From deep learning to LLMs to generative AI, we live in a world where AI has the power to transform just about everything, whether it's handling customer support or generating code.

AI wasn't always so pervasive, though.

The first AI tools had significant limitations.

They couldn't understand context or learn and improve on their own.

The first chatbots, for example, were rule-based, meaning they were developed based on predefined rules or scripts.

This severely restricted their capacity to just what was fed into them, the same way any computing software can only perform the tasks that it was programmed to perform.

AI-based chatbots have come a long way since then.

They're leveraging advancements in machine learning and deep learning capabilities to improve the understanding of natural language.

Now with the adoption of LLMs, we're seeing a new evolution.

Large language models use massive amounts of data and a combination of deep learning algorithms, neural networks, and natural language processing techniques to generate human-like responses to queries.

Let's show what we're talking about with Watson X Assistant, a conversational AI platform designed to build and deploy AI-powered chatbots.

With the emergence of generative AI, Watson X Assistant is working to transform user experiences and deliver more intelligent human-like responses.

Today we'll show you how to leverage Neuralseq, which is a search and natural language generation system that we'll integrate with Watson X Assistant.

First, we need to set up Watson X Discovery, which is where our data will be stored.

Let's use Robotic Vacuum Manuals as our example and go into Improve and Customize to test it.

Let's ask, how do I change the filter?

The answer we get isn't great, so let's try asking it another question, how often should I change the mopping pad?

This time it doesn't answer the question we were asking at all.

To improve our answers, let's set up Neuralseq to work with Discovery and Assistant.

Now let's go into the initial setup page to further fine-tune our extension.

We've already pre-filled the initial information here, like what we're talking about, and we've also connected with Watson Discovery.

Then in the About section, it came up with this paragraph on its own.

In tune, we're basically seeing how it should look, as well as whether we should be looking at newer or older documents.

And now in Q&A, this actually looks at all of our documents and generates some questions for us to bootstrap our actions in Watson X Assistant.

So it's saying that based on this data, we think people are going to be asking these questions.

Now click on the Integrate tab.

From here, make note of your API key and download your custom OpenAPI file.

Now open the Integrations page in Watson X Assistant and click Build Custom Extension.

Follow the steps to import your OpenAPI file and click Finish.

Now scroll down and click Add on your Neuralseq tile to add the extension.

Follow the steps to set up the authentication for Neuralseq.

For the authentication type, be sure to set it to API Key Auth and paste your Neuralseq API key from earlier in the indicated field.

Click through to finish.

Now that we've set up our Neuralseq extension, let's go ahead and add it to our dialog.

First create a new action skill.

Next we're going to create an action and quick start with a template before adding our Neuralseq starter kit, which we just generated.

Here you can see we have our Neuralseq search action.

We need to configure it.

So you can see that it's generated all this information for us.

We need to edit our extension, select the extension we just created, and for the operation we would like to seek an answer from Neuralseq.

For parameters, we'll set question equal to query underscore text.

Click Apply and then save the action.

Now we want to go to no action matches, which means that if we can't match any of the phrases to anything we've set up in Watson X Assistant, it's going to go out to Neuralseq and get that answer for us from Discovery.

Let's remove the pre-filled text.

So this is basically like the anything else node.

Now let's go to a sub action and go to Neuralseq search before hitting apply, then save, then preview.

Let's ask Watson X Assistant, how do I change the filter?

It's going out to Neuralseq and here you can see the response that it's generated.

To change the filter, open the filter door, remove the filter by grasping the tab, shake off debris, reinsert the filter, and close the filter door.

Replace the filter every two months.

This is a really good and accurate response.

Now let's ask our other question.

How often should I change the mopping pad?

It's going back to Neuralseq again and gives us the correct response.

As you can see, our Neuralseq extension can help your chatbots carry out conversations just as well as any human could, and that's due to its generative AI capabilities.

To learn more about how you can leverage generative AI with Watson X Assistant, please visit the Watson X Assistant page on ibm.com.

Thank you.

If you have questions, please drop us a line below.

If you want to see more videos like this in the future, please like and subscribe.