Watch how LEIP accelerates time to deployment at GTC 2024 March 18-21 Learn more

 

Why AI Needs to be Ergonomic to Succeed

by Caleb Tung | Posted Feb 17, 2023
Share
Why AI Needs to be Ergonomic to Succeed

What do smartphones and AI have in common?

They both share the letter “A”.

They also share status as consumer technology. But they’re not “just” stuff that everyone uses daily; they are tools that can empower us to do more as we build our world.

Smartphones have certainly done that: just think about what happened when you needed to tell your friend (let’s call him “Jeff”) to bring a bag of chips to the party, pre-smartphone era. You actually had to make a phone call, hope you caught Jeff at home, and that he picked up. Even with a message, if you mis-timed your call there was still a good chance there would be no chips at the party. But with smartphones, you can quickly text your request, hit SEND, slip the phone back into your pocket, and get right back to whatever you were doing, confident that Jeff would be notified. 

Granted, getting chips is small potatoes compared to the rise of the Internet. Now, the world’s information is at our fingertips. Unfortunately, it’s also rather easy to argue that the smartphone, perhaps too often, “empowers” us to do less. Colorful, addicting “gotcha” mobile games and oh-so-doom scrollable social media means that we’re spending a third of the day on our phones, absorbed in apps. Rather than helping us build our world, the smartphone has often wound up becoming our entire world.

That’s not something we want AI to have in common with the smartphone. In a rapidly progressing world where search engines can literally chat with you and apps can generate incredible artwork on command, it’s more important than ever to ensure that as AI improves, it should always help us find answers, not become the answer. That means that AI builders need to design their tech to be as ergonomic as possible, specifically when used as tools to aid humanity.

Ergonomic. That word gets used to describe a tool when it contours to your hand, when it allows you to maintain a comfortable posture during operation, when it fits perfectly with your workflow. But, when you attempt to misuse the tech, the ergonomics work against you (imagine trying to use an otherwise ergonomic power drill to apply paint – misusing the tool is not easy, and that’s the idea!)

Latent AI believes that verifiable, repeatable AI technology is a key part of achieving those ergonomics. AI technology needs to consistently give similar outputs when given similar inputs. When a model makes a decision, it has to be extremely easy and transparent for a human to verify if the model got it right. And we want to help AI builders everywhere make sure their technology does just that.

Latent AI’s Efficient Inference Platform (LEIP) empowers AI builders to design the future with ergonomic principles in mind, across the entire AI pipeline. No matter the step of the development cycle, LEIP provides the tools. Our compilers and intermediate representation scheme allow us to ingest different model types and strip out inconsistencies hiding under the hood. LEIP recipes are rigorously tested for reliable accuracy at inference time. Our runtime allows you to deploy the same model the same way on any device, ranging from embedded systems to high performance computers.

AI isn’t going to affect the world of consumer technology the same way smartphones did, but like smartphones, it has the potential to be both used and misused. At Latent AI, we want those developing AI to create the most ergonomic tools the planet has ever seen, and we’re building LEIP to give the industry its best shot. After all, sharing the letter “A” is pretty good when it’s the grade on the report card for humanity’s progress.

To learn more, visit latentai.com or contact us at info@latentai.com

Share
View All

Related