ROCK AI, Responsible Owner Curated Knowledge
WYSIWYG is still most common user interface, but language-based user experience (chat) is finding momentum.
Some users are more comfortable chatting than finding the information through a graphical user interface(aka GUI), the new chat-based UX can be called a CUI(conversational user interface)
LLM (Large Language Model) is making chats an enjoyable experience, chat is a new trend in user interface and will remain..
AI is not only a chat, but also an assitant, both to the owner and the consumer.
There is market awareness in the public about this shift to Conversational UI.
There are also cool AI services such as image tags, scheduleing, etc. that have market
LLM systems are language-based, they are not made for accuracy but to simulate human language, they have a tendency to hallucinate. .
The machine learning of some AI is based on large amount of data, which may include false or imaginative information.
The chatbots do not take responsibility, the fine prints of all of them is that the response can be inaccurate. How can a business owner rely on an CR who does not take responsiblity for what they say?
When AI learns, it does not unlearn. If a seat on a flight or a house is sold, AI must know about it immediately, and remove it from its knowledge. And if an employee, specially a key position, leaves the company, the AI must immediately unlearn the previous knowledge about them.
The AI who acts as a sales rep to a brand, must not recommend merchandise from the competing brands, even if they are better. There is no guarantee that an AI will not recommend a competitor.
The company's data as well as the consumer's data, can be classfied. AI must be aware what data is classfied and what is not.
This solution, promoted by Microsoft among others, adds business specific data to the machine learning process. It helps to add truthfulness to data. But it does not guarantee it.
Involves having a human checking the data generated by ML. It is not feasible for smaller businesses and it is somewhat reductive, also it still does not guarantee perfect results.
It is possible to ask the ML system to only display the results on a certain website or in a certain set of files. This can improve the truthfulness, but may also produce unexpected results.
