What Happens if AI Doesn’t Live Up to the Hype?
While AI has been projected as the next frontier in technology, there are doubts being raised over the effectiveness of some aspects of machine learning. AI may not live up to the hype.
Popular opinion has it that artificial intelligence is the future—machine learning has delivered some groundbreaking technologies in the hands of the human race that will change the way people live, work, travel and even eat and exercise.
But how much of this hope in AI being the next frontier in technology is just hype? And what if AI fails to deliver as promised?
This question has indeed suddenly gained currency following a few developments, some known and others not so much in the public domain.
Here are a few points to address:
Machine Learning and Deep Learning: Opportunities and Considerations
Machine learning is the technology behind artificial intelligence. If you are using a mobile phone or a computer with the appropriate capabilities, the device keeps track of what you do and tries to detect and analyze patterns. If you keep using a set of apps regularly, it tries to push them to the top of the interface, making it easy for you to access them. This is the process behind Apple’s Siri suggested shortcuts feature in iOS-run devices.
As its name suggests, machine learning means your device gathers information from you and starts acting accordingly. This same technology has been taken to autonomous cars where the basic requirements of skills for driving are fed into the system and it is able to navigate through traffic, making the adjustments a human driver would do.
But here is the catch—according to experts who have shared their insights recently, one finding is that machine learning or AI technology has limitations when confronted with complex situations in real life. Staying with the same driverless car scenario, concerns have been brought forward that AI systems do not know if they should save the passengers inside the car or the pedestrian crossing the road, in the event of an accident.
Deep learning is part of machine learning, and it is possible here that technology tries to tell us that it cannot replace humans in certain skills. Areas like investment finance, too, have been found to be complex for AI-driven algorithms to arrive at the correct decision or solution which a trained human could do with alacrity.
Some Indirect Hints That AI Could Be Losing Its Sheen
Those who have gone public with their comments and apprehensions on how the hype on AI could be slowly waning have cited a few instances to drive their point home.
One relates to how companies like how Facebook and Baidu went out of their way to hire high profile AI experts only to see them leave the job or be shifted to lesser roles within the organizations. The reference is to Andrew Ng and Yann LeCun. Ng left Baidu last year to start his own AI initiatives, and back in January LeCun stepped down as the head of Facebook’s AI Research Lab and took a new position serving as chief AI scientist.
These companies had set up dedicated laboratories to work on AI. Corporate businesses are very sharp; they make huge investments where they see the prospects, and the moment they sense it may not deliver, they are equally fast at refocusing the projects.
But one particular paper by AI researcher Gary Marcus of New York University seemed to carry a lot of substance. In the paper, which first appeared this year, Marcus gives clear examples of how deep learning has its limitations. He pointed out how, in the popular video game Breakout, the algorithm “DeepMind” comes out quite confused when one of the tools in the game is changed to a different size. In a similar situation, a human being is able to adjust quickly and play the game better.
These points have added up to building a narrative in which AI is not seen to be as powerful and futuristic as it is hyped up to be.
The other inference on how the future of AI-based technologies could be not as rosy as they’re made out to be is from the way some investors are reacting to the developments in the industry. There are figures to show that the enthusiasm with which venture capitalists used to eagerly invest in AI startups is no longer rising at the same heights reached in recent years. Venture funding in AI technology has begun to plateau in the United States.
The investments are either being held back or there is a clear realization that the timelines earlier imagined for many of these deep learning projects to deliver are being extended, keeping the investors guessing if it was prudent to invest good money over bad money.
But It Is Not All Gloom and Doom
One must hasten to add here that these developments do not immediately signal the end of the road for whether machine learning or artificial intelligence as an application technology.
In specific areas, AI has been found to be extremely useful in offering solutions. Some of the verticals where experts feel AI has been able to bring some difference in the way operations were earlier carried out include the fields of medicine, agriculture and even insurance.
The underlying message appears to be that where a large amount of data has to be analyzed in half the time and the solution presented for the humans to take the next step, these algorithms seem to be doing well. Tasks such as checking the creditworthiness of applicants for loans or other credits, based purely on recorded data, are done through AI-based technologies successfully. Firms have found this method more reliable than using people to do these tasks, and it’s highly cost-effective too.
Given the above points, it can be determined here that machine learning is useful, though AI cannot fully displace humans from some key decision-making scenarios. And to that extent, the hype may turn into one of long-term hope—at least in a certain faction of applications.