Machine Learning Made Easy: Q&A with Snowflake Head of Artificial Intelligence and Machine Learning Strategy Ahmad Khan

Why AI has everyone’s attention, what it means for different data roles, and how Alteryx and Snowflake are bringing AI to data use cases

There’s a llama on the loose! Well, more specifically, LLaMA (Large Language Model Meta AI), along with other large language models (LLMs) that have suddenly become more open and accessible for everyday applications. With all the hoopla around AI, there’s a lot to get up to speed on—especially the implications this technology has for data analytics.

And who better to chat about ChatGPT with than Ahmad Khan, Head of artificial intelligence (AI) machine learning (ML) strategy at Snowflake? Ahmad has not only witnessed the emergence of AI in the data space but helped shape it throughout his career. At Snowflake, Ahmad works with customers to solve AI use cases and helps define the product strategy and vision for AI innovation, which includes key technology partners like Alteryx. 

I interviewed Ahmad for the Alteryx Alter Everything podcast, and we had a great time talking about the past innovations that led up to the AI boom we’re seeing today, and what that means for data analytics.

Some takeaways?

  • AI will open doors for people in the same way the internet did. And potentially with an even bigger impact.
  • AI unlocks new data use cases. With the ability to handle unstructured data types and larger volumes of data, AI gives us the tools to tackle more complex, exciting problems.
  • Alteryx and Snowflake give you a place to start taking advantage of AI. You have everything you need when you combine the two: a centralized, governed repository for data; scalable compute power; quality data sets and APIs; and an AI-powered, user-friendly UI that lets non-technical employees start creating value fast.

Read on for the interview highlights and listen to the podcast episode for the full conversation.

On how we made it to the “iPhone moment” for AI

Alex Gnibus: You really have been hands-on with AI, actively building and shaping this technology and making it available over the past several years. You’ve witnessed firsthand the evolution of AI, and it feels like things have really snowballed recently.

Between ChatGPT and all the other technologies popping up, it feels like the barrier to entry for experimenting with AI really just dropped out of nowhere. People are calling it the iPhone moment for AI. So, I wanted to talk through what just happened in the past few years and even months … or this week!

Ahmad Khan: You’re absolutely right. And I can tell you, this has been slowly evolving. At the end of 2022 and early 2023 is when Generative AI and LLMs got to a point where it captured the public’s imagination. But really, with the innovations that have been happening, you can go probably all the way back to the early 2010s.

And even before that, there were multiple waves that joined in to create the tsunami. And there, it’s the proliferation of cloud computing—being able to store large amounts of data, access to large clusters of compute, usage of specialized hardware such as GPUs that enables faster processing for the math behind machine learning (ML). 

The key paper that drove this moment came out in 2017 was titled Attention Is All You Need. And that was the basis of this new architecture of how neural networks are arranged. And that led to next-level performance on natural language processing. That’s where you see the emergence of ChatGPT, right after the paper came out. 

And so it’s been going on for a while. It’s just that the public at large was not exposed to the research or the academic side of it. And I would definitely agree that, in my mind at least, this will have a big, big impact on our world, perhaps even bigger than the internet.

On the new possibilities for data use cases

Alex: Suddenly it’s now much easier to start working with AI, solving your own problems and experimenting without needing access to the resources that you normally would’ve needed access to.

For Alteryx customers and Snowflake customers who are thinking about this for data analytics, what does this mean for people who want to start experimenting with AI for data use cases?

Ahmad: Yeah, I mean, the possibilities are really endless, right? We’re at the beginning of this, and when I used to talk to customers about classical ML, like regression use cases, and essentially what you’re doing is looking at historical data and then you’re trying to predict something about the future.

But when you take deep learning, ML, AI and LLMs, this is really enabling a newer kind of insights from a newer kind of data set, and data that we’ve never even looked at. 

I was looking at some statistic that at any typical company, more than 80% of the data is unstructured. We’ve been doing analytics on structured data only. And so it’s not just about doing some kind of sales forecast anymore. Those are very important things to businesses. But now this enables a newer kind of insights from all this unstructured data that has been untapped so far.

Think about how you have maybe a corpus of support call recordings, and you unleash an AI agent on those call recordings and transcribe them to your primary language, maybe English or French. And then you unleash another AI agent on that to get the sentiment out of it. Then you’re building a sort of a dashboard in Alteryx and saying, “Hey, give me the customer sentiment, day by day.”

On how anyone can leverage AI

Ahmad: What this has done is taken ML and AI from academia and created a commercial stack. An everyday programmer can go in and create amazing applications very quickly. 

And a non-technical person can do this as well. Not only can you generate code and build a lot of stuff, but what I foresee is features and functionality in all sorts of tools, including Alteryx, where you would be able to give a prompt to the tool and it will do things for you. And not only that, but the tool itself will maybe inject a prompt and use an LLM in the background to generate something for you. 

These are newer patterns that we are seeing in this space. So, I would say the possibilities are endless. If you’re somebody technical, it makes you ten times more productive, a hundred times more productive.

And if you’re somebody who’s non-technical, it also makes you ten times more productive, right? You could be a solo entrepreneur and just get started very quickly because you can build all this stuff using ChatGPT and other tools.

On the combination of Snowflake and Alteryx for AI

Alex: And it goes back to what we like to talk about with Snowflake and Alteryx, where we support workers who—even if they don’t know a coding language or have a specific skill—can start getting upskilled on best practices for knowing how to interface with these new technologies. What are some ways Snowflake and Alteryx are helping businesses develop AI capabilities?

Ahmad: For our customers, Snowflake is the main repository of data. So think of Snowflake as the central place where you can bring all sorts of data. What we’ve done over the years is enable structured and unstructured data. The key piece there is you have to be able to govern that data. Your data is very, very valuable, and you have to have a place where you keep the data governed and give the right access to the right people.

And maybe I want to transform and process in that same environment. Snowflake supports the data analysts that are writing SQL, the data engineers that are writing stuff in Python, and the data scientists that are writing in Python.

And I think where Alteryx really complements Snowflake is the UI component, where there are personas such as business analysts and data scientists who like to go and experiment very, very quickly with the visual environments as well.

Alex: So if you’re able to get your hands on the data you need, you can get started on building. You can prepare a training data set and automate the workflow using Alteryx Designer or Designer Cloud. You can put that data set in Alteryx Machine Learning. You can combine it with data from Snowflake Marketplace.

There’s so much connectivity between the two platforms that I think it’s the perfect tee-up for being able to go in and just start working with AI yourself in many ways.

Ahmad: Absolutely. I think in order to learn all this new stuff and play with it, the two tools are going to be super beneficial. 

And the other thing I like about Alteryx is it gives you programmatic access as well. So for any of the technical personas, they can go in and code this stuff. And non-technical personas can use Alteryx Designer and use its ML capabilities both in Alteryx and then in Snowflake. Snowflake is also making it easy for our customers to integrate with external APIs, such as GPT4 and other commercial vendors. So that’s something I’m really excited about.

Alex: To wrap up here, what would you suggest for someone who is super new to AI and doesn’t know how to code? How can they get started?

Ahmad: There are simple use cases. The number one is definitely forecasting. If I were someone getting started, I’d start with the most simple applied ML. What is being applied in business today? It’s stuff like simple time series forecasting, regression, classification. And when I talk to customers, a lot are still in early stages of operationalizing even the classical ML use cases, like being able to better predict customer churn. 

Look at these use cases being applied in business today, and that builds a solid foundation and gives you insight on how to apply the latest AI techniques. People are figuring this stuff out. It’s still super early. 

Alex: Very early days. Which means it’s the perfect time to get started.

Note: This interview has been edited for length and clarity.

Additional resource

The post Machine Learning Made Easy: Q&A with Snowflake Head of Artificial Intelligence and Machine Learning Strategy Ahmad Khan appeared first on Snowflake.

LATEST ARTICLE

See Our Latest

Blog Posts

admin June 12th, 2024

Bringing machine learning (ML) models into production is often hindered by fragmented MLOps processes that are difficult to scale with […]

admin June 12th, 2024

Discovering and surfacing telemetry traditionally can be a tedious and challenging process, especially when it comes to pinpointing specific issues […]

admin June 12th, 2024

Today’s data-driven world requires an agile approach. Modern data teams are constantly under pressure to deliver innovative solutions faster than […]