{"id":12699,"date":"2024-01-04T07:42:52","date_gmt":"2024-01-04T07:42:52","guid":{"rendered":"https:\/\/www.salesforce.com\/?p=12699"},"modified":"2025-08-05T07:13:27","modified_gmt":"2025-08-05T07:13:27","slug":"generative-ai-hallucinations","status":"publish","type":"post","link":"https:\/\/www.salesforce.com\/in\/blog\/generative-ai-hallucinations\/","title":{"rendered":"Is Your Generative AI Making Things Up? 4 Ways To Keep It Honest"},"content":{"rendered":"\n

Generative AI chatbots are helping change the business landscape. But they also have a problem: They frequently present inaccurate information as if it\u2019s correct. Known as \u201cAI hallucinations,\u201d these mistakes occur up to 20% of the time<\/a>.<\/p>\n\n\n\n

\u201cWe know [current generative AI] has a tendency to not always give accurate answers, but it gives the answers incredibly confidently,\u201d said Kathy Baxter, principal architect in Salesforce\u2019s ethical AI practice. \u201cSo it can be difficult for individuals to know if they can trust the answers generative AI is giving them.\u201d <\/p>\n\n\n\n

You might hear those in the computer science community call these inaccuracies confabulations<\/a>. Why? Because they believe the psychological phenomenon of accidentally replacing a gap in your memory with a false story is a more accurate metaphor for generative AI\u2019s habit of making mistakes. Regardless of how you refer to these AI blunders, if you\u2019re using AI at work, you need to be aware of them and have a mitigation plan in place. <\/p>\n\n\n\n

The big trend<\/h3>\n\n\n\n

People have gotten excited (and maybe a little frightened<\/a>, especially when used at work<\/a>) about generative AI and large language models<\/a> (LLMs). And with good reason. LLMs, usually in the form of a chatbot, can help you write better emails and marketing reports, prepare sales projections, and create quick customer service replies, among many other things. <\/p>\n\n\n\n

\n\t
\n\n\t\t
\n\t\t\t

What is an LLM?<\/strong><\/h2>\n\t\t\t\t\t\t\t

Large language models (LLMs) are a type of AI that can generate human-like responses by processing natural-language inputs. LLMs are trained on massive datasets, which gives them a deep understanding of a broad context of information. This allows LLMs to reason, make logical inferences, and draw conclusions.\u00a0<\/p>\n\t\t\t\n\t\t\t\n\t\t\t\t\t\t\t

\n\t\t\t\t\tCheck out our AI glossary<\/a>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\n\t\t
\n\t\t\t\t\t<\/div>\n\t<\/div>\n\n\t\t\t
<\/div>\n\t\n\t\t\t\n\t\t\"\"\n\n\t\t\n\t\t\t\t\t\"\"\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\t\t\t\"\"\n\t\t\n\t<\/div>\n\n\n\n

In these business contexts, AI hallucinations may lead to inaccurate analytics, negative biases, and trust-eroding errors sent directly to your employees or customers. <\/p>\n\n\n\n

\u201c[This] is a trust problem,\u201d said Claire Cheng, senior director, data science and engineering, at Salesforce. \u201cWe want AI to help businesses rather than make the wrong suggestions, recommendations, or actions to negatively impact businesses.\u201d<\/p>\n\n\n\n

It\u2019s complicated<\/h3>\n\n\n\n

Some in the industry see hallucinations more positively. Sam Altman, CEO of ChatGPT creator OpenAI, told Salesforce CEO Marc Benioff<\/a> that the ability to even produce hallucinations shows how AI can innovate. <\/p>\n\n\n\n

\u201cThe fact that these AI systems can come up with new ideas, can be creative, that\u2019s a lot of the power,\u201d Altman said. \u201cYou want them to be creative when you want, and factual when you want, but if you do the naive thing and say, \u2018Never say anything you\u2019re not 100% sure about\u2019 \u2014 you can get a model to do that, but it won\u2019t have the magic people like so much.\u201d<\/p>\n\n\n\n

For now, it appears we can\u2019t completely solve the problem of generative AI hallucinations without eradicating its \u201cmagic.\u201d (In fact, some AI tech leaders predict hallucinations will never really go away<\/a>.) So what\u2019s a well-meaning business to do? If you\u2019re adding LLMs into your daily work, here are four ways you can mitigate generative AI hallucinations.<\/p>\n\n\n\n

\n\t
\n\n\t\t
\n\t\t\t

Train your own LLM<\/h2>\n\t\t\t\t\t\t\t

Thinking of adding generative AI to your business? You don\u2019t need to train your own. A simple API can connect your data to an existing platform.<\/p>\n\t\t\t\n\t\t\t\n\t\t\t\t\t\t\t

\n\t\t\t\t\tFind out how<\/a>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\n\t\t
\n\t\t\t\t\t<\/div>\n\t<\/div>\n\n\t\t\t
<\/div>\n\t\n\t\t\t\n\t\t\"\"\n\n\t\t\n\t\t\t\t\t\"\"\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\t\t\t\"\"\n\t\t\n\t<\/div>\n\n\n\n

1. Use a trusted LLM to help reduce generative AI hallucinations<\/h2>\n\n\n\n

For starters, make every effort to ensure your generative AI platforms are built on a trusted LLM<\/a>. In other words, your LLM needs to provide an environment for data that\u2019s as free of bias and toxicity<\/a> as possible. <\/p>\n\n\n\n

A generic LLM such as ChatGPT can be useful for less-sensitive tasks such as creating article ideas or drafting a generic email, but any information you put into these systems isn\u2019t necessarily protected<\/a>. <\/p>\n\n\n\n

\u201cMany people are starting to look into domain-specific models instead of using generic large language models,\u201d Cheng said. \u201cYou want to look at the trusted source of truth rather than trust the model to give you the response. Do not expect the LLM to be your source of truth because it\u2019s not your knowledge base.\u201d<\/p>\n\n\n\n

When you pull information from your own knowledge base<\/a>, you\u2019ll have relevant answers and information at your fingertips more efficiently. There will also be less risk the AI system will make guesses when it doesn\u2019t know an answer. <\/p>\n\n\n\n

\u201cBusiness leaders really need to think, \u2018What are the sources of truth in my organisation?\u2019\u201d said Khoa Le, vice president of Service Cloud Einstein and bots at Salesforce. \u201cThey might be information about customers or products. They might be knowledge bases that live in Salesforce or elsewhere. Knowing where and having good hygiene around keeping these sources of truth up to date will be super critical.\u201d<\/p>\n\n\n\n