Artificial Intelligence

5 Trends in the Artificial Intelligence Space that will Dominate 2018-2019

By: lisamorrison118
0Shares

AI, machine learning, and deep learning have been buzzwords in the tech arena. The past year saw an explosion in implementation of AI and machine learning solutions, but with the focus shifting away from a basic analytics approach to a more advanced machine learning era.

As practical applications of AI become more common the hype and miscalculations surrounding this technology grows. We cut through the hype and focus on the top five trends that will be on the agenda of the top CIO’s and IT leaders.

Read on to learn how these advanced AI-based applications and tools influence businesses in 2018-2019.

1. Biases will continue to be a problem for AI

Today’s machine learning algorithms often use the information contained in our collective hive mind. No amount of programming can get rid of this dependency. Since this information is generated by humans, it contains all the biases that color human thoughts.

This influences the AI behavior, and Tomer Shiran, CEO of Dremio, believes that a debate centered on data sets is inevitable in the near future. Other experts also agree that enterprises need to improve their data in order to optimize AI training.

Adding AI to products makes them smarter and autonomous to some extent, but the bias is an obvious problem. The models driving these products are trained on data sets and any problems in the data set will eventually crop up in the resulting decisions.

Here are a few instances where this problem arises.

 

  Trump Tweetbot

A Twitterbot created by MIT is modeled after Donald Trump’s Twitter feed. The bot (called DeepDrumpf) analyzes President Trump’s language patterns and creates similar sentences based on the learned style. The resulting tweets clearly embody all the characteristics of Trump’s speeches and tweets.

Of course, it doesn’t exactly make sense most of the time as shown by this example

Nonetheless, it has gotten eerily close to the linguistic style of the original.

 

  AI-Judged Beauty Contest

Researchers decided to define the parameters of beauty and wrote five algorithmic “judges.” Criteria included facial symmetry, amount of wrinkles, and chronological age vs perceived age. The results included 44 winners, most of whom were white, a few Asians and almost none that had dark skin. This might be a funny SNL skit-worthy joke about racist robots, but it shows a deeper flaw.

 

  Tay the Nazi-Quoting Bot

Built by Microsoft, Tay was supposed to be a fun and smart Twitter chat-bot that interacted with 18-24 year olds. It was programmed to become smarter as it engaged more and more with humans. Sadly online interactions changed it to a racist bot that praised Hitler! The bot was deactivated after 16 hours and has been quiet ever since.

Things are not entirely bleak in our quest to remove biasness. A startup project, Knowhere News, has tried to combat the menace of lack of objectivity in news coverage by creating an AI that writes indifferent literal reporting. That being said, AI still has a long way to go.

 

2. The hype will still outpace the actual breakthroughs

Ramon Chen, the chief product officer at Reltio says that AI predictions over the past few years have claimed that major breakthroughs are just around the corner. However, in reality, businesses that invested in AI and machine learning have yet to see any major quantifiable benefits. This hype has increased interest, but most enterprises still seem reluctant to make long-term commitments. This is mostly due to lack of expertise and lack of reliable training data.

Chad Maley, VP (Marketing) of the analytics firm Teradata reiterates Chen’s view. He believes this year will witness a backlash against last year’s hype. As a result, things will settle down as a balanced approach is advocated when it comes to utilizing AI in industry settings.

 

3. Rapid cloud adoption will accelerate and facilitate AI

According to Horia Margarit, Senior Director of Data Science Solutions at Qubole, enterprises will be looking to change their processes and infrastructures to make them AI-ready. These efforts will focus on cloud-based solutions and tools because of its scalability and support for big data. Large-scale data collection and manipulation is easier with cloud technologies that allow on-demand computational power and storage. Another aspect that affects cloud adoption is available internet speeds for businesses. Spectrum internet delivers more reliable and secure connectivity.

 

4. AI audit trails will become a requirement

The decision-making process of AI is often impossible to track. The algorithm simply reaches a decision and presents a viable solution. This is referred to as a black box process, where the steps taken to create the end-result can’t be observed and therefore can’t be audited.

Nima Negahban, co-founder and CTO of the high-performance firm Kinetica says that in order to encourage wider adoption, especially in highly regulated industries, AI solutions will have to create audit trails. According to him, AI applications are being used in industries that directly affect human lives, such as driverless vehicles and drug discovery. Tracking the exact steps that led to an erroneous decision will be essential in these cases. These errors can be corrected with human-written code to eliminate such problems.

 

5. Enterprises will ramp up AI usage

Enterprises are currently using AI-based solutions on a smaller scale, such as chatbots to increase customer engagement. This limited usage will be a thing of the past as enterprises look to integrate AI on operational scales. Negahban says that firms have spent the past several years educating themselves about available AI-based tools. This preliminary evaluation phase will result in a wider deployment of automated solutions.  According to Negahban, companies will look towards developing tools that automate and streamline their machine learning lifecycle.

 

Last Word

Judging from these trends and expert opinions, wide adoption of AI and machine learning will become a reality. However, the rapid pace implied by all the hype is an exaggeration; things are moving along at a slower speed. Companies have plenty of time to evaluate the feasibility of present solutions and integrate them into their day-to-day operations.

BECOME A CONTRIBUTOR

Join Our Blogging Community and Become a Contributor Today! We are proud to have some of the brightest minds in the industry share their thought leadership and experience with our audience. These contributors provide valuable insights.

Read the Guest Posting Guideline Carefully and Register to become contributor.

Note: The Links allowed are No follow. For Dofollow links ( Sponsored post ) please contact admin[at]myvigour.com.
Back to Top