REV3: We will clarify Tab.1 in the paper and add derivations of Eqs. 5, 6. DESIRE[21] is variational. The second is an online planning algorithm (Shooting), where Ego's future action sequences are38 optimized tomaximize fortheplanning rewardunderthelearned MFP dynamics model.

Stories about artificial intelligence have been with us for decades, even centuries. In some, the robots serve humanity as cheerful helpers or soulful lovers. In others, the machines eclipse their human makers and try to wipe us out. "The Creator," a sci-fi film that hits theaters Friday, turns that narrative around: The United States is intent on wiping out a society of androids in Asia, afraid the artificially intelligent beings threaten human survival. Do any of these stories reflect our real-life future?

Pre-trained Language Models such as BERT are impressive machines with the ability to memorize, possibly generalized learning examples. We present here a small, focused contribution to the analysis of the interplay between memorization and performance of BERT in downstream tasks. We propose PreCog, a measure for evaluating memorization from pre-training, and we analyze its correlation with the BERT's performance. Our experiments show that highly memorized examples are better classified, suggesting memorization is an essential key to success for BERT.

Tom Cruise runs in a scene from the film'Minority Report', 2002. Back in 2002, the science fiction film Minority Report once again reignited futuristic imaginations about a world and police state gone too far. At the time, the movie inspired plenty of speculation about the future of our society, how computers would interact with us, and how law enforcement would be carried out proactively based on intent. In the movie, they combined technology with the psychic abilities of the "precogs," to proactively prevent crimes. The precogs had the ability to predict when crimes were about to be committed ahead of time, enabling law enforcement to act early.

It may sound like the plot of the 2002 movie Minority Report, but artificial intelligence can predict any arrest within three years of a prisoner being released on parole. The machine learning was designed to determine the risk of releasing a prisoner early by analyzing 91 variables, including age, race and previous arrests. Scientists from The University of California, Davis (UC Davis) used the data of more than 19,000 inmates scheduled with the New York State Parole Board from 2012 to 2015. Court documents show 4,168 individuals were released, but that AI determined the board could have released double the inmates without increasing the subsequent arrest rate. The film, set in 2054, is about a specialized police department that apprehends criminals using foreknowledge provided by three psychics called'precogs.'

A memory leak in an application deployed on the cloud can affect the availability and reliability of the application. Therefore, to identify and ultimately resolve it quickly is highly important. However, in the production environment running on the cloud, memory leak detection is a challenge without the knowledge of the application or its internal object allocation details. This paper addresses this challenge of online detection of memory leaks in cloud-based infrastructure without having any internal application knowledge by introducing a novel machine learning based algorithm Precog. This algorithm solely uses one metric i.e the system's memory utilization on which the application is deployed for the detection of a memory leak. The developed algorithm's accuracy was tested on 60 virtual machines manually labeled memory utilization data provided by our industry partner Huawei Munich Research Center and it was found that the proposed algorithm achieves the accuracy score of 85\% with less than half a second prediction time per virtual machine.

An intelligent agent operating in the real-world must balance achieving its goal with maintaining the safety and comfort of not only itself, but also other participants within the surrounding scene. This requires jointly reasoning about the behavior of other actors while deciding its own actions as these two processes are inherently intertwined - a vehicle will yield to us if we decide to proceed first at the intersection but will proceed first if we decide to yield. However, this is not captured in most self-driving pipelines, where planning follows prediction. In this paper we propose a novel data-driven, reactive planning objective which allows a self-driving vehicle to jointly reason about its own plans as well as how other actors will react to them. We formulate the problem as an energy-based deep structured model that is learned from observational data and encodes both the planning and prediction problems. Through simulations based on both real-world driving and synthetically generated dense traffic, we demonstrate that our reactive model outperforms a non-reactive variant in successfully completing highly complex maneuvers (lane merges/turns in traffic) faster, without trading off collision rate.

For autonomous vehicles (AVs) to behave appropriately on roads populated by human-driven vehicles, they must be able to reason about the uncertain intentions and decisions of other drivers from rich perceptual information. Towards these capabilities, we present a probabilistic forecasting model of future interactions of multiple agents. We perform both standard forecasting and conditional forecasting with respect to the AV's goals. Conditional forecasting reasons about how all agents will likely respond to specific decisions of a controlled agent. We train our model on real and simulated data to forecast vehicle trajectories given past positions and LIDAR. Our evaluation shows that our model is substantially more accurate in multi-agent driving scenarios compared to existing state-of-the-art. Beyond its general ability to perform conditional forecasting queries, we show that our model's predictions of all agents improve when conditioned on knowledge of the AV's intentions, further illustrating its capability to model agent interactions.

Samsung's AI assistant, Bixby, is getting an update on the shiny-new Galaxy S10 with Bixby Routines, a feature that learns your habits to preemptively launch apps or settings when you're most likely to need them. For instance, getting into your car could automatically trigger Spotify or turn on Do Not Disturb mode, while heading to bed could boot up battery-saving modes. Bixby Routines will work best once the software has a chance to learn your habits, so we weren't able to test it for our hands-on preview of the S10 and S10 . The phone that learns your habits to blend seamlessly into your day. Bixby has historically trailed behind competitors including Google, Amazon and Apple when it comes to virtual assistant technology.

It's 6:15 a.m. and my alarm clock is ringing, but 15 minutes ahead of schedule. My clock is always connected with Smart, my virtual assistant, an artificial intelligence software that autonomously manages my calendar (among other tasks) and from which it has learned that I will need to catch a flight out of town this morning. Smart has also checked the morning traffic, received from the V2I (vehicle-to-infrastructure sensors installed in 2020 on primary communication routes), which notes some delays along the route that I would normally take to the airport. Welcome to the future of AI. Of course, before waking me, Smart checked to make sure that the flight would be on time.