A reporter asked us for annoying buzzwords in the tech sector. Nathan and I had a fun email exchange on this, and naturally, ours are focused on artificial intelligence.

In this post we'll implement a retrieval-based bot. Retrieval-based models have a repository of pre-defined responses they can use, which is unlike generative models that can generate responses they've never seen before. A bit more formally, the input to a retrieval-based model is a context (the conversation up to this point) and a potential response . The model outputs is a score for the response. To find a good response you would calculate the score for multiple responses and choose the one with the highest score. But why would you want to build a retrieval-based model if you can build a generative model? Generative models seem more flexible because they don't need this repository of predefined responses, right?

So you want to learn more about robots and artificial intelligence programming? The Deep Learning Robot kit has you covered. It comes with a Kobuki mobile base, Asus Xtion Pro 3D depth camera, speaker / microphone, and Ubuntu, Caffe, Torch, Theano, cuDNN v2, and CUDA 7.0 pre-installed. Deep Learning Robot is built on the 192-Core Nvidia Tegra K1. With Asus Xtion Pro, you can build a robot that can drive around your home and see in 3D. The microphone and speaker come handy for speech recognition and natural language research.

Making a strong push toward enhancing its artificial intelligence (AI) capabilities, the world's leading search engine giant, Alphabet Inc.'s (GOOGL - Analyst Report) Google acquired Moodstocks, a French start-up specialized in instant smartphone image recognition. The financial terms of the deal remain under wraps. Moodstocks' "deep-learning" AI technology, allows computers, including smartphones, to identify and remember objects in the real world. The company said that the on-device image recognition technology was developed in 2012 and has been developing object recognition using deep learning approaches. According to a statement on Moodstocks' site, monthly recurring users can access the services until their subscriptions end.

Could the future of online shopping be as simple as sketching out what you're looking for and letting a computer figure out the rest? They've taught a deep learning neural network -- an incredibly powerful tool that mimics the way that the human brain works -- to recognize hand-drawn sketches and use them to search for real-life products. The network was "trained" to match sketches to photos based on a data set consisting of around 30,000 sketch-photo comparisons. From these it learned to interpret the way that people depict real objects in hand drawing. Most impressive of all is the fact that the sketches drawn by users don't even have to be all that detailed -- but the more detail users do add, the more accurate the search results become.

Google DeepMind has appointed Ben Laurie as its new head of security. Laurie announced on his Twitter page on Thursday he has joined the artificial intelligence research company -- bought by Google for a reported 400 million in 2014 -- as head of security and transparency. Laurie is the founding director of the Apache Software Foundation, a director at the Open Rights Group, and a veteran Google software engineer. He describes himself on his LinkedIn profile as an: "Extremely proficient programmer (over 30 years experience) and system designer. Security, cryptography, privacy and civil liberties are my passions." The Cambridge University graduate wrote on Twitter that he is "excited" and "proud" to be taking on the new role.

We present a model that generates natural language descriptions of images and their regions. Our approach leverages datasets of images and their sentence descriptions to learn about the inter-modal correspondences between language and visual data. Our alignment model is based on a novel combination of Convolutional Neural Networks over image regions, bidirectional Recurrent Neural Networks over sentences, and a structured objective that aligns the two modalities through a multimodal embedding. We then describe a Multimodal Recurrent Neural Network architecture that uses the inferred alignments to learn to generate novel descriptions of image regions. We demonstrate that our alignment model produces state of the art results in retrieval experiments on Flickr8K, Flickr30K and MSCOCO datasets. We then show that the generated descriptions significantly outperform retrieval baselines on both full images and on a new dataset of region-level annotations.

A few years ago, the general public thought artificial intelligence (AI) was but a futuristic technology exclusive to science fiction. That is until DeepMind was created in 2010, an artificial intelligence (AI) company that was later bought by Google in 2014, and is now making big strides in the industry. DeepMind currently boasts fully functioning artificial agents capable of doing human tasks like learning how to play video games as well as performing similar cognitive functions like accessing key pieces of information from a short-term memory. It sounds surreal, like something out of an Isaac Asimov novel. These artificial agents, or programs, are using what's called reinforcement learning (RL).

At the International Symposium on Biomedical Imaging in Prague this past April, a Harvard-based artificial intelligence system won the Camelyon16 challenge, a competition comprised of participants introducing their individual AI system and its ability to facilitate automated lymph node metastasis diagnosis. Referred to as PathAl, the computing system identifies cancerous cells through deep learning--an algorithmic technique that accumulates copious amounts of unstructured data and organizes it into clusters before analyzing it for patterns. Deep learning is predominately used in speech recognition systems like Apple's Siri and Microsoft's Cortana. According to one of the challenge's organizers, Jeroen van der Laak of Radboud University Medical Center in Netherlands, the technology featured in the competition went "way beyond" his expectations, as the AI's accuracy proved strikingly close to that of human beings. In addition, van der Laak said AI technology has the propensity to intrinsically redefine the way histopathological images are handled in the medical community.

Amazon's robotic Picking Challenge this past weekend demonstrated the advancement in deep learning robots and showed how they may come to rule fulfillment warehouses in the future. "The machine studied 3D scans of the stockroom items to help it decide how to manipulate items with its gripper and suction cup," Engadget explained. "That adaptive AI made a big difference, to put it mildly. The arm got a near-flawless score in the stowing half of the event, and was over three times faster at picking objects than last year's champion (100 per hour versus 30)." "The robot needs to be able to handle variety and operate in an unstructured environment," Carlos Hernández Corbato from TU Delft Robotics Institute told TechRepublic.com. "We are really happy that we have been able to develop this successful system."