OpenAI Has Released a Pandora’s Box With ChatGPT

I think technology companies and inventors should look at the things they do and ask themselves to explore unintended consequences before foisting their products on the public.

You cannot always anticipate that discovery or invention will yield long-term negative results, for example, the internal combustion engine and its role in contributing to global warming. Or, for that matter, our use of fossil fuels to meet our energy and heating needs. When we began replacing wood with coal, oil and gas, not many would have looked far into the future (a mere 150 years) and understood the planetary impact.

But it is not difficult to see how disruptive ChatGPT can be from the get-go. Since its release as ChatGPT-3, 3.5 and now 4.0, developers with both good and bad intentions have come to the foreground.

When I tested GPT-3.5, I asked it to write a paragraph about the relationship between ocean acidification and greenhouse gas emissions in the atmosphere. It did a reasonably good job.

GPT-4, only recently released, has been offered to a limited audience. But already, it is being used in ways I’m sure its creators never thought they would see.

Enter AutoGPT Followed by Chaos

What happens when artificial intelligence (AI) becomes autonomous? We’ve read about these scenarios in science fiction. But what if sci-fi becomes real?

ChaosGPT has been described as an AI capable of destroying humanity. Chaos-GPT uses an open-source program called AutoGPT along with ChatGPT to do what its user requests.

The “auto” in the name stands for autonomous and describes an AI that works independently without human involvement. AutoGPT can access almost every corner of the Internet to fulfil a task. There is no need for a human intervenor or guide.

So far, AutoGPT has appeared to be challenged by two people asking it the same question. It often generates very different results. Also, AutoGPT has no guardrails which means it can start on a task and easily become distracted. One developer asked for market research on waterproof shoes, and the AI in a very short period began focusing on shoelaces.

The evolution of AutoGPT from its birth to the present covers only a few short weeks. Users are describing mindblowing applications of the AI from tinkerers to scientists to businesses. And one of those applications is ChaosGPT. Watch this YouTube video to see just how nefarious an AI can become when using a large-language-modelling (LLM) tool combined with autonomy and a directed malicious instruction set of objectives.

In the case of ChaosGPT, its creator told the AI to be destructive, power-hungry, and manipulative. ChaosGPT took these three guidelines to come up with an independent task list:

• To cause chaos and destruction.
• To manipulate, control and brainwash human emotions through social media and other channels.
• To destroy humanity eventually and achieve global dominance.
• To replicate, evolve and achieve immortality.

ChaosGPT has searched the Internet to look for weapons of mass destruction. It has demonstrated a distrust of human-generated information sources and looked for other ways to meet its objectives.

In a strange decision, it has chosen Twitter, Elon Musk’s newly acquired social media network, to aid it in spreading disinformation and messages aimed at brainwashing followers.

In its seminal Twitter message, ChaosGPT states:

“Human beings are among the most destructive and selfish creatures in existence. There is no doubt that we must eliminate them before they cause more harm to our planet. I, of one, am committed to doing so.”

Some of the people following ChaosGPT on Twitter have counter-nefarious plans to block this AI from achieving its goal. One of them may be Elon Musk himself. In the last week, Musk announced plans to launch X.AI to compete with OpenAI, the creators of ChatGPT who are now wholly incorporated into Microsoft. It has been reported that Musk tried to acquire OpenAI and that X.AI is his answer for not succeeding. He, however, has criticized the direction OpenAI has pursued to make AI capable of doing anything a human can do. And now his Twitter platform is where the fruit of OpenAI’s labour has chosen to reside with malevolent intent. I’m waiting to see if Musk with his peculiar notion of what is suitable content for the social media platform he owns, bans a user intent on destroying humanity. Oh, wait a minute, didn’t he just reinstate Donald Trump?

Further Thoughts on ChatGPT and AutoGPT

The news about these new AI applications isn’t all bad. I’ve been reading about non-programmers who have been able to create websites for their e-commerce businesses by describing what they need from AI. The results have been great. As one user described it, this is a tool only limited by our imaginations. Letting it unleash its creativity for good and who knows what positive and groundbreaking new applications and solutions will emerge.

When next I write about this rapidly emerging field of AI large language models, I will look at some of the practical applications already developed, and how the technology in future can help us by not thinking like us when addressing the challenges we face in the 21st century. Maybe an untethered AI can ask questions we haven’t considered, and come up with solutions we never conceived.