2018 has been a bumper year for AI journalism. While it’s up for debate whether this has been a help or a hindrance to the field as a whole, there’s no doubt that the AI ecosystem is an endless well of fascinating topics to explore. Whether you prefer a technical deep dive or a more casual thought piece, there have been some gems hidden amongst the hype.
We’ve already reviewed our own content for 2018, but we’ve also taken a moment to look back at stories from around the web that kept us glued to our screens this year. Covering everything from AI’s progress with video games to psychopathic algorithms, here are 12 of our favorite articles about AI and machine learning from the last 12 months.
Our Favorite AI Articles from 2018
Stop Feeding Garbage to your Model: The 6 Biggest Mistakes with Datasets and How to Avoid Them, Hacker Noon – From low quality to unbalanced classes, there are plenty of things that can prevent your dataset from being the best it can be. The advice Julien Despois sets out here will put you on the path to making your dataset great.
The Most Important Skills for a Data Scientist, semanti.ca – In one of Reddit’s top machine learning articles of the year, semanti.ca have put together this great checklist of all the skills you’ll need to continue your growth as a data scientist into 2019.
How Teaching AI to be Curious Helps Machines Learn for Themselves, The Verge – Some games are more difficult for machines to beat than others. In a new feature, James Vincent explores a method that is proving more effective than reinforcement learning at beating Montezuma’s Revenge, its limitations, and the work that’s being done to improve it.
AI Keeps Mastering Games, But Can It Win in the Real World?, The Atlantic – It’s common knowledge that machine learning models like AlphaGo can play board games at a superhuman level. This article by The Atlantic dives into what we’ve learned from building machines to master these environments, and the challenges we face in applying that knowledge to the real world.
Through All the Hype, Self-Driving Cars Remain Elusive, The New York Times – Given the public furore around self-driving cars, balancing expectation and reality can be a difficult task. This piece by The New York Times does a great job at dealing with the hype in a measured way, before profiling some of the new, underlying players taking advantage of the upcoming shift in the way we drive.
AI Has Started Cleaning Up Facebook, but Can It Finish?, Wired – 2018 has been a difficult year for Facebook. Rocked by several major scandals, the need for stronger content moderation on the platform has become alarmingly apparent. Wired’s story explores the algorithms and human-in-the-loop systems currently dealing with Facebook’s content moderation, before detailing the problems they’re facing in trying to prevent the platform’s misuse.
How Cheap Labor Drives China’s A.I. Ambitions, The New York Times – Li Yuan profiles the workers and business owners behind China’s new ‘assembly line’ – the mass production of annotated data that is powering the country’s booming AI industry.
AI as Talent Scout: Unorthodox Hires, and Maybe Lower Pay, San Francisco Chronicle – In an interesting dive into recruitment, this piece explores how AI is being used to fill positions in tight labor markets, such as data science, and its potential effects on the way future recruiters will scout for talent.
Can We Trust AI if We Don’t Know How it Works?, BBC – This story unpacks neural networks, the millions of parameters behind them, and what they could mean for the sections of society that value clarity above all else.
Unbiased Algorithms can Still Be Problematic, TechCrunch – The effect of human bias on algorithms has been much discussed by AI journalists. However, this article suggests that the issues won’t simply go away once we’ve improved our training data. Through discussion with several experts, Megan Rose Dickey looks into some of these problems and why they’re so difficult to resolve.
Are you Scared Yet? Meet Norman, the Psychopathic AI, BBC – Through an unnerving experiment with inkblots, this article illustrates the potentially dramatic consequences of flawed data. Norman’s shocking interpretations of the pictures within demonstrate the unbreakable bond between the quality of your training data and the makeup of the model you create.
The Spooky Genius of Artificial Intelligence, The Atlantic – Derek Thompson’s article plays with the idea of what it means to be smart, using parallels between AI and the natural world to challenge conventional thinking about the limitations of machine learning.
As a final bonus, here’s our favorite article from Lionbridge AI this year. Happy reading, and stick with us for more content about AI, machine learning and training data in 2019!
The 50 Best Free Datasets for Machine Learning – Whether you need data for sentiment analysis, the financial market, or computer vision, our article is the perfect place to start your search.