Updated: Jul 18, 2019
AI has been generating hype in the past few years. This can be attributed to the futuristic media coverage about superhuman AI. Furthermore, films and television sci-fi content has projected artificial intelligence as a threat to mankind. Real commercial applications and challenges of AI seem to be lost deluge of such content.
For a large number of start-ups, the road to AI implementation has not been easy. Most simple AI applications in the industry face some challenges in the area of implementation. This is due to the fact that good AI talent is costly and having customized solutions for the data in the organization for each function is a costly affair. Therefore, a number of AI-enabled applications have come for specific applications like lead generation, manufacturing analytics, etc. But, most large corporates are still struggling with AI tool with the flexibility to create customized solutions without requiring a PhD in artificial intelligence!
But, there seems to be a new category of AI tools rising which allow for integrating AI in the current systems without having great expertise. The US-based startup, DataRobot was one of the early players to identify the opportunity. It allows the user to enter the data of any kind and then predicts the machine learning model suitable for such data. It also auto-tunes the parameters to give an optimal AI training model for the task to be accomplished which may be identifying patterns, classification, prediction, etc.
But, a range of major cloud providers with AI services are catching up with solutions in this regard. Amazon Web Services’ SageMaker and Azure Machine Learning Service from Microsoft Azure are some of the prominent examples. Nikkei reports that “the automatic tuning function (Automated Machine Learning) of the Azure Machine Learning service has developed an AI for recognizing handwritten numbers from 0 to 9, and achieved an accuracy rate of 98.3% compared to that of machine learning experts. “ The results were compared with Kaggle participating teams which are usually professional quality. Also, important to note was that the time required for automatic tuning of AI was only 18 minutes which is not feasible for most human beings.
The results can potentially predict an important trend of the AI cloud providers coming up with tools so that automatic tuning of AI becomes a norm. This would allow the development of basic AI enabled applications in the company itself without having machine learning experts in-house.
It is also important to note here that data preparation and execution still requires understanding machine learning but doesn’t require expert level knowledge as is the case today. Also, the applications created with such a method have simple, defined outcomes like identifying patterns, classification, etc. which are already heavily researched tasks and optimal AI algorithms for such cases have already been identified and proven. AI cloud providers would continually improve on this capability to enlarge the set of outcomes possible using an automatic tuning. This could pave the path for truly large scale adoption of AI in the current business systems. A non-exact analogy would be the way templatized website builders like Wix or Squarespace helped in creating a large number of websites for people who were not web developers!
To deep dive and stay continuously updated about the most recent global innovations in Artificial Intelligence and learn more about applications in your industry, test drive WhatNext now!
Video Courtesy : www.openai.com