{"id":6348,"date":"2024-02-05T12:46:24","date_gmt":"2024-02-05T12:46:24","guid":{"rendered":"https:\/\/tup.kxe.temporary.site\/how-to-protect-against-and-benefit-from-generative-ai-hallucinations\/"},"modified":"2024-02-05T12:46:24","modified_gmt":"2024-02-05T12:46:24","slug":"how-to-protect-against-and-benefit-from-generative-ai-hallucinations","status":"publish","type":"post","link":"https:\/\/okdesign.ca\/en\/how-to-protect-against-and-benefit-from-generative-ai-hallucinations\/","title":{"rendered":"How to protect against and benefit from generative AI hallucinations"},"content":{"rendered":"


\n<\/p>\n

\n
\n
\"Optimove\"Optimove<\/figure>\n<\/div>\n

As marketers start using ChatGPT, Google\u2019s Bard, Microsoft\u2019s Bing Chat, Meta AI or their own large language models (LLM), they must concern themselves with \u201challucinations\u201d and how to prevent them.<\/p>\n

IBM<\/a> provides the following definition for hallucinations: \u201cAI hallucination is a phenomenon wherein a large language model\u2014often a generative AI chatbot or computer vision tool\u2014perceives patterns or objects that are nonexistent or imperceptible to human observers, creating outputs that are nonsensical or altogether inaccurate.<\/p>\n

<\/p>\n

\u201cGenerally, if a user makes a request of a generative AI tool, they desire an output that appropriately addresses the prompt (i.e., a correct answer to a question). However, sometimes AI algorithms produce outputs that are not based on training data, are incorrectly decoded by the transformer, or do not follow any identifiable pattern. In other words, it \u2018hallucinates\u2019 the response.\u201d<\/p>\n

Suresh Venkatasubramanian, a professor at Brown University who helped co-author the White House\u2019s Blueprint for an AI Bill of Rights, said in a CNN blog post<\/a> that the problem is that LLMs are simply trained to \u201cproduce a plausible-sounding answer\u201d to user prompts.<\/p>\n

\u201cSo, in that sense, any plausible-sounding answer, whether accurate or factual or made up or not, is a reasonable answer, and that\u2019s what it produces. There is no knowledge of truth there.\u201d<\/p>\n

He said that a better behavioral analogy than hallucinating or lying, which carries connotations of something being wrong or having ill-intent, would be comparing these computer outputs to how his young son would tell stories at age four.<\/p>\n

\u201cYou only have to say, \u2018And then what happened?\u2019 and he would just continue producing more stories,\u201d Venkatasubramanian added. \u201cAnd he would just go on and on.\u201d<\/p>\n

Frequency of hallucinations<\/h2>\n

If hallucinations were \u201cblack swan\u201d events \u2013 rarely occurring \u2013 they would be something marketers should be aware of but not necessarily pay much attention to.<\/p>\n

However, according to studies from Vectara, chatbots fabricate details in at least 3% of interactions \u2013 and as much as 27%, despite measures taken to avoid such occurrences.<\/p>\n

\u201cWe gave the system 10 to 20 facts and asked for a summary of those facts,\u201d Amr Awadallah, Vectara\u2019s chief executive and a former Google executive, said in an Investis Digital blog post<\/a>. \u00a0\u201cIt is a fundamental problem that the system can still introduce errors.\u201d<\/p>\n

According to the researchers, hallucination rates may be higher when chatbots perform other tasks (beyond mere summarization).<\/p>\n

What marketers should do<\/h2>\n

Despite the potential challenges posed by hallucinations, generative AI offers plenty of advantages. To reduce the possibility of hallucinations, we recommend:<\/p>\n