It seems easiest to use AI as an assistant for existing members

It seems easiest to use AI as an assistant for existing members.

Currently, there seem to be few cases where AI is being used effectively in business, but I believe this is because what AI should be doing has not been clarified or articulated.

For example, if you are introducing a chatbot to interact with users, the clarification and articulation have already been done, and there are track records elsewhere, so the effects of introducing it should be easy to estimate.

Even in such cases, the important thing is a high-quality dataset to feed into the AI, I think.

Especially with chatbots and the like, you should secure many examples of ideal responses.

As for the specific procedure for creating a chatbot dataset, I think it would be around the following:

1. Have humans check the data input into the chatbot and the chatbot's responses to evaluate the validity of the answers. If the response is inappropriate, include the desired response in the dataset. 2. Articulate and organize know-how that has become personalized and include it in the dataset. 3. Keep a record of the basis for the data included in the dataset and the background of how that data was created to improve the quality of PDCA in reviewing the dataset content.

I think that covers it.

Even in this process and task of dataset creation, if you can break down the processes and tasks well and run many PDCA cycles along the way, you can expect significant accuracy improvements without large-scale training.

Furthermore, as for what humans will do after doing those things,

4. Use the free time to proceed with tasks that only humans can currently do.

That is what it should come down to.

#GenerativeAI #Gemini #Grok
#DailyNote #Column #DailyUpdate #note #DailyPost #noteDailyUpdate #Daily
#RecentLearning #DailyUpdateClub #ThoughtColumn #Creative