Artificial Intelligence is the technology bringing profound changes in our lives. However, it is not yet the technology delivering the promise of machines with human intelligence. While AI matures and adapts to reach that state, businesses can continue to take advantage of artificial intelligence.
Christopher Mims, a columnist at The Wall Street Journal, says “tried-and-true algorithms, applied to narrowly defined problems, are far more useful today” than the most seemingly intelligent artificial intelligence getting all the attention. Mims also explains how organisations can build AI that actually works for them.
Lack of material impact
Some of the most popular AI systems grabbing all the headlines and attention are programs like DALL-E or GPT-3. DALL-E can create images based on text suggestions while GPT-3 is a large language model that can not only generate text but is also capable of writing scientific papers.
Another AI system gaining attention these days is LaMDA, a chatbot designed by Google capable of producing humanlike conversations. It is so good at producing conversations like humans that one of Google’s engineers called it to be sentient. The search giant not only rejected the claim but also fired the engineer.
With these programs and AI systems gaining attention online, there seems to be little or no material impact from these systems. “The AI systems that currently matter the most to companies tend to be far more humble,” Mims writes.
AI designed to simplify
Mims uses the example of Phuc Labs to show how artificial intelligence is now being used to simplify some of the common processes. Founded by entrepreneur Phuc Vinh Truong, the startup is working on a new way to use AI to make recycling electronic waste more efficient. The idea is to start with “chopped-up debris left after recyclers of batteries and other e-waste crush old electronics.”
Usually, this waste is processed with a variety of techniques, including chemical separation. Phuc Labs suspends the particles in water and then channels the slurry through tiny tubes. During this process, a camera is used to capture its passage at 100 frames per second.
Once the camera captures these frames, a computer running machine-vision algorithm analyses each of these frames to distinguish between the metal particles valuable to recyclers and everything else. This vision valve technology designed by Phuc Labs is in its early stages and Phuc Labs is working on a pilot program with IRI, one of the biggest recyclers of e-waste in the Philippines.
Doing less with AI
The story points out how engineers have found that trying to do less with AI will ultimately lead to success. Mims gives examples of autonomous driving systems and its failure to deliver on early promises of full autonomy.
The WSJ story also gives the example of Dr. Stork, co-founder of New York-based PreciTaste, who is using AI-based sensors and algorithms to predict how much food will be ordered by people at any given moment. PreciTaste uses an array of wall-mounted cameras equipped with machine vision to track an order from the moment its raw ingredients leave a refrigerator and “ready to be handed to a customer.”
PreciTaste wants to use the technology to help restaurant chains waste less food and do more with fewer workers. This idea of reducing waste at a restaurant chain shows how AI can be used to do less while also creating maximum impact in the process.
To make its systems work, PreciTaste is training depth-sensing cameras to recognise how much of an ingredient remains in a prep tray. A prediction algorithm is designed and fed with data such as the demand for a particular product, external factors such as weather and local holiday, to help restaurant chains with their kitchen.
Mims says these processes show how AI is still not as efficient as humans and requires engineers and data scientists to do a “lot of hand-holding.” All these examples prove that business can build AI that works for them but in most situations, AI systems lack common sense.