What robots want: Using machine-learning to teach effectively

By Hyunsoo (Hyun) Kim, co-founder and CEO of Superb AI

AI is having a moment. One need only casually scan the news each week to see that the topics of artificial intelligence and machine learning have grown like ivy, extending their tendrils into stories as varied as racial bias, hiring, and of course, identifying spiders.

But for all the diverse applications of AI across our inboxes, magazines and evening news, few outside of the engineering community have a robust understanding of what the terms actually mean, or how the robots and algorithms we increasingly rely upon come to know how to do the complex jobs humans assign to them.

For starters, the machines involved in machine learning are increasingly more likely to take the form of a disembodied hivemind than a humanoid assistant.

Nearly 60 years after Rosie the robotic maid first enchanted American prime time television viewers on The Jetsons, robotic minds and algorithms instead are in demand within nearly every sector of business.

Filling these machine minds with context and experience requires teaching and training. But humans can only teach artificial intelligence so much – or at least at only so great a scale.

Machine Learning is thus the field of study beyond that scale, in which the algorithms and physical machines in question are taught using enormous caches of data. Machine learning has many different disciplines, with Deep Learning being a major subset of that.

Deep Learning utilizes neural network layers to learn patterns from datasets. The field was first conceived nearly three decades ago, but didn’t achieve popularity due to the limitations of that generation’s computational power.

But just as Moore’s Law dictated that the number of transistors on a microchip would double every two years even as the cost was halved, humanity’s ability to teach machines to think for themselves has grown exponentially since then. In fact, the speed at which AI is learning is now wholly outpacing Moore’s Law.

These conditions mean that Deep Learning is finally experiencing its star turn, driven by the explosive potential of Deep Neural Network algorithms, which require enormous amounts of computations but can ultimately be very powerful if one has enough computational capacity and datasets.

But now that machines are capable of learning incredibly vast and complicated datasets, who teaches the machines? Who decides what AI needs to know?

First, engineers and scientists decide how AI learns. Domain experts then advise on how robots need to function and operate within the scope of the task that is being addressed, be that assisting warehouse logistics experts, medical imaging specialists, or security consultants.

How AI processes these inputs falls into two distinct categories: Planning and Learning.

Planning involves scenarios in which all the variables are already known, and the robot just has to work out at what pace it has to move each joint to complete a task such as grabbing an object.

Learning on the other hand, involves a more unstructured, dynamic environment in which the robot has to anticipate countless different inputs, reacting accordingly along the way.

Learning takes place via many different forms, but three among them are: Demonstrations involve physically training machine movements through guided practice. Simulations take place via 3D artificial environments.

Finally, machines can be fed videos or data of a person or another robot performing the task it is hoping to master for itself. All three of these represent types of Training Data, sets of labeled or annotated datasets that an AI algorithm can use to recognize and learn from.

Training Data is increasingly necessary for today’s intricate Machine Learning behaviors. For ML algorithms to pick up patterns in data, ML teams need to feed it with a large amount of accurate training data.

Accuracy and abundance of data are critical for success. A diet of inaccurate or corrupted data will result in the algorithm not being able to learn correctly, or drawing the wrong conclusions.

If your dataset is focused on trains and you input a picture of a lion, then you would still get a train.

This is known as lack of proper data distribution. Insufficient training data will result in a stilted learning curve that might not ever reach the full potential of how it was designed to perform.

Enough data to encompass the majority of imagined scenarios and edge cases alike is critical for true learning to take place.

Machine Learning is currently being deployed across a wide array of industries, from real estate and financial planning to even literature and poetry.

Unmanned vehicles are currently assisting the construction industry, deployed across countless live work sites.

Construction companies use data training platforms like Superb AI to create and manage datasets that can teach ML models to avoid humans and animals, and to engage in assembling and building.

In the medical sector, research labs at renowned international universities deploy training data to help Computer Vision models recognize tumors within MRIs and CT Scan images.

These can eventually be used not only to accurately diagnose and prevent diseases, but also to train medical robots for surgery and other life-saving procedures.

A properly trained robotic tumor-hunting assistant can perform its job all night long, well after even the doctors and nurses on graveyard shift have gone home for the day.

There’s a tremendous opportunity for Training Data, Machine Learning, and Artificial Intelligence to finally help robots to live up to their potential in unlocking medical and technological breakthroughs, relieving humans of monotonous and difficult labor, or even reducing the length of the 40-hour work week.

Technology companies employing complex Machine Learning initiatives have a responsibility to educate and create trust within the general public, so that these advancements can be permitted to truly help humanity level up.

But humans also bear responsibility here as well, in that they owe it to themselves to educate and familiarize themselves with these emerging fields of study.

It will fall upon engineers and data analysts to do the lion’s share of the work in teaching and training machines how to best assist us.

But public opinion is a powerful lever all its own, and certainly one that can be wielded to help shape and frame our future of man-machine teaching and cooperation.

About the author: Hyunsoo (Hyun) Kim, co-founder and CEO of Superb AI, is an entrepreneur on a mission to democratize data and artificial intelligence. He has a background in Deep Learning and Robotics, through his Ph.D. studies at Duke University, and a career as a Machine Learning Engineer.