WHILE artificial intelligence is already making important decisions about our lives, people still don’t trust it when left unsupervised, according to a new study.
Instead, it turns out that people prefer human intervention in digital decision-making.
1
While AI has established itself as on par with human brain functions in many ways, decision-making plays out in a different realm.
Artificial intelligence draws laser-like inferences based solely on facts and algorithms, but that’s not the scaled down and empathetic way humans interact with the world, according to Harvard Business Review.
In a recent poll conducted by multiple parties including Intel and Forbes, one in four executives said they had to reverse AI decisions due to impassable errors, according to the outlet.
Up to 38 percent of respondents said the machine provided erratic or dissonant conclusions, while another 34 percent said what the robot was doing was overall inappropriate.
The trolley problem raises a theoretical question: if a trolley car were blocked in its path and about to kill several people tied to the track, an AI could successfully decide to move the trolley from its original track to one at the last second remove alternate track where only one person would die?
It’s considered the pinnacle of moral questioning, and scientists aren’t sure AI has such a profound ethical capability at this point.
These issues crop up throughout the artificial intelligence world and are not always a matter of life and death.
self-directed
Uber was forced to abandon a self-driving test after a pedestrian was killed.
A test vehicle put on the road by the ride-sharing giant struck a civilian in Tempe, Arizona, Harvard Business Review reported.
While a human would have seen and stopped the person many times, the self-propelled AI mechanism failed to recognize the unforeseen Jaywalker, resulting in their death.
Although a backup human driver was on board and watching the streaming at the time of the accident, temporarily giving the person the lion’s share of blame, the National Transportation Safety Board ultimately decided it was primarily a technological glitch.
Recruitment 101
Amazon has developed an AI tool to recruit the best technical minds in an era of constantly outsized technological advances.
Aimed at recruiting the best talent, the tool used decades of data to find the best things in the best resumes.
Sexism ended up being a problem with the tool as the data was mostly collected from men and the algorithm falsely started to eradicate bias against women’s activities like ‘Women’s Chess Club’.
After finding that the tool refused to be gender neutralized, Amazon shut down the recruitment tool entirely.
TAY’s world
Microsoft’s chatbot has a name that stands for Thinking About You.
Left to their own devices, TAY began using racist language and other slurs on Twitter users.
As a self-learning mechanism, it should teach itself from human behavior and interaction.
Unfortunately, the bot started mimicking the troll-like behavior in addition to spreading false information without bothering to fact-check it.
Just 24 hours later, Microsoft had to revoke the experiment.
artificial morality
AI also carries the risk of bias due to biased or infected data.
Data, like anything else, is subject to human bias or error, and since artificial intelligence works with algorithmic data, it can easily mirror these human biases.
Even the most advanced robotic systems lack human values, according to the Harvard Business Review.
While the aspect of human nature can be taught to machines, they have yet to exemplify this instinct themselves.
Scientists have found that BERT, GPT-3 and Jurassic-1, advanced language transformation mechanisms, are getting closer and closer to working accurately without human correction.
The artificial strain’s intelligence can help evaluate data and aid in decision-making.
However, it is still a flawed, emotional, organic intelligence that makes the best decisions for other fallible people.
