Artificial intelligence is already having a immense effect on our society. An impact which promises to become even greater as the technology is becoming more sophisticated. It doesn’t mean there are only advantages out of it.
We will discuss few disadvantages of artificial intelligence, which we all need to be aware of:
1. Unemployment:
With increase in fears that the automation and AI will change the course of the way we used to work and force the people into unemployment. Questions about which jobs needs to be replaced by the machines in the near future are being raised. Some experts say that the potential shifts in occupations are imminent by 2030, estimating that around 75 million to 375 million workers need to switch jobs and should learn new skills. This shows a really huge gap in the predictions which are ranging from optimistic to very pessimistic, and highlights that many experts from various technologies and business sectors do not share a common point of view on the future of our labour market. it’s really hard to determine how many jobs will actually be lost by people.
The transition to a more automated world will be a major challenge for many countries as ensuring that workers have the skills and the support needed for the transition to the new jobs is not easy. It is because the impact of automation is more effective for low-skilled jobs, such as administrative tasks, construction, and logistical services. Hence, the diffusion of robotics and AI contributes to the reduction in much available jobs for the less educated and has a negative effect on the lower paid jobs. This particular disadvantage of AI could lead to a growth in mass unemployment.
2. Lack of transparency:
Al can also be wrong in many ways which is the main reason of why the transparency is extremely vital. The input data can be entered with many errors or be poorly cleansed. Perhaps the data scientists and the engineers that trained this particular model, selected much more biased data sets to begin with. But with these many things that could go wrong, the real problem is certainly the lack of visibility, which is not knowing the reason why the AI is performing so poorly. In a typical application development, there is a quality assurance as well as testing processes and also there are tools which can quickly spot any bugs if present.
But AI is not just a code, the underlying models could not be examined to see where the bugs are. Some machine learning algorithms could not be explained and are kept in secret. This leads us to limited understanding of the bias or the faults AI could cause. The problem is such that there is little oversight and transparency on how these tools actually work.
3. Biased and discriminatory algorithms
This leads us to our next topic. Bias is not just a social or a cultural problem. It is equally found within the technical sphere. Design flaws and imbalanced data that is being fed into the algorithms can lead to a biased software and technical artifacts. So, AI just reproduces race and age bias that already exists in our society and deepens the social and economic inequalities. You have probably read about Amazon’s experimental hiring a few years back. That tool used artificial intelligence to search for the candidates by ranking them from a range of one to five stars, much like how the shoppers rate products on Amazon. It was discriminatory against women, because Amazon’s computer models were trained to select the applicants by observing the patterns in resumes submitted to the company over a 10-year period, effectively preferring only male candidates and penalizing resumes that included the word “women”.
4. Profiling
AI can be used to build precise profiles of the people. Algorithms have been developed to find out the patterns, so when testing their abilities and gathering personal data in a particular contest, it became very clear that they were able to predict a person’s likely future location by observing the previous location history. The prediction was more accurate when it started using the location of the users’ friends and social contacts. Sometimes this disadvantage of artificial intelligence is downplayed. You might think that you do not care about who knows your movements, after all you have nothing to hide. First of all, chances are that this is not entirely true. Even if you do not do anything wrong, you may not want your personal information available to everyone. After all, you would not move in a house built with transparent walls. Definitely not. Information is power, and information we give up is like giving someone power over us.
5. Environmental impact
Although AI could have a positive environmental impact, for instance by enabling the smart grids to match the electrical demand or enabling smart and low-carbon cities. However, one of the disadvantages of artificial intelligence is that it can also cause equal significant environmental damage due to its intensive energy use. A 2019 study found that a particular type of AI (deep learning in natural language processing) has a huge carbon footprint due to the fuel the hardware requires. Experts say that training a single AI model produces 300,000 kg of Carbon dioxide emissions which is roughly equivalent to 125 round trip flights from NYC to Beijing or 5 times the lifetime emissions of an average American car. The training of the models is not the only source of emissions. The carbon impact of the infrastructure around big tech’s deployment of AI is also significant. The data centres need to be built and materials that needs to be used should be mined and transported.
