The 5 Why’s which the CXOs asking on Generative AI?

Last week in my weekly technology article - I have spoken on - Open Door Discussion with CXOs on Generative AI – What should we Focus ?? I am sharing the link again in case people have missed reading it so far.

https://www.dhirubhai.net/pulse/open-door-discussion-cxos-generative-ai-what-should-kingshuk-biswas-h3irc

This week I will tell you - The 5 Why’s which the CXOs are asking on Generative AI and I will try to Put the AI into Action…?

• Where does the organization send their data and who can see the data actually? According to a CEO study, 61% are concerned about data lineage.
• What about Data Security and Data Privacy? According to CEO study, 57% are concerned about security.
• Will the model providers use the data to train their own models? This is a very big concern of many large enterprise organizations.
• What is the cost for running these models to calculate the business ROI? This is again very big concern of many large enterprise organizations.
• How to Reduce Large Scale Deployment Costs of Gen AI?

So, when we are going to any open-door discussions with CXOs and Customer Leadership teams, be well prepared with these 5 Why’s. They are super important.

You can refer my published articles on Generative AI in LinkedIn for addressing all the above questions nicely.

https://www.dhirubhai.net/pulse/generative-ai-foundation-models-value-creator-kingshuk-biswas - How can Enterprise become Value creator in Gen AI and embark on their Gen AI journey.

https://www.dhirubhai.net/pulse/generative-ai-fms-llms-how-reduce-cost-large-scale-kingshuk-biswas - How to reduce cost in large scale deployment of Gen AI and Foundation Models

Let’s Put the AI into Action…

How to Define Generative AI Use Cases

Hey, I have got an interesting use case for, then it is worth asking these questions :

• Does it matter if the output is correct, or if any of risks trigger?
• In lot of cases, being correct really does matter, and in that case, you need to make sense that we can detect, or we can tell the difference between good quality information and bad information.

Language Models are not Knowledge Models

Probably this is the one message I'd like you to know in Gen AI transformative space.

In the case of a language model, they're very good at understanding which words normally come before or after which other words, when talking about a particular topic, but they don't really understand what those topics are that they're talking about.

The words that are used are just symbols and are just tokens that they're manipulating. And there is no real or deep understanding of the knowledge and the concepts that sit behind that. I'm not saying they're not useful. They are very much helpful, and they are going to be transformative.

Let me give an example to demonstrate this :

To me, medical diagnosis, it's a not a good use case. We shouldn't be using generative AI for medical diagnosis. If you took a medically trained chat GPT and put it in the hands of patients, that would not be a good idea because the patients can't tell if the advice, they're getting from that chatbot is correct or not. However, if you take that same chat bot and use it to augment the abilities of a human doctor to use it to make suggestions to that doctor about diagnosis, they might not have considered which they can then decide whether to accept and use or to filter out feels much more acceptable because the person consuming the data or the person consuming the output of the model is already expert enough to tell the difference between good information and bad.

How to Implement Generative AI Solution

Next week, It will be even more interesting, since I will be covering the following topics :

• Steps to build a Generative AI solution
• How to train Large Language Models
• Generative AI Framework

#HappyWeekendAndStayBlessed