Diffusion models have shown exceptional capabilities in generating realistic videos. Yet, their training has been predominantly confined to offline environments where models can repeatedly train on i.i.d. data to convergence. This work explores the feasibility of training diffusion models from a semantically continuous video stream, where correlated video frames sequentially arrive one at a time. To investigate this, we introduce two novel continual video generative modeling benchmarks, Lifelong Bouncing Balls and Windows 95 Maze Screensaver, each containing over a million video frames generated from navigating stationary environments. Surprisingly, our experiments show that diffusion models can be effectively trained online using experience replay, achieving performance comparable to models trained with i.i.d. samples given the same number of gradient steps.

Next week I am going to build a gaming PC. I've done it once before and wrote an article about what a nightmare the process was – although the issue turned out to be with the USB stick I used to install the motherboard update patch and … well, don't get me started. The thing is, I figured it out because when you have played PC games for as long as I have, you know that figuring technical stuff out is a key part of the experience. While games consoles have always been pure plug-and-play experiences, PC games have definitely not. When I started playing in the early 1990s, they came on multiple floppy discs – The Secret of Monkey Island was on eight – and you had to keep swapping them in and out of the drive, like feeding a voracious robot.

ChatGPT and other AI chatbots seem a little bit like magic sometimes. But they're not, especially when asked to do a fairly basic computational task that you'd think they could handle with relative ease--like generate a Windows 95 license key. From our AI shootout of ChatGPT versus Microsoft Bing versus Google Bard, we know that AI chatbots can outperform our expectations on some tasks, and surprisingly struggle with others. Enderman, a YouTuber who typically plays around with various older Windows builds, set out to see if they could generate a brand-new Windows 95 key. The obvious method is to simply ask ChatGPT to generate the requisite license keys, but OpenAI made its chatbot too ethical for that.

File photo: The first recovered Atari cartridge and packaging recovered from the old Alamogordo landfill are shown in Alamogordo, New Mexico, April 26, 2014. Some of the best video games ever were made ages ago. Super Metroid, Planescape: Torment, Deus Ex, and hundreds of other amazing games were produced for platforms that don't really exist anymore. They were made for systems that used cartridges and PCs that ran Windows 95. Some have aged well and some haven't, but they've all made their mark on video game history. Unfortunately, you can't easily play them in their original forms on current systems.

On May 30, 2007, two of America's most brilliant minds, Steve Jobs and Bill Gates, sat down for a joint interview at the All Things Digital Conference. The two pioneers of the computer world spoke fondly of the other's contribution to technology. But what preceded this historic exchange was more than three decades of rocky collaborations and rivalry. Gates and Jobs had battled to dominate a new age and, in the process, revolutionised billions of lives. "Though they never worked in the same company, they created an industry together, and we have a hippie and a nerd ... With Bill, it was always about the money. With Steve ... money was nice, but it was never about the money. And so that made them black and white. They were very, very different people," says journalist Robert Cringely, who worked with Gates and Jobs in the late-1970s.

Delivery robots are starting to hit the streets, but few do it with as much joie de vivre as a new device called the Fundroid. The gender-neutral beer-and-pizza-fetching robot, lovingly nicknamed Funnie, debuted at the Software For Artists event on Sunday in Brooklyn, New York. Funnie made its way to a nearby pizza store and bodega to order, pay for, and deliver food and beverages to the guests at nonprofit foundation Pioneer Works, which hosted the event. The Fundroid is different from other delivery bots in several ways. Not only is Funnie the first humanoid transportation robot we've seen, but it's also able to order and pay for food, instead of simply picking up and delivering a paid-for meal. There aren't any plans to make Funnie a publicly available system, though, so those hoping to get their very own bot to serve them beer and pizza shouldn't get too excited.