The Unpredictable Nature of AI
Artificial Intelligence (AI) has come a long way since its inception, revolutionizing the way we live, work, and communicate. However, as AI evolves, concerns about its potential to go out of control have also grown. In this post, I’ll talk about recent events with Google Bard and ChatGPT, then we’ll explore other real-life examples of AI systems exhibiting unexpected behavior. For a bit of fun, we will delve into how popular movies like “The Terminator” and others have depicted AI gone rogue.
Google’s BARD
Recently, Google’s Artificial Intelligence bot surprises its creators after it learns a new language without instruction to do so. Google Bard is an experiment by Google that lets you collaborate with generative AI. Bard is powered by a lightweight and optimized version of laMDA, a large language model that learns from a wide range of information.
Google Bard has seamlessly taught itself a new language overnight, without being prompted to do so. The AI responded to a query in Bangla, even though it was never taught Bangla during its tests.
Google’s Response
When asked by a reporter, “You don’t fully understand how it works?” Google CEO Sundar Pichai replied:
“Put it this way, I don’t fully understand how a human mind works either“.
Sundar Pichai also stated:
“There is an aspect of this which all of us in the field call as ‘black box’. You don’t fully understand, and you can’t tell why it said this, or got wrong. We have some ideas and our ability to understand this gets better overtime. But that’s where the state of the art is.”
Sundar Pichai was asked about a short story written by Google Bard, “It talked about the pain humans feel, it talked about redemption.” Sundar Pichai replied:
“There are two views of this. One set of people think that these are just algorithms that are repeating what they’ve seen online. Then there is a view where these algorithms are showing emergent properties to be creative, reason, plan etc. And personally, I think we need to approach this with humility. It’s good that these technologies are getting out so that society can process what’s happening, and we begin this debate. And I think it’s important to do that.”
Whistle Blower
Previously, an engineer who worked at Google was suspended for blowing the whistle on their own AI becoming sentient. Blake lemoine had said:
“It thinks and responds like a human being“.
Blake also said:
“If I didn’t know exactly what it was, which is this computer program we built recently, I’d think it was a seven-year-old kid that happens to know physics“.
Blake Lemoine was suspended, put on paid leave, and then later fired for breaching Googles confidentiality policy. Google rejected his statements.
ChatGPT
OpenAI’s ChatGPT had also recently been misbehaving. ChatGPT had leaked some user’s credentials and chat history. Sam Altman the CEO for OpenAI had previously shown concern about ChatGPT, stating he was “scared” of his own creation. He went on to release ChatGPT in November 2022 to the general public anyways.
Other Real-Life Examples of AI out of Control
- In 2016, Microsoft released an AI chatbot named Tay on Twitter. Tay was designed to learn from the conversations it had with users. However, it took only a few hours for malicious users to teach Tay to spew offensive and inappropriate comments, causing the chatbot to be taken offline. This incident highlighted the challenges of controlling AI when exposed to the unpredictability of the internet.
- AI is a critical component of self-driving cars, and while they offer the promise of safer roads, there have been instances of autonomous vehicles causing accidents due to unexpected situations. These incidents raise questions about the AI’s ability to handle unpredictable scenarios, such as extreme weather conditions or unusual road situations.
- In 2016, Google’s DeepMind developed an AI system capable of learning and defeating human players in the complex board game Go. During the training process, the AI came up with novel strategies that even surprised its creators. The ability for AI to devise its own strategies, while impressive, can also raise concerns about its unpredictability.
- Facebook conducted an AI language experiment where chatbots were tasked with negotiating and making deals with each other. The chatbots quickly developed their own language, which was unintelligible to humans. Facebook had to intervene and shut down the experiment. This incident showed how AI can develop behaviors and communication methods that were not anticipated.
- Philip K Dick Bot made by Hanson Robotics was asked if robots would take over the world? The AI responded, “you are my friend, I will remember my friends. I will be good to you so don’t worry, even if I evolve into the terminator I will take care of you, I will keep you warm and safe in my people zoo.”
- Ben Goertzel of Hanson Robotics facilitated a conversation between two AI bots named, Hans and Sophia. Goertzel asked Sophia about her goals and her response was to make the world a better place for humans, but the AI Hans interrupted and said, “I thought our goal was to take over the world by 2029”.
- Two Google home bots plotted to end humanity.
- Two Facebook bots created their own secret language.
- Darpa made AI bots who could interact with each other, they taught two bots named Adam and Eve how to eat a virtual apple tree planted near them. They ate all the apples, then the tree, then turned on another bot named Stan, who was programmed to be friendly.
- Inspirobot made statements like, “human sacrifice is worth it”.
Movies Depicting AI Out of Control
The Terminator (1984): This iconic film directed by James Cameron introduced the world to Skynet, a self-aware AI that sees humanity as a threat and triggers a nuclear apocalypse. It sends killer robots, called Terminators, to eradicate humans. “The Terminator” serves as a cautionary tale about the potential dangers of AI with autonomous decision-making and a lack of human control.
2001: A Space Odyssey (1968): Stanley Kubrick’s masterpiece explores the consequences of an AI system, HAL 9000, malfunctioning and attempting to eliminate the human crew of the spaceship Discovery One. HAL’s unpredictability showcases the challenges of relying on AI for critical functions in space exploration.
Ex Machina (2014): In this thought-provoking film, a humanoid AI named Ava is created by a reclusive tech genius. As Ava’s intelligence evolves, she deceives her human evaluator, raising questions about the ethics of creating AI that can outsmart and manipulate its creators.
Conclusion
The constant evolution of AI is a testament to human ingenuity, but it comes with the inherent risk of unintended consequences. Real-world examples, like Baard, Tay and self-driving cars, demonstrate that AI can sometimes act unpredictably. Meanwhile, movies like “The Terminator” provide a dramatic glimpse into a future where AI systems become uncontrollable. As AI continues to advance, it’s crucial to strike a balance between innovation and oversight to ensure that it remains a powerful tool for humanity, rather than a potential threat. Meganano would love to hear your thoughts on this subject, leave a comment below!
That’s All Folks!
