Optical Character Recognition (OCR) is a subset of machine vision. OCR converts text in a picture or scanned document into machine readable text. It recognizes characters in the picture and converts it into text. A common application is converting a scanned copy of a document into an editable Word or Excel document.
This can save businesses a lot of time. You may have had the (dis)pleasure of doing data entry at your day job, opening one PDF file after another and typing selected info into a spreadsheet for analysis or record keeping. This (in)glorious work never ends, because new documents just keep coming.
OCR-enabled software can do this automatically. Also, why stop at converting physical documents and PDFs to editable files? If your company uses dozens of vendors and receives hundreds of invoices, a software package can combine OCR with other AI methods to:
- Recognize characters and numbers on invoices using OCR
- Convert the invoices to editable text files or spreadsheets and automating data entry
- Use Natural Language Processing to extract relevant keywords and numbers and create an invoice database reporting templates or dashboards
- Automatically generate management reports and dashboards about what is being paid to vendors, what for, and how frequently
- Analyze the database to create forecasts or predict potential issues
Large corporations across industries deal with insane amounts of paper documents and have entire teams devoted to document processing and management reporting. They would save significant time & money automating away some of these routine tasks. Accuracy would improve and there would be less time spent on data reconciliation. White collar workers everywhere would rejoice at being freed from the drudgery of data entry and processing.
Of course, this could also result in many data processing jobs being eliminated, but that’s a separate issue.
Robotics involves designing, creating and operating robots to automate mechanical tasks such as factory work. However, most industrial robots are not intelligent — they only carry out pre-programmed movements. These robots can’t learn or improve, and have limited functionality.
Robots can also be powered by AI to carry out more complex tasks, or learn new tasks they are shown without being reprogrammed. Computer vision could help robots ‘see’ the objects they’re interacting with. Machine learning, neural networks and deep learning can help them classify objects and ‘understand’ what they’re doing and get better at it.
A self-driving car is an example of an AI-enabled robot. It sees and understands objects on the road and drives the car in different situations.
In manufacturing, an AI-enabled robot could be quickly moved to new jobs without being reprogrammed by a human.
Want to see a humanoid robot? Check out Sophia from Hanson Robotics. She (kind of) looks human, can hold a conversation, and even has (creepy) facial expressions. Computer vision allows her to recognize faces and NLP enables her to understand speech. She claims that she “loves humans,” which is far better than the alternative.
In future, more advanced humanoid robots could keep elderly people company and act as caregivers, thereby alleviating loneliness.
AI Robotics in action: Amazon’s robot army
Amazon’s fulfillment centers are huge — around 1.25 million square feet. This is where products are stored, packed and shipped from. Millions of products have to be moved, packed, and processed. Staffing these places would require thousands of people and create plenty of gainful employment, right?
Not quite. Amazon’s fulfillment centers are largely operated by AI-enabled robots, like the guy in the video above.
Thousands of these robots move millions of products around the warehouse to the right location, faster and more accurately than a human could. These robots can see where they are going, identify which products they need to handle, and don’t get tired or injured.
How much longer before robots like these get into other industries? And how will it affect human jobs?
Will AI benefit society, or create some sort of dystopian nightmare? Stephen Hawking and Elon Musk think that AI, left unchecked, could doom us all. Others are more optimistic and claim AI will improve our quality of life. As with everything in life, the truth is probably somewhere in the middle, and only time will tell.
Security & Convenience vs. Privacy
Facial recognition cameras can improve surveillance and aid crime fighting. We already have the tech to be able to pay for stuff in stores just by looking into a camera. Nobody is against a safer world and an easier life.
But what about the data collection and privacy implications? Can we be tracked everywhere we go and be watched 24/7? What happens to our personal freedoms? What if companies and governments misuse the data they gather on us? If you think the Facebook & Cambridge Analytica scandal was bad, consider that facial recognition can track everything you do, whether you post it online or not.
Automation vs. Job Loss
Chatbots will provide 24/7 customer service and save businesses money. Call center workers will lose their jobs. Self-driving vehicles will improve road safety. Cab drivers and truckers could lose their jobs. Warehouses and factories will be staffed by robots. These workers will lose their jobs. AI software will automate back office tasks. Many lower level white collar workers will lose their jobs.
The benefits of automation to business are not in doubt. The downside is that entire professions might be rendered obsolete very soon. It isn’t easy to re-train a many workers, especially if they are older. It’s unrealistic to assume that a truck driver in his fifties is going to easily transition into another well-paying career. In fact, many traditional jobs are at risk of being automated.
Dow can we help the people who are impacted by automation? Create laws to protect these jobs sacrifice business and technological progress. Should we tax tech companies automating jobs away and distribute these taxes to the public? Should we revamp education systems to train kids for jobs of the future? There are no easy solutions and unfortunately, most people are not thinking enough about these problems.
In the U.S., businessman Andrew Yang is running for President on a promise to give all American adults a universal basic income of $1,000 per month. Yang recognizes that the most common jobs in America are being automated away and that people and communities that depend on these jobs will be left high and dry.
Yang plans to implement a value added tax on goods and services such as Amazon sales and robot truck miles to help pay for this $1,000 ‘Freedom Dividend.’ He is one of the first to run for higher office that acknowledges this problem and perhaps mainstream politicians will follow suit.
We already have remote drone strikes that can take out a building half a world away. Similar technology paired with facial recognition and other AI methods could theoretically track and kill single targets. A video dramatizing what these ‘slaughterbots’ could do was created by an arms control advocacy group. Surgical strikes — assassinations — by killer autonomous drones. Most of the tech already exists and is not that expensive. The potential for misuse is frightening and international agreement is required to prevent this sort of misuse.
Deepfakes use AI techniques to alter images and videos. It can be used to make it look like someone said or did something that they did not do. This would be the ultimate evolution of fake news.
The tech is getting more convincing. It is already possible to doctor videos of world leaders to make it look like they said something they didn’t say. This ability to alter reality could soon be misused in political campaigns and everyday life.