01.02.2024 | News IntraFind debunks five common myths about generative AI

When it comes to technology, no other topic has been driving the world as much as generative AI since 2022. In particular, Open AI's ChatGPT, a chatbot based on the Large Language Model GPT (Generative Pretrained Transformer), is on everyone's lips. As a result, some misconceptions have spread, which IntraFind will clear up below:

Myth #1: There is no way around ChatGPT

The "conversational chatbot" ChatGPT is by far the most popular tool when it comes to text generation using generative AI. This is not without reason: The tool, which is based on the Large Language Model (LLM) GPT, is extremely powerful, especially in its latest version. With all the hype surrounding this chatbot, the competition is of course losing focus. However, it exists and is growing day by day. Google’s Bard, for example, as well passed the Turing test and is therefore indistinguishable from a real person in a written conversation. There are also European alternatives such as Luminous from the German provider Aleph Alpha and some very good open source models.

Myth #2: Companies cannot use generative AI in compliance with data protection regulations

Generative AI becomes stronger the more data is available to it. Companies can certainly use this technology as long as they observe a few guidelines and create the necessary conditions. To this end, it is advisable to only use the relevant tools on-premises, i.e. on your own IT infrastructure. By integrating generative AI into an enterprise search engine that takes into account the read and access rights of employees, companies prevent that sensitive information reaches unauthorized persons. Thus, there is no objection to use it under data protection law. The fact that every employee only sees the data that they are authorized to see must not be ignored when using generative AI. Enterprise search products that serve as an access guardian to authorized information are the prerequisite for this.

Myth #3: ChatGPT is perfectly adequate for finding corporate content

Unfortunately, this is not true. ChatGPT is undoubtedly useful, but if the underlying LLM has not been fed with company information or does not have access to it, even the most intuitive and powerful AI cannot find anything. It also does not make sense to use it without a corresponding enterprise search solution, as ChatGPT on its own delivers unfiltered information to potentially unauthorized persons. To obtain an excellent search result with valid content, an enterprise search that incorporates the organization's own knowledge in a rights-checked manner is required in addition to generative AI. Let's assume that a new document with new content first has to be "introduced" into an existing LLM as part of a training course - the constant retraining of a model is simply far too expensive and takes far too long. "Conventional" search is and remains a valid function that cannot be replaced by an LLM.

Myth #4: We have no use for chatbots in our company

ChatGPT is a chatbot, that's true so far. However, generative AI can do much more than just hold a deceptively real conversation or respond to questions. There are many different use cases for this future technology. For example, it can summarize information from different and very long documents. It can also be used as a digital assistant in first line customer support or to take over time-consuming routine tasks, such as extracting data from files.

Myth #5: The main thing is that we have an AI tool - we can decide later for which purpose

It is understandable that companies do not want to waste time when using generative AI and are currently trying out a proliferation of innovative ideas: Installing AI tools without a concrete task and objective is not advisable. Companies should first define the problem to be solved with an artificial intelligence tool and then work with experts in the field to evaluate which tool is suitable for the specific use case. As part of the planning process, they should clarify which AI model is best suited to which scenario. This also includes the question of which framework conditions the tool must fulfill to meet the guidelines of the company or authority with regard to hardware, data storage and data protection. 

"Generative AI can take over tedious routine tasks and make work in companies and public authorities much more efficient," explains Franz Kögl, CEO of IntraFind. "Artificial intelligence with access to company data can be a powerful ally, as long as clear limits are set: Companies must apply the highest security standards to protect their sensitive data when using generative AI - an enterprise search solution helps to enforce them."