Generative AI in Everyday Life

Guest -Karthick:

Sure. Thanks for hosting this and thanks for inviting me here. The new generation of AI, generative AI, is massively different, completely different from traditional AI. I’ll give you a human-like example and then probably a technological one. Generative AI essentially means that it has the capability of generating new content, be it text, video, audio, or anything. The most popular tools are like ChatGPT. If you ask a question like, “What is the capital of India?”, it is based upon data to train this model. From the trained data, it will be able to get the next best or next best available result. So obviously, the highest probability answer will be Delhi. In this way, for each and every syntax, it is able to generate a composition, which was not possible in the earlier generation of AI. It’s able to generate new content.

Use Cases of Generative AI

Neha: Great. Okay. That’s how it is. I did not know it. I am sure our audience must not be knowing it. That’s great insight. Since you have mentioned ChatGPT and all that, are there examples of the use cases of generative AI in everyday life?

Karthick: Use cases have always been there, like every time you use a credit card, AI is used to advise of potential fraud. But now the usage of AI technology is much more commonly available to every part of the world. Next time you go to a company website or your banking website and you are chatting and asking a specific question, it may not be a real person; an AI agent is answering you. The new generation AI is capable of generating answers for almost every question. You are able to get human-like answers or responses from those machines. This is very common whether you are booking a ticket in the Indian IRCTC ticketing system or you go to any bank.

Generative AI and Chatbots

Neha: So Karthick, I thought chatbots are just like, you know, computerized kind of thing. Now, since we know that chatbots are mostly using generative AI, can you tell us how actually generative AI technologies are applied to chatbots?

Karthick: ChatGPT is essentially a chatbot application. It’s a wrapper on top of generative AI technology. Generative AI is the new generation of AI that is able to generate new content. When you ask a question, it is passed on in the back end to a language model, in this case a GPT model created by OpenAI. The model understands the question, reasons it out, and gives you back an answer.

The Role of Robots

Neha: Many people have this thought: Tesla has created a robot, and people are using it in their daily lives, even making them do their normal household chores. Since they are interacting with us like a friend or like a house help, do you think robots can rule our lives at certain points?

Karthick: I will not make a very bold statement and say robots are going to rule our lives. But there will definitely be a proliferation of robots. The Tesla robots are designed for doing everyday chores. But AI technologies are mostly going to be driven by software. For example, they are definitely going to play a huge role in code generation. Even scenarios like where documents are going to be read and understood, AI is going to play a huge role. These AI agents or AI robots are not visible to you; they are going to do a lot of jobs at the back end. There are already a lot of physical robots. Military-based robots are going to places that are probably not reachable by humans.

Risks of Relying on AI

Neha: What will be the risks of relying on AI in personal decision-making?

Karthick: AI is still literally just a year or two old at this moment. I would not give too much importance and still be conscious of whatever answer AI is going to give back to you. Whenever you ask a question to ChatGPT, don’t take it at face value. Go ahead and make sure that it is correct because all these large language models have updates. The information is only as good as the information that it’s been trained on, and it’s only till a particular cutoff date. The risks are that AI technology is still evolving. There are terms like “hallucinations,” which means it can confidently give you a wrong answer as though it’s the right answer. So whenever you get an answer from an AI bot, make sure that you double-check those answers. There are also a lot of risks regarding data privacy.

Neha: Do the AI models we use daily have data privacy implications, or should we use more offline AI models?

Karthick: For example, ChatGPT’s terms and conditions say that answers they give will be part of their research. But if you are taking a paid subscription with them, they will not use your answers and your data for any other purpose. There is definitely a risk involved for using free subscription of ChatGPT. When using a free subscription, be mindful of the fact that you are not asking any private questions about yourself or giving any private data about yourself.

Neha: Please do not ask any private questions while you are using free versions of ChatGPT or any other AI models because that can be used to train the model. If you want to ask certain kinds of private questions, maybe use a paid version.

How to Interact with Generative AI Tools

Neha: Many people face problems with how to interact with generative AI tools to get the desired answers.

Karthick: We are conditioned to a specific way of doing things. For the last 10-20 years, we have been slaves to searching in Google in a specific way. There is a huge difference between searching in Google and how you ask questions to generative AI. When you are doing searches, you typically don’t ask questions. You just type keywords. The first suggestion I am going to give you is to talk to that particular AI as you are talking to a human. Don’t consider it as a machine. Be very specific and clear about what you want to achieve. Instead of asking, “Tell me about history,” tell me about the history of a specific empire. The more specificity, the more relevant answers you get. Provide a context on why you are asking this question. Instead of blindly asking, “Give me details about climate change,” provide a context, saying, “Consider yourself as an expert in climate change. I am writing a research paper on climate change. Can you please provide me data on global temperature during the 21st century?”. Break down complex requests. For instance, first explain the causes of World War One, and then ask about the key events that led to its conclusion. Research papers come in a specific format. You can also specify the format in which you need the answer. You can say, “Can you give me the list of the top five largest economies in the world this year?” The word “list” is important because now it is going to give you the answer as a list. Keep hydrating and refining your prompt.

Neha: To keep in mind: You are not Googling. Ask in a way you are interacting with a human. Give the AI a context.

Karthick: The input that you are giving is called prompting. In the initial stages, it may not be that easy. Slowly get rid of using keywords.

Neha: We are still evolving and getting experts at asking questions from AI.

Conclusion

Neha: Before we wrap up, would you like to share any additional information?

Karthick: When you initially start using generative AI technologies, the input that you are giving is called prompting. In the initial stages, it may not be that easy. Once you get more comfortable, you will start asking the AI questions instead of your wife. Take the answers with a pinch of salt and make sure to verify the answers. The AI models are evolving every week.

Neha: Thank you so much, Karthick.

Karthick: Thank you for having me here.