The Next Economy will be driven by 10 key emerging technologies, underpinned by artificial intelligence. The vision is that it will become a general purpose technology, like a utility, that is likely to scale across every industry, every sector of our economy and nearly every aspect of science, society and culture.
Depending on who you ask, this vision also has its drawbacks. Specifically, the fear is that A.I. will replace people. This is what gets the most media attention, but A.I. is nowhere near the cognitive capabilities of a human and no one agrees when that will happen; it’s still in the early stages.
Current AI training methods require reams of labeled data
The limitation is the amount of data needed, and the time and labor to label it properly. The main way computers learn to do tasks is by getting fed information by us. Just like a child needs to be taught what’s what and what isn’t, so do computers.
“Today, in order to teach a computer what a coffee mug is, we show it thousands of coffee mugs. But no child’s parents, no matter how patient and loving, ever pointed out thousands of coffee mugs to that child.” — Andrew Ng, former AI lead at Google and Baidu
This is where we are.
The question is not whether A.I. will take over our jobs, it will, it’s how soon and how. One thing is for sure: it will happen in phases; first it will create new ones where it will augment our abilities not replace them.
It’s just like the early days of the internet had webmasters, people who looked over websites. Today the new job is tagger: people who prepare data so algorithms can learn to do our jobs.
Read the NY Times article titled “How cheap labor drives China’s A.I. ambitions” to see how early stage we are and how this is already happening:
“I used to think the machines are geniuses,” Ms. Hou, 24, said. “Now I know we’re the reason for their genius.”
A.I. has to be taught. It must digest vast amounts of tagged photos and videos before it realizes that a black cat and a white cat are both cats. This is where the data factories and their workers come in.
In the summer I was having lunch with a friend and I mentioned that I foresaw a business opportunity in data tagging, specifically because most companies that want to develop their own internal A.I. capabilities will have that obstacle. This is entirely obvious if you work in the A.I. industry, so I’m not surprised that data factories are cropping up in China where the demand is even higher.
We have to be thoughtful about how we label data
One of the biggest obstacles to adoption is the acquisition or development of ones own datasets. Organizations have loads of data about their customers, but that data needs to be cleaned and labeled before it’s fed into an algorithm that will make sense of it all. This is not something most organizations are prepared for, getting a properly labeled data set is a real stumbling block for companies designing new AI systems.
My company, Netek, does emotion recognition technology through cameras and EEG. We’ve accumulated around 1 million faces with various emotions, and used this dataset to teach our algorithm to detect when someone is happy, sad, angry, contempt, surprised or disgusted. Gathering and labeling this data didn’t happen in a couple of weeks, but in months; with only a couple of people working on it.
Anybody who is supposedly doing A.I. work has to go through the same process of acquiring and labeling data before feeding it to an algorithm. This is the most important part of the process because it’s not just about accumulating data, we have to be thoughtful about how we label data.
So, A.I. will not replace people anytime soon. It will take over some tasks first, that’s a given, but this only happens when humans prepare data so it can properly learn.
Bottom line: Everyone wants well-trained A.I., but it takes a lot of manual labor by humans to get it. With that said, the new role for humans is to prepare data so A.I. can learn to do our jobs.