Self-supervised learning has been a powerful approach for learning meaningful representations from unlabeled data across various domains, reducing the reliance on large labeled datasets. Inspired by BERT's success in capturing deep bidirectional contexts in natural language processing, similar frameworks have been adapted to other modalities such as audio, with models like BEATs extending the bidirectional training paradigm to audio signals using vector quantization (VQ). However, these frameworks face challenges, notably their dependence on a single codebook for quantization, which may not capture the complex, multifaceted nature of signals. In addition, inefficiencies in codebook utilization lead to underutilized code vectors. To address these limitations, we introduce BRIDLE (Bidirectional Residual Quantization Interleaved Discrete Learning Encoder), a self-supervised encoder pretraining framework that incorporates residual quantization (RQ) into the bidirectional training process, and is generalized for pretraining with audio, image, and video. Using multiple hierarchical codebooks, RQ enables fine-grained discretization in the latent space, enhancing representation quality. BRIDLE involves an interleaved training procedure between the encoder and tokenizer. We evaluate BRIDLE on audio understanding tasks using classification benchmarks, achieving state-of-the-art results, and demonstrate competitive performance on image classification and video classification tasks, showing consistent improvements over traditional VQ methods in downstream performance.

In the winter of 2020, on one of my aimless, frigid quarantine walks around my silent neighborhood, I remember being struck by a thought: did a medieval European peasant know that he was living through what is now widely known as the Dark Ages? Was there some moment when he leaned against his hoe in the fields, gazed up at the uncaring sky, and dimly perceived that he was unlucky enough to have been born into a bad century, perhaps even a bad millennium, too late for classical antiquity and too early for the Renaissance? I was sympathetic toward that notional peasant, because I was feeling the same way. The tide of history was overwhelming; I was minuscule, my life brought to a terrifying standstill by an airborne virus. I thought that if the humans who survived into the year 2500 looked back on my era, they would see it as cursed or benighted, the beginning of a downward slide.

The Goldilocks theory of policy is simple enough. If Mummy Bear says your latest government bill is too hot, and Daddy Bear says your latest government bill is too cold, then you can tuck in knowing that the actual temperature is just right. Unfortunately, the Goldilocks theory sometimes fails. You learn that what you actually have in front of you is less a perfectly heated bowl of porridge and more a roast chicken you popped in the oven still frozen: frosty on the inside, burnt on the outside, and harmful to your health if you try to eat it. To its supporters, the online safety bill, which was dropped from the legislative calendar last Wednesday to make space for a no-confidence motion in the government, sits firmly in the Goldilocks zone. The bill is a monster piece of legislation, with its roots in a green paper published way back in October 2017.

Interspecies was once a technical term used in science to describe how one species got along with another. Now it is a word of more consequence: it evokes the new connections between humans and non-humans that are being made possible by technology. Whether it is satellite footage tracking geese at continental scale, or a smartphone video of squirrels in a park, people are seeing the 8.7m other species on the planet in new lights. In "Ways of Being", James Bridle, a British artist and technology writer, explores what this means for understanding the many non-human intelligences on Earth. Your browser does not support the audio element.

Futurism is a mug's game: if you're right, it seems banal; if you're wrong, you look like the founder of IBM, Thomas Watson, when he declared in 1943 that there is room in the world "for maybe five computers". David Adams knew these risks when he wrote about the future of technology in the Guardian in 2004 – even citing the very same prediction as an example of how they can go awry. And from our vantage point in 2020, Adams certainly did a better job than Watson. When he looked ahead to today, he avoided many of the pitfalls of technology prediction: no promises about flying cars nor sci-fi tech such as teleportation or faster-than-light travel. But in some ways, the predictions were overly pessimistic.

I like to think that while I may have misgivings about much of what the current technological revolution is visiting on us, I yet manage to resist that dread ascription "luddite". It's one Bridle also wishes to avoid; but such is the pessimism about the machines that informs his argument, that his calls for a new "partnership" between them and us seem like special pleading. As futile, in fact, as a weaver believing that by smashing a Jacquard loom he'll stop the industrial revolution in its tracks. If we're in ignorance of what our robots are doing, how can we know if we're being harmed? At the core of our thinking about new technology there lies, Bridle suggests, a dangerous fallacy: we both model our own minds on our understanding of computers, and believe they can solve all our problems – if, that is, we supply them with enough data, and make them fast enough to deliver real-time analyses.

Is it a silly prank, a Pagan ritual, or a genius discovery about the next era of mass transit? In a picture posted to Flickr by artist James Bridle--known for coining the term, "New Aesthetic"--a car is sitting in the middle of a parking lot has been surrounded by a magic salt circle. In the language of road markings, the dotted white lines on the outside say, "Come On In," but the solid white line on the inside says, "Do Not Cross." To the car's built-in cameras, these are indomitable laws of magic: Petrificus Totalus for autonomous automobiles. Captioned simply, "Autonomous Trap 001," the scene evokes a world of narratives involving the much-hyped technology of self-driving cars.

We spend a lot of time and words on what autonomous cars can do, but sometimes it's a more interesting question to ask what they can't do. That's what this little bit of performance art tells me, anyway. You can see the nature of "Autonomous trap 001" right away. One of the first and most important things a self-driving system will learn or be taught is how to interpret the markings on the road. This is the edge of a lane, this means it's for carpools only, and so on.

We derive criteria for training adaptive classifier networks to perform unsupervised data analysis. The first criterion turns a simple Gaussian classifier into a simple Gaussian mixture analyser. The second criterion, which is much more generally applicable, is based on mutual information.

We derive criteria for training adaptive classifier networks to perform unsupervised data analysis. The first criterion turns a simple Gaussian classifier into a simple Gaussian mixture analyser. The second criterion, which is much more generally applicable, is based on mutual information.