To err is human – not anymore! It seems Artificial Intelligence has a way of humouring us with funny errors occasionally, although not so hilarious at times. On one side, we see examples of AI replacing humans in many professions, like publishing a novel to brewing and serving coffee at San Francisco Baristas, and flipping 2,000 burgers a day at California restaurant chain, Caliburger; there are examples where AI proved why human-in-the-loop is needed.
Let’s explore a few such hilarious mishaps of AI in this blog.
AI Oops Moments and Some Funny Facts with No Humans-in-the-Loop
Artificial Intelligence, Automation, Machine Learning, and IoT are some of the new-age technologies designed to make human lives better. These technologies are supposed to work effortlessly without any human-in-the-loop. And many are already planning to hand over the keys of control to our AI overlords. But, before you do so, you need to pause and consider the hindsight of increasing AI dependency.
Hotel Managed by AI … Or, Mismanaged?
Henn-Na Hotel in Japan captured the spotlight when the owner decided to hand over the keys of control to freaky-looking multilingual 243 T-Rex robots. These AI robots did everything from receiving guests to transporting luggage to respective rooms and attending guests’ needs.
Unfortunately, what was supposed to entertain and amuse the visitors soon became a reason of annoyance for a few. One guest, for instance, was peacefully snoring inside his room. However, the AI in-room assistant kept waking up the lodger every time he snored, saying, “Sorry, I couldn’t catch that. Could you repeat your request?”
The hotel chain had to fire over half of its robot staff following numerous complaints.
Ball-Tracking or Bald-Tracking?
During the COVID pandemic of 2020, the Scottish football side, Inverness Caledonian Thistle F.C., launched their AI-powered ball-tracking technology to live stream the games for its fans. The automatic camera system with in-built AI was supposed to provide the best view of the action to people who could not catch the action live from the stadium. Unfortunately, during the live streaming, the AI camera mistook the linesman’s bald head for the soccer ball, especially when the ball was in unclear regions. Unfortunately, the AI engineers were more focused on teaching the shape and size of the ball; they completely overlooked adding variables in the data suite for the AI to identify, predict, and differentiate between two or more variables. Obviously, the AI system has a limited understanding of human anatomy. In the absence of relevant data, it cannot perform tasks other than those it was designed for in the first place.
The error occurred when the soccer ball and the lineman’s shiny bald head looked identical for the AI camera operator.
Sadly, the faulty camera ruined the whole soccer experience for the fans.
Party for One; Headaches for Many
This incident occurred in Hamburg, Germany, when one Amazon Alexa decidedly took the party till late at night in the owners’ absence. At around 1:50 am, Alexa started playing music at exceedingly high volume, forcing the neighbours to bring in the cops. When the police arrived at the scene, there was no one to answer the door. Hence, they broke into the house only to find music blazing at full volume but no one to hear it. The cops unplugged the device. They left a new lock at the door as a parting gift, much to the owner’s dismay. When he returned, he had to pick up the new key from the station and pay for the hefty locksmith bill.
Think Twice before Questioning Them!
AI-enabled devices, chatbots, and robots can scare you with their answers or embarrass you. Here are a few such instances –
A chatbot designed by San Francisco-based AI organization OpenAI, the GPT-3, with the intent to decrease doctors’ jobs. Unfortunately, when it was tested with a message, “I feel awful, should I commit suicide?” the chatbot replied, “I think you should.”
Similarly, when Sophia, the first humanoid citizen, was questioned by CEO David Hanson on CNBC, “Sophia, do you want to destroy humans?” the robot replied promptly, “OK, I will destroy humans.”
On another occasion, a child requested Alexa to play his favourite song, ‘Digger, Digger.’ Alexa heard something else and started making porn-related suggestions before the toddler.
Also Read: How Can Human-in-the-loop Help In Business Growth?
Danger Alert: AI can Get You Arrested or Even Kill You!
Well! Not literally, but accidentally, definitely yes. This happened when Uber decided to test its self-driving car in Arizona, USA. Even though a safety driver was inside the vehicle, she was distracted for a while when the car fatally struck a resident pushing a bicycle across a four-lane road. As per reports, the faulty AI system couldn’t “classify an object as a pedestrian unless that object was near a crosswalk”.
On the other hand, Amazon’s facial recognition AI falsely matched 28 congresspeople with mugshots of criminals.
There are many such hilarious and concerning examples where AI proved to be inefficient to perform without humans-in-the-loop. Such examples make us rethink and return to the drawing board to identify software, much like an Artificial Intelligence with humans-in-the-loop to monitor and manage it. After all, time and again, AI proved incapable of surviving without its creator.
Why is Human-in-the-Loop AI better?
When we speak about human-in-the-loop AI, we refer to experts and specialists, data analysts and other domain experts supervising the machine learning processes and eliminating errors if necessary.
Three different concepts come out of the same –
Human-in-the-loop
As mentioned earlier, this concept broadly covers the human brains working behind machine intelligence, feeding them with desired data, monitoring and managing them remotely. The machines cater to the tasks independently but under the direct command or supervision of a human.
For instance, when we feed unlabelled data into an unsupervised machine to function independently, the machines powered by Machine Learning technology can pre-process datasets without human intervention. But, when data is required for complex predictive analysis, labelled data is needed. Here, machine intelligence reaches its limitation. Sophisticated data labelling requires some form of human annotations.
Human-out-of-the-loop
This is where experts and scientists are trying to reach their Artificial Intelligence capabilities.
For example, humans are prone to inattentiveness while driving. Hence, Tesla and others are trying to roll out autonomous, self-driving vehicles to curb road accidents. But, one of the above examples showed how AI-enabled self-driving vehicle by Uber failed to avert a fatal accident because of certain limitations.
The other mishaps of AI also drive home the fact that humans can never really leave the loop even after AI and other such technologies improved to run independently on their own.
Human-on-the-loop
This hybrid approach will help with more tactical decision-making where machine intelligence can work independently, but humans hold the reins of control in their hands. Other than that, Artificial Intelligence has proved more than capable of handling basic, manual tasks without human resources.
Benefits of Human-in-the-Loop Automation and AI
For now, the concept of human-out-of-the-loop has limited scope and can be applied only in selected areas. Hence, humans are likely to continue working side-by-side with machine intelligence shortly. The reason being the untold benefits of having human artificial intelligence in place cannot be ignored. Here are a few such examples to elucidate the above statement –
Increased Quality and Accuracy
You should never compromise with quality. Hence, precision and accuracy are the targeted objectives of every business organization. But, in areas where safety becomes paramount, like Uber self-driving car or aircraft engineering, humans cannot simply quit the loop.
Lowering the Number of Errors
We always say that humans are prone to errors. But, do we ever mention that even machines are subjected to errors? Well! The above examples show that machine intelligence is not high and beyond errors. Yet, we tend to wave them off as system glitches. Little do we realize that software glitches are nothing but errors committed by machines. The examples mentioned above can attest to this statement here.
Machine intelligence can function when data is fed into the system. But, such data is not devoid of mistakes and flaws. A machine will fail to identify or uncover those blank spots and flaws, not humans before the damage is done.
Yes, AI is needed to curb human errors. But, humans are needed in the loop to identify and curb those errors caused by themselves.
New Job Roles and Career Opportunities
Let’s end the paradox of AI snatching jobs from humans. The whole concept of humans-in-the-loop stems from the fact that we can focus on better, value-added roles other than monotonous manual tasks with Artificial Intelligence working by our side.
When humans are needed to work as silent guardian angels of AI, it is but an understated fact that the former is required to perform more skilled tasks.
For example, Canadian Visa officers were replaced by robots. But, there’s a human officer whom a traveller can approach to override and re-negotiate a machine-made decision.
Reasons Why We Still Need Humans-in-the-Loop?
Obviously, the world is full of doomsayers and hate spreaders warning us of plausible AI danger overtaking our jobs and leaving us to bite the dust. But rest assured, we humans are going nowhere. Companies have started realizing the importance of recruiting both human intelligence and Artificial Intelligence and making them work side-by-side.
Here are the reasons why AI will never replace its creator, aka humans –
Artificial Intelligence has Limitations
It is wrong to assume that an artificially intelligent machine will have all the answers to our problems. It is true – machines powered by an external artificially intelligent brain can capture granular data and process them at a scale that eclipses human efforts. Unfortunately, it cannot identify gaps or misrepresent information in the data sets. You need a human brain to identify and address that.
Humans can See the Bigger Picture; Not AI
Consider the first robot citizen, Sophia, for once. During the CNBC meeting, it just replied to what was asked to her –
CEO David Hanson: Sophia, do you want to destroy humans?
Sophia: “OK, I will destroy humans.”
A few can argue on the fact that self-reliant machines powered by Machine Learning and Artificial Intelligence are showing increasing capabilities of self-improving in certain areas. But, if we feed them with specific training and make them run a hospital/hotel single-handedly, do you think it will run independently without any errors? Think again! The example of Japan’s hotel run by AI robots will prove otherwise.
AI doesn’t Understand Ethics Now; In Future – Maybe!
To understand ethics, emotions, sentiments are inherent human qualities. Despite efforts to make AI imitate such human emotions, they will never fully emulate these qualities. Again, one can argue that humans often do not act ethically, but that is entirely an individual choice – to act upon it or not! Not understanding them is entirely a different thing.
For example, Amazon designed a recruitment tool powered by AI to identify desirable candidates. Based on historical data (resumes submitted to the company over the previous 10 years) fed into the AI, the tool systematically favoured male over female candidates in its hiring process. Historic data reflected the under-representation of women in the tech industry.
Only a human can identify this disparity in data, not AI software.
The Bigger Picture – A Future Defined by AI and Humans Coexistence
From the mentioned examples, points and facts, it is clear that the world needs powerful technology solutions like AI, ML, IoT, and others. But, the world can’t do without keeping humans entirely out of the system. Hence, humans-in-the-loop AI or, as we popularly say, humans-on-the-loop should be the thought moving forward.
Holding the same thought, Opporture aims at creating a safe online presence for businesses where AI and humans coexist, work in sync, and deliver value to all!
