising less reliable sources.

Identifying the Truth

Many foundational large language models (LLMs) have been trained on the wider Internet, which as we all know contains both reliable and unreliable information. This means that they’re able to address questions on a wide variety of topics but they have yet to develop the more mature, sophisticated ranking methods that search engines use to refine their results. That’s one reason why many reputable LLMs can hallucinate and provide incorrect answers.

Related:Revolutionizing Health Care Delivery With Artificial Intelligence

One of the learnings here is that developers should think of LLMs as a language interlocutor, rather than a source of truth. In other words, LLMs are strong at understanding language and formulating responses but they should not be used as a canonical source of knowledge. To address this problem, many businesses train their LLMs on their own corporate data and on vetted third-party data sets, minimising the presence of bad data. By adopting the ranking techniques of search engines and favouring high-quality data sources, AI-powered applications for businesses become far more reliable.

Knowing Your Limits

Search has become quite accomplished at understanding context to resolve ambiguous queries. For example, a search term like “swift” can have multiple meanings – the author, the programming language, the banking system, the pop sensation and so on. Search engines look at factors like geographic location and other terms in the search query to determine the user’s intent and provide the most relevant answer.

However, when a search engine can’t provide the right answer, because it lacks sufficient context or a page with the answer doesn’t exist, it will try to do so anyway. For example, if you ask a search engine, “What will the economy be like 100 years from now?” or “How will Manchester United perform next season?” there may be no reliable answer available. But search engines are based on a philosophy that they should provide an answer in almost all cases, even if they lack a high degree of confidence.

This is unacceptable for many business use cases and so generative AI applications need a layer between the search or prompt, interface and the LLM that studies the possible contexts and determines if it can provide an accurate answer or not. If this layer finds that it cannot provide the answer with a high degree of confidence, it needs to disclose this to the user. This greatly reduces the likelihood of a wrong answer, helps to build trust with the user and can provide them with an option to provide additional context so that the generative AI app can produce a confident result.

Outside the Black Box

Explainability is another weak area for search engines but one that generative AI apps must employ to build greater trust. Just as secondary school teachers tell their students to show their work and cite sources, generative AI applications must do the same. By disclosing the sources of information, users can see where information came from and why they should trust it. Some of the public LLMs have started to provide this transparency and it should be a foundational element of generative AI-powered tools used in business.

Healthy Scepticism

Make no mistake, it is still challenging to create AI applications that make minimal mistakes. But the benefits are so clear and measurable that business leaders cannot afford to ignore the idea. The key thing is to approach AI tools with open eyes, in the same way that the internet has taught us all to be sceptical about news and information sources, so too must business leaders develop a healthy scepticism around trustworthy AI. Key to this is to always demand transparency from AI applications, seeking explainability at every stage and remaining conscious of the ever-present risk of bias. 

Applications built in this way hold out the promise of transforming the world of work entirely. To deliver on this promise, they must be built to be reliable and trustworthy first and prioritise accuracy above all. Search engine technology is designed with different use cases in mind but has useful lessons in how to surface accurate responses from bewildering amounts of data. By taking this knowledge and adding new techniques to build on this accuracy, business leaders can build generative AI apps that deliver enormous potential.