When it comes to being competitive in today’s market, AI (Artificial Intelligence) is a must-have. Nevertheless, even for some of the world’s top companies implementing that technology was challenging. It is because there are issues with creating models that generate enough ROI, finding the right talent and data problems.
Thus, many artificial intelligence projects fail. Only about 35% of organizations succeed in getting models into production successfully, according to IDC.
Ankur Goyal is the CEO at Impira. He said that while they see ai performing a swath of incredible feats such as AlphaGo, Google Translate, and solving a Rubik’s cube. Thus, it is hard to tell which business problems AI is apt to solve. Goyal adds that it has led to a lot of confusion. The confusion was that a vendor community took advantage of it by labeling things like artificial intelligence when they are not. It is very similar to what happened in the early last decade. That time cloud technologies took off, and, as Goyal says, they had a lot of cloud washing going on. When their offerings were vaporware, they had vendors marketing themselves as cloud players. Ankur Goyal says that they are going to something similar to artificial intelligence washing now.
We all should keep in mind that artificial intelligence is not magic. Artificial intelligence can’t solve all the problems of the company. We need to take a more realistic approach to technology.
How to use Artificial Intelligence
Santiago Giraldo is the Senior Product Marketing Manager of Data Engineering at Cloudera. He is saying that machine learning (ML), unlike traditional data analytics, is not always going to offer clear-cut answers. Giraldo adds that to implement artificial intelligence into the business requires experimentation. Nevertheless, not every experiment is going to drive ROI. It is often built on top of many failed data science experiments when artificial intelligence is successful. Taking a portfolio approach to artificial intelligence and machine learning enables greater longevity in projects and in the ability to build successes more effectively in the future.
Gus Walker is the Senior Director of Product Management at Veritone. He says that it frequently happens that businesses take on artificial intelligence projects, and they don’t realize that it could be cheaper to continue a process manually. Thus, they invest vast amounts of money and time into building a system that does not save the company money or time.
Of course, no one wants to spend money and time on a project and then realize there are compliance or legal restrictions. That can mean having to abandon the effort.
Debu Chatterjee is the senior director of platform ai engineering at ServiceNow. He said that firstly, the data of customers must not be used without permission. Secondly, he adds, that bias from data must be mitigated. It is because any model, which is a black box and can’t be tested through APIs for bias, must be avoided. In nearly any artificial intelligence model, the risk of bias is present, even in algorithmic decisions.
- Trading Instrument