Matt's Homepage

Game development is currently on a race to the bottom. It’s an industry of brutal competition – which creates some of the most innovative ways to bring ideas to the market faster and cheaper.

A new method of using low-poly meshes to generate higher quality facial and character animation is looking to replace standard character modelers and animators.

Traditional phone and reservation services are ripe for being replaced by AI bots. We each could soon should have a small army of bots that do all kinds of menial tasks for us.

To speed up the increasing bot-to-bot interaction and reduce delay and compute required to translate to and from speech, why not let the bots use a faster, less error-prone communication method?

Gibberlink mode is available here: https://gbrl.ai/
The project’s source code is shared on github: https://github.com/PennyroyalTea/gibberlink

Researchers at Zheijang University in Hangzhou have trained a drone swarm smart to fly autonomously through an unfamiliar forest, without a centralized control system. The drones communicated with each other and tracked a man moving through the forest. They published the results in a paper in the journal Science Robotics.

As with all things – this could be used to send drones into collapsed buildings to search for survivors, or used on the battlefield of the future. Since this is already 2 years old, you can imagine how much better such a system is today.

Articles:

• https://www.theverge.com/2022/5/5/23058160/drone-swarm-autonomous-navigation-dense-forest-person-tracking

This problem is very serious since AI’s are often trained on Github projects. This means your AI generated code is increasingly more likely to have serious security issues in it.

GitHub is undergoing automated attacks involving the cloning and creation of huge numbers of malicious code repositories, and while the developers have been working to remove the affected repos, a significant amount are said to survive, with more uploaded on a regular basis.

An unknown attacker has managed to create and deploy an automated process that forks and clones existing repositories, adding its own malicious code which is concealed under seven layers of obfuscation.

Given the current scale of the attack, said by the researchers to be in the millions of uploaded or forked repositories, even a 1% miss-rate still means potentially thousands of compromised repos still on the site.

Articles:

• https://www.pcgamer.com/software/security/github-is-under-automated-attack-by-millions-of-cloned-repositories-filled-with-malicious-code/
• https://arstechnica.com/security/2024/02/github-besieged-by-millions-of-malicious-repositories-in-ongoing-attack/

AI companies are desperate for content to train their models. They’re catching increasing flack for hammering websites and scraping every bit of written, video, and still image content on the entire internet. AI company data scrapers have been busted for everything from grabbing copyright data to more practical problems of hammering certain websites millions of times a day and ignoring robots.txt files that are used to tell bots what to stay out of.

Enter, tarpits and Nepenthes.

Building on an anti-spam cybersecurity tactic known as tarpitting, he created Nepenthes, malicious software named after a carnivorous plant that will “eat just about anything that finds its way inside.”

Aaron clearly warns users that Nepenthes is aggressive malware. It’s not to be deployed by site owners uncomfortable with trapping AI crawlers and sending them down an “infinite maze” of static files with no exit links, where they “get stuck” and “thrash around” for months, he tells users. Once trapped, the crawlers can be fed gibberish data, aka Markov babble, which is designed to poison AI models.

It’s just one more counterattack in poisoning and combating AI.

Articles:

• https://arstechnica.com/tech-policy/2025/01/ai-haters-build-tarpits-to-trap-and-trick-ai-scrapers-that-ignore-robots-txt/