Artificial Intelligence: The Risks in AI Revolution

0
273
Artificial Intelligence Risk

Miss-classification is the biggest risk posed by AI. The Risks in AI Revolution may be classifying a friendly aircraft as hostile, to blocking your legitimate card transaction as fraud. The proportion of this anomaly might be low but the impact is heavy. So the smart AI may allow a smart spammer’s email into your inbox while blocking an Amateur non spammer. Using AI for screening candidate profiles can be too dangerous at times. Assume a good candidate might not use good keywords in the resume. Well its possible as he is not aware of how systems auto classify documents these days. On the other side a not so good candidates resume might be picked up just because of great keywords. Not that humans don’t make mistake but at least they have a review mechanism which seems difficult in AI enabled solutions. We cannot avoid using AI as its already a part of many things we do today but a cautious approach needs to be selected.

 

Miss-classification is the biggest risk posed by AI

Preparing for the future is not something you can do once. You have to think about it every day. So building an AI powered security solution that makes good use of it is a must. Since cybersecurity is a continuous process and not a once time product. It has to be revisited regularly. AI cannot be seen as a replacement of humans in the short run as this is not a one size fits all solution. It has to be customized according to industry. The kind of companies adopting AI are often small and medium enterprises (SMEs). This not only allows them to grow faster but also deal with issues faced by them. But there are ethical questions raised by AI and the implications they have on society are growing with each passing day.

The problem with AI

The problem with AI is that it is still a far cry from the creators’ original vision of it. Because there were certain areas that were left undefined, the AI process became unpredictable. The creators’ “code” or algorithms became inefficient because the operators could not gauge their unpredictability. With AI comes unpredictability because it does not have an identical set of rules to the code that controls us. What this means is that the applications created based on “machine learning” algorithms will likely bring unpredictability and change in the future. If you are developing an application based on machine learning or algorithms, you should take into account the unpredictability of AI.

Need to Limit Compromises

No matter how advanced the technology is today, the concept of unpredictability is still quite common, in any industry. In the field of enterprise applications, AI would be applied in a vast amount of ways. A customer could spend their time being frustrated because they cannot reach the correct, fast, efficient, and accurate service they are expecting. In addition to that, an employee could spend his time trying to fix the supplier and wasting time trying in dong so. Let’s look at some applications of AI in brief but with a perspective as to what unexpected may happen.

AI Risk in HR

There are web based APIs which have a cloud based deployment of AI services. Often these use in backend the ML solutions of Microsoft, Google or IMB. Some of these services are using Generative Pre-Trained Transformers. Typically, the version would be 2 and more recently we have the GPT 3 version. HR automation has happened wherein the incoming resumes/CV’s for any JD is typically shortlisted by an AI solution. The Risks in AI Revolution lies in the fact that the system tries to match the keywords in the JD with the resume document. Higher the match proportion better the chances of getting shortlisted initially. At a later stage the human interview process would make sure the suitability of the candidate. Although this could come at the cost of loosing a very good candidate, who wasn’t good at picking the keywords. Risks in AI Revolution need to be seen with a retrospective analysis.

AI Risk in Card Authorization denial

Have you every been locked out on your own account while doing a transaction. I have been forced to get a new card on couple of occasions because the banks AI based security management system had blocked it. The customer care could only explain that the new AI system has blocked it permanently and you have to get a new one issued. Its frustrating how an algorithm starts deciding that your legit transaction is fit of denial.

AI Risk in Digital Marketing

Have you ever tried putting up a campaign for your self on Facebook or google. Face book and google have been the pioneers in ML and AI. They use it extensively to mark up a post as spam. A couple of time my campaigns very not running as the proportion of text in my image was higher than what’s expected. After a backend reach with FB, we could get it running. This came at a cost of 15 days of delayed run.

So there could be many advantages of AI. But some ethical, social and personal loss on account of non audited usage of AI. We need to minimize the risks in this AI Revolution.

How to minimize the risks

The best way to minimize risks is not to believe a machine that is predicting bad outcomes. To overcome that, use it judiciously. A machine can never know the nuance or personality of an individual. When a profile of a potential candidate is being generated, an algorithm is likely to look at the language used in the resume. So it would be better to stick to hiring people with good grammar and use AI to support people. Any business that has to deal with personal data will need to manage the public interactions better. Exhaustive use of AI can do this with little effort and that can be an aid to the screening. Handle the risks since algorithms are imperfect, they are designed to improve as they are fed more data.

Conclusion

Huge monetary benefits are associated with AI. With rapidly increasing competition for a huge market a lot is to be gained. However, huge investments are being made and much still to be lost. What really matters is the quality of AI that it delivers and how it is deployed. Machine learning is an innovative solution but its deployment may be too flawed to deliver on its promises. So it is important to select the AI with a right set of algorithms for the right type of problems. This would help to deliver the best possible solution. The solution delivered is also important and do not be caught by fancy words but by the application. Suggest you to go through my free machine learning course which will help you to understand Machine Learning better.

LEAVE A REPLY

Please enter your comment!
Please enter your name here