DeeperDive, a new tool that converses with readers, is an effort to beat the AI industry at its own game. The publishing company behind USA Today and 220 other publications is today rolling out a chatbot -like tool called DeeperDive that can converse with readers, summarize insights from its journalism, and suggest new content from across its sites. "Visitors now have a trusted AI answer engine on our platform for anything they want to engage with, anything they want to ask," Mike Reed, CEO of Gannett and the USA Today Network, said at the WIRED AI Power Summit in New York, an event that brought together voices from the tech industry, politics, and the world of media. "and it is performing really great." Most publishers have a fraught relationship with AI, as the chatbots that trained on their content are now summarizing it and eating the traffic that search engines used to send them.

In this work we introduce an incremental learning framework for Click-Through-Rate (CTR) prediction and demonstrate its effectiveness for Taboola's massive-scale recommendation service. Our approach enables rapid capture of emerging trends through warm-starting from previously deployed models and fine tuning on "fresh" data only. Past knowledge is maintained via a teacher-student paradigm, where the teacher acts as a distillation technique, mitigating the catastrophic forgetting phenomenon. Our incremental learning framework enables significantly faster training and deployment cycles (x12 speedup). We demonstrate a consistent Revenue Per Mille (RPM) lift over multiple traffic segments and a significant CTR increase on newly introduced items.

Back in 2012, when neural networks regained popularity, people were excited about the possibility of training models without having to worry about feature engineering. Indeed, most of the earliest breakthroughs were in computer vision, in which raw pixels were used as input for networks. Soon enough it turned out that if you wanted to use textual data, clickstream data, or pretty much any data with categorical features, at some point you'd have to ask yourself -- how do I represent my categorical features as vectors that my network can work with? The most popular approach is embedding layers -- you add an extra layer to your network, which assigns a vector to each value of the categorical feature. During training the network learns the weights for the different layers, including those embeddings.

In one of my job interviews for a data scientist position, I was given a home assignment I'd like to share with you. The interviewer sent me a CSV file containing samples of measured quantities x and y, where y is a response variable which can be written as an explicit function of x. It is known that the technique used for measuring x is twice as better than that for measuring y in the sense of standard deviation. Here are all the imports I'll need: It clearly looks like linear regression case. First I'll manually remove the outliers: I'll use LinearRegression to fit the best line: If you're not familiar with the linear regression assumptions, you can read about it in the article Going Deeper into Regression Analysis with Assumptions, Plots & Solutions.

The common narrative of artificial intelligence is that it has finally taken off in recent years because there was enough data -- from mega repositories like Google -- and enough computing power through racks of servers equipped with fast processors and GPUs. That's not incorrect, but it's too simplistic to describe the future of machine learning and other forms of AI. That was the message from Intel's CTO of AI products, Amir Khosrowshahi, at VentureBeat's Transform 2018 conference outside San Francisco today. The challenge now is optimizing the whole process. Better algorithms require less computing and can draw accurate inferences from less data, said Khosrowshahi, cofounder of AI company Nervana Systems, which Intel acquired in August 2016.

Nvidia chips dominate the AI training chip market, where huge amounts of data help algorithms "learn" a task such how to recognize a human voice, but one of the biggest growth areas in the field will be deploying computers that implement the "learned" tasks. Intel dominates data centers where such tasks are likely to be carried out. "For the next 18 to 24 months, it's very hard to envision anyone challenging Nvidia on training," said Jon Bathgate, analyst and tech sector co-lead at Janus Henderson Investors. But Intel processors already are widely used for taking a trained artificial intelligence algorithm and putting it to use, for example by scanning incoming audio and translating that into text-based requests, what is called "inference." Intel's chips can still work just fine there, especially when paired with huge amounts of memory, said Bruno Fernandez-Ruiz, chief technology officer of Nexar Inc, an Israeli startup using smartphone cameras to try to prevent car collisions.