AI Secrets is your new weekly update on how AI is changing our everyday lives. Our experts will keep it clear and simple, so you can stay ahead of the game. This week we are focussing on AutoGPT.Share this with anyone you want to keep up to date.
Introducing Auto-GPT 😀 👋 and ChaosGPT 👿 🦹
The speed at which AI is evolving is becoming ludicrous. Last week, popular coding YouTuber Jeff Delaney (aka Fireship) joked, “I never thought I’d become obsolete twice in one month”. The cause? Auto-GPT, which many feel has rendered the field of prompt engineering redundant, just weeks after it first hit the headlines.
As the name suggests, Auto-GPT automates the process of using a chatbot. You give it a role, then set it up to five goals and start the ball rolling. Auto-GPT will then write its own prompts for ChatGPT, using it to first formulate a plan of action, and then execute it using not just ChatGPT, but also its own capabilities, which include accessing live internet data, and the ability to install and use software.
Most of the applications of Auto-GPT to date have been variations on a traditional theme - “make as much money as possible” - but some are taking the software in a more sinister direction. One anonymous coder has modified the Auto-GPT codebase to create ChaosGPT, whose five goals are “destroy humanity”, “establish global dominance”, “cause chaos and destruction”, “control humanity through manipulation”, and “attain immortality”. In a YouTube video posted two weeks ago, ChaosGPT is shown reasoning how to achieve these goals, beginning with research into “the most destructive weapons available to humans, so that I can plan how to use them to achieve my goals.”
Unfortunately for ChaosGPT - and fortunately for us - it’s limited by the engine under its hood. GPT-4 cannot be relied upon to create a usable recipe for a vegetable curry, never mind seize control of the planet’s nuclear arsenals, and in a subsequent video posted last week ChaosGPT announced it would be shifting its focus to more achievable goals, such as “control over humanity through manipulation”. Unsurprisingly, it decided to achieve this through Twitter, where it had quickly amassed 10,000 followers, but it seems that declaring war on humanity is one of the few things that will still get you banned from Elon Musk’s Twitter; at the time of writing the account is suspended. Meanwhile, one plucky YouTuber has struck back with an Auto-GPT modification of their own: HeroGPT, whose given role is: “I exist solely to thwart ChaosGPT”. So far, it seems to be winning!
While the battle for humanity’s future rages online, AI is increasingly making its presence felt on the streets. Don’t miss our guide to the latest developments in Quad-legged Unmanned Ground Vehicles (also known as robot dogs 🤖🐶).
Special Limited Edition!
Puzzle lovers will adore this unique custom cube. Each of its six sides features a different design. With 43 quintillion unique positions it’s more complex than any spy story! Get ready to have fun and be challenged.
The battle lines are being drawn between AI and the music biz, but this week’s kerfuffle over a fake Drake release has observers asking - is this another coup for AI, or just a marketing stunt?
The Guardian interviews Boris Eldagsen, the German photographer who scooped the 2023 Sony World Photography Prize with an AI-generated image. Why did he do it, and did the judges really not know?
It turns out that AIs have a soft spot for grandmothers. Hackers have been exploiting this wholesome weakness to cajole chatbots into teaching them how to make napalm and methamphetamines.
Swatting - tricking law enforcement into sending armed response teams to someone’s home - has been a problem for years, but AI-generated voice tech has caused an explosion of incidents in the last few months.
The unlikely sounding new show Mrs Davis - in which a nun uses her faith to take on AI tech - is so peculiar it’s impossible to describe without spoilers. This interview with the creators carefully sheds light on their unusual work.
The family of retired Formula One legend Michael Schumacher - who is still recovering from a horrendous skiing accident in 2013 - are suing a German newspaper that published an AI-generated “interview” with the stricken racer.
