Intel's Wei Li on low-code/no-code AI and sentience

Wei Li, Intel's vice president and general manager of AI and analytics, says he's on a mission to make AI more accessible to developers.

Intel, the world's largest manufacturer of CPUs and semiconductors, is well known for its chips, though it has struggled to stay at the top of the chip market. But hardware isn't Intel's only focus, according to Li. His team, the AI software group, is dedicated to bringing low-code/no-code AI software to the masses.

One of the ways Li said Intel hopes to make AI more accessible is with AI reference kits, which include a chatbot and a document analyzer. The kits, which support Python and C++, contain pre-written code, training data, libraries, machine learning pipeline instructions and Intel oneAPI components for compatibility across different processors and accelerators.

In this Q&A, Li talks about the utility of reference kits, a low-code/no-code future and the sentient AI debate.

How do reference kits help developers to introduce AI into workflows?

Wei Li: They're designed for people who may want to develop [for example] a chatbot and they don't know how to start. Most people already have some kind of interface, such as a GUI interface, and some basic infrastructure. Then they say, 'Hey, how can I get a conversation engine from this thing?' The chatbot reference kit gives people a starting point to create a basic chatbot model.

Wei Li

It used to take weeks to develop a model. Now, because of reference kits, you can get it down in days.

We provide a state-of-the-art machine learning model inside the chatbot reference kit. The model is already provided -- you don't have to go through the exercise of figuring that out. We provide the code ... we provide access to the data set, and we provide a machine chatbot architecture.

All that work used to take quite a bit of time to do. It's not only the amount of time to go through looking for things and deciding things, but it also takes expertise. If you're in the middle of everything, it's easier for you to do the thing. But if you're new to this, it's even harder because it's a learning curve as well. So that's why I believe [reference kits] reduce starting costs significantly.

Can reference kits be described as no-code?

Li: Even though we're giving you code, you don't need to write the code yourself. If you're happy, you can take it, and you can just start running.

Even though we're giving you code, you don't need to write the code yourself. If you're happy, you can take it, and you can just start running.

It's also open source. Once you become better at it, and you find out 'Hey, I want to be creative, I want my chatbot to be a little bit more sentient than others,' you can add more things to it.

So chatbots are sentient?

Li: [Laughs.] Somebody claimed that already ... it would be no fun for me to claim something and not be the first one.

What would you say to people who are wary of losing control to sentient AI?

Li: Sometimes people say we don't know what's going on [with AI], but that's not true because it's a statistical thing. It's not even an analytical thing. Yes, we have unknowns, but are [chatbots] human? I don't think they're human, but on the flip side of it, we don't really understand 'human' either -- biology is less understood compared to math or physics. But are machines human? Maybe in a scientific sense they can be considered human.

Do the decisions in reference kits consider responsible AI?

Li: We are very interested in the topic of responsible AI ... We are developing tools, for example, to evaluate the AI model to make sure there's no bias, because quite often there are biases in the original data set. There are many different issues involved in the responsible AI side, and that, as a community, is still beginning. And that is a good thing. We're pointing out these challenges ... but I don't think it's a solved problem.

Can you explain more about what a no-code AI future looks like?

Li: We will have AI everywhere. AI everywhere means we have different developers and different data scientists with different skill sets. That's how we go to scale. The software bridge I'm building here will have different lanes: It will be fast lane or slow lane for different people with different capabilities. We will provide some people with the knobs they can tune to get the best things out of it -- if I give them no-code, they will get upset because they want to control it. But the opposite is people who don't care -- they just want to have something you can run already. We do see there's a no-code persona we're trying to pursue here.

Editor's note: This Q&A has been edited for clarity and conciseness.