The world of generalized computing is an interesting place, and week after week there's always something new and exciting to talk about. I like to keep a general eye on what's happening in the world of tech, and some of it can be incredibly nerdy. This is a weekly column that I'll be writing called "Bit by Bit," talking about a mixture of topics that I find interesting and have my own thoughts and opinions on.

Counter-Strike: Neural Offensive

You could just play the real thing, though

counter-strike-2-games-like-valorant.jpg

I'm an avid Counter-Strike enjoyer, having competed within Ireland and Europe over the last decade. I started with Counter-Strike: Global Offensive, and I'm a huge fan of Counter-Strike 2. One story that caught my eye was one that sounds bizarre at first... and probably still sounds bizarre after explaining it.

DIAMOND (DIffusion As a Model Of eNvironment Dreams) is a diffusion world model that can generatively create and render an entire world for you, and one of the demo games shown was Counter-Strike: Global Offensive, on the iconic Dust 2 map. For Counter-Strike, the researchers collected the following data:

We use the 3.3 hour dataset (190k frames) of high-skill human gameplay, captured on the 'dust_2' map, which contains observations and actions (mouse and keyboard) captured at 16Hz. We use 2.6 hours (150k frames) for training and 0.7 hours (40k frames) for evaluation. We resize observations to 64×64 pixels, and use no augmentation.

This isn't the first time that Counter-Strike has been used for model training and environment learning, either. In 2022, the game was also used as a playground for behavioral cloning, where the model could be trained on movement from real players on actual servers, and an AI could then learn to mimic that and play like a real human. AI itself has a lot of uses, and developments like this can help further our understanding of using AI in real-world physical environments. It may not be helpful to society that an AI can traverse Dust 2, but it can help with developments that allow it to move about in the real-world in future.

In a sense, this is a very similar concept to DOOM being run in Stable Diffusion. The models basically recreate the game based on training data, and that can then be replicated.

Related

These are the best AI tools that I use to boost my productivity

Maybe you'll get some use out of them, too.

Large Language Models are destroying the internet

But maybe not in the way that you think

Source: Unsplash

If you feel like the internet is less helpful since the advent of ChatGPT, you were probably on to something. As it turns out, a recent research paper noticed a 25% decline in responses on Stack Overflow within six months of the tool's launch. In contrast, Chinese and Russian counterparts, where AI usage is limited, the equivalent websites were more or less unchanged.

In other words, it's very likely that generative AI like ChatGPT is contributing to the internet being less useful overall, and for a number of reasons. Not only has AI-generated content become incredibly pervasive, but people are also deterred from contributing to the internet and helping others. As the abstract of the paper states:

We find no significant change in post quality, measured by peer feedback, and observe similar decreases in content creation by more and less experienced users alike. Thus, LLMs are not only displacing duplicate, low-quality, or beginner-level content. Our findings suggest that the rapid adoption of LLMs reduces the production of public data needed to train them, with significant consequences.

What that means for the future of the internet isn't clear, but it's obviously not good. It's very likely a problem that will only become worse as time goes on, and looks like it may well be a catalyst to bringing the dead internet theory to reality... if it's not already.

You might be able to run Linux apps on your Android soon

Just like on Chromebooks

On Chromebooks, you can run Linux programs to use apps that you wouldn't normally be able to find on Chromebooks. While that makes a lot of sense for an operating system built primarily to be portable, there isn't so much of a lack of options on Android. There's basically an app for everything, and you can even virtualize Linux on your smartphone, too. However, a report from Android Authority seems to suggest that Linux app support will soon come to Android, with a new developer option that will enable a "Linux terminal app that runs inside the VM."

According to the report, Google's upgraded Terminal app will download and configure Debian to run in the VM, with the Terminal app seeming to be an all-in-one app for downloading, configuring, and running an instance of Debian inside of a VM. It was apparently tested on two devices; "tangorpro" and "komodo," the codenames for the Pixel Tablet and Pixel 9 Pro XL respectively.

If you're looking to run Linux apps on your smartphone, then you may be in luck in the future. It's possible the feature could arrive next year, or it could eventually be shelved. Still, if you're looking for ways to use old smartphones, then Linux apps are a great way to get new life out of them.

Related

Best Android phones in 2024

Need a new phone? Look no further than this selection of best Android phones you should buy, including the best from Google, Samsung and more!