Wearable-based Human Activity Recognition (HAR) is a key task in human-centric machine learning due to its fundamental understanding of human behaviours. Due to the dynamic nature of human behaviours, continual learning promises HAR systems that are tailored to users' needs. However, because of the difficulty in collecting labelled data with wearable sensors, existing approaches that focus on supervised continual learning have limited applicability, while unsupervised continual learning methods only handle representation learning while delaying classifier training to a later stage. This work explores the adoption and adaptation of CaSSLe, a continual self-supervised learning model, and Kaizen, a semi-supervised continual learning model that balances representation learning and down-stream classification, for the task of wearable-based HAR. These schemes re-purpose contrastive learning for knowledge retention and, Kaizen combines that with self-training in a unified scheme that can leverage unlabelled and labelled data for continual learning. In addition to comparing state-of-the-art self-supervised continual learning schemes, we further investigated the importance of different loss terms and explored the trade-off between knowledge retention and learning from new tasks. In particular, our extensive evaluation demonstrated that the use of a weighting factor that reflects the ratio between learned and new classes achieves the best overall trade-off in continual learning.

Self-supervised learning (SSL) has shown remarkable performance in computer vision tasks when trained offline. However, in a Continual Learning (CL) scenario where new data is introduced progressively, models still suffer from catastrophic forgetting. Retraining a model from scratch to adapt to newly generated data is time-consuming and inefficient. Previous approaches suggested re-purposing self-supervised objectives with knowledge distillation to mitigate forgetting across tasks, assuming that labels from all tasks are available during fine-tuning. In this paper, we generalize self-supervised continual learning in a practical setting where available labels can be leveraged in any step of the SSL process. With an increasing number of continual tasks, this offers more flexibility in the pre-training and fine-tuning phases. With Kaizen, we introduce a training architecture that is able to mitigate catastrophic forgetting for both the feature extractor and classifier with a carefully designed loss function. By using a set of comprehensive evaluation metrics reflecting different aspects of continual learning, we demonstrated that Kaizen significantly outperforms previous SSL models in competitive vision benchmarks, with up to 16.5% accuracy improvement on split CIFAR-100. Kaizen is able to balance the trade-off between knowledge retention and learning from new data with an end-to-end model, paving the way for practical deployment of continual learning systems.

The relatively unheard-of studio Ocelot Society did the impossible with Event0: they made a chatbot interesting. Set within the standard but comforting framework of a claustrophobic spaceship exploration game, Event0 is really about building a relationship between the player and the ship-bound AI, a chatbot named Kaizen. Being able to see the limitations of Kaizen's technology (basically the same stuff that's been used since we were all tooling around with AIM bots a decade ago) doesn't stop him from being a convincing and sympathetic companion as you try to figure out how to return to Earth. Kaizen may not be trustworthy, but that just makes him all the more bewitching.

The central hook to Event[0] is a character named Kaizen, a chatbot/AI who is surprisingly good at casual conversation. Other than you, the protagonist, Kaizen is literally the only other "living" thing in the game. There are other characters, sure, but they only exist in the past through chatlogs and recordings. Kaizen is all we have for companionship, so it's a damn good thing that the AI is such a compelling companion. You're originally part of a mission to Europa, but something goes wrong along the way, and you end up in an escape capsule, drifting through space. Eventually you drift your way to the Nautilus, a prototype ship now devoid of all life save for Kaizen, the caretaker AI.

As I drift through the lobby of the Nautilus, a derelict spaceship I've found myself stranded on, the shipboard AI starts playing music. I see a terminal across from me in the room. I go to it, and Event[0], out now for PC, starts revealing itself to me. Developed by an independent team called Ocelot Society, Event[0] wants you to talk to it. The terminal has a real-time typing interface and a conversation partner in the form of that shipboard AI.

Event[0] ReviewPrice: 14.99 Developer: Ocelot Society Publisher: Ocelot Society Platform: PC Hidden inside one of Event[0]'s chunky CRT terminals is an event log in which two crewmembers of the Nautilus space yacht discuss living with an Artificial Intelligence. One of the crew mentions being lonely on account of having nobody but a stupid AI to talk to. The other remarks that an AI doesn't have to be intelligent to be good company, it only needs to appear human. It's an astute self-observation by developers Ocelot Society, and gets to the nub of why Event[0] succeeds where so many other games fail. In gaming, AI isn't so much about actual intelligence as it is creating a convincing illusion of intelligence, which is often achieved through means other than algorithms and machine learning.

I'm aboard the Nautilus, a derelict vessel in orbit around Jupiter's frozen moon of Europa, and the only thing I've come in contact with so far is the ship's artificial intelligence. It even has a friendly name--Kaizen, which translated into English means "Change for better." Kaizen's waiting for me to respond, there in the friendly blink of a command prompt on a terminal straight out of the 1970s. But I've seen plenty of science fiction films, and so I still hold my breath as I ask Kaizen to open the airlock doors, prepared for an "I can't do that, Dave," and the slow strains of "The Blue Danube" to accompany my floating corpse into the blackness of space. Kubrick's classic science fiction film is a clear inspiration, seen in everything from the pseudo-'70s retrofuturism of the Nautilus to its all-seeing security cameras--black, with a cherry dot in the center.

One of the big games coming out this week is Event[0]--available for Windows and Mac on September 14th. It's a sci-fi game set in an alternate retrofuture reality in which humanity built a starship in 1985 and has since embraced artificial intelligence even more than we have now. The events depicted in the game take place in 2012, on board a starship, where you, the player, are left alone with an insecure AI known as Kaizen. The idea is to explore the ship (which requires completing some hacking puzzles) in order to gather items and information. With this, you can then talk to Kaizen through the terminals located across the ship, in order to make it trust you a little more.

Funny, strange, sometimes alarming things can occur when you chat up one of those artificial intelligence bots capable of convincingly chatting back. Look no further than Cleverbot. It's all right there in the parental-advice disclaimer: "things it says may seem inappropriate"; "whatever it says, visitors never talk to a human"; "use with discretion and at YOUR OWN RISK." These web apps can be amusing to screw around with, but what if natural-language AI was baked into a full-fledged video game? That's what Ocelot Society is going for with its upcoming narrative exploration game Event[0], which is now set for a September 14, 2016 release on Steam.

Event[0] is a game about a stranded astronaut talking to an artificial intelligence to help them get back to earth. Like Firewatch before it, much of this solitary adventure is centered around conversing with your colleague. Unlike Firewatch, your colleague is a computer recalling 2001's HAL. Also unlike Firewatch, you get to manually type in the questions you'd like to ask. There's no prescribed dialogue trees here, folks.