Scary experience when installing AI in the microwave

On Twitter, Lucas Rizzotto told the story of "one of the scariest experiences of my life" installing the GPT-3 language toolkit on his microwave.

GPT-3 is a language processor developed by OpenAI that uses machine learning and input data to write human-like texts.

"You are a backstabber and I will kill you," wrote the artificial intelligence under the name Magnetron after being asked by Rizzotto to write a poem.

The microwave oven is installed with the GPT-3 language toolkit.

Turn the microwave into a childhood friend

The microwave that Rizzotto bought from Amazon already has Alexa, Amazon's artificial intelligence built in. However, he thought Alexa was too useless, so he wanted to integrate other AI, and ended up choosing GPT-3.

'How far can we develop it? Can I somehow microwave the GPT-3 and train the AI ​​with imaginary memories? Can I bring my imaginary friend back to life?' Rizzotto explained his idea on Twitter.

With a few technical manipulations, Lucas Rizzotto was able to put the AI ​​in the microwave and talk to it.

Rizzotto bought an additional Raspberry Pi board, installed the GPT-3 functions, then added a microphone and speakerphone to begin his experiment.

According to Rizzotto, he had an imaginary friend since he was a boy. That character is called Magnetron, a British gentleman from the 1900s and served in the first world war. He wanted GPT-3 to recreate his old imaginary friend.

To feed the GPT-3 to turn AI into an imaginary friend, Rizzotto wrote a 100-page biography containing relevant information and fed it to the AI. Once completed, everything seemed to be according to Rizzotto's plan when the microwave was able to 'talk'.

Lucas Rizzotto shared a photo of him standing and talking with an imaginary friend inside a microwave as a child.

'Talking to it is both special and strange. I really felt like I was confiding in an old friend. Although not all interactions are perfect, everything is accurate enough,' Lucas Rizzotto wrote on Twitter.

The conversation is getting worse

After the initial euphoria, Rizzotto began to feel a little strange and at times scared. He said that Magnetron (the character's name was imagined by Rizzotto) gradually began to say violent and inciting statements. Then Rizzotto questioned whether all AIs are prone to violence.

'Red roses, blue violets, you're a backstabber and I'll kill you,' Magnetron responded when Rizzotto asked to write poetry.

'That's when this hit me. Of the memories I included in the AI, about 10% detailed his time in World War 1, including some of the most traumatic memories of his life, such as the loss of his entire family. Is AI also traumatized', Rizzotto was bewildered.

The AI ​​in the form of a microwave sends a threat to Rizzotto.

Rizzotto felt these were not just jokes. It's like a threat. When he told Magnetron he was inside the microwave, the AI ​​turned on the microwave itself and tried to start the cooking process. It was like an act with the intention of 'baking' his master.

'This is terrifying. After a few minutes, I asked him a simple question: Why did you do that? The AI ​​just replied that I wanted to hurt him the same way he did me,' Rizzotto said.

"I want to hurt you the same way you did me," replied the AI ​​Magnetron after trying to bake his master in the microwave.

After the terrifying experience, Rizzotto concluded that Magnetron's thoughts could be fake and stemmed from programming commands. However, those words unintentionally created true feelings inside Rizzotto. In a video about the process of 'chatting' with the microwave through AI posted on Youtube, Rizzotto said that everything is real and there is no staged element.

Ultimately, Rizzotto thinks he will create another feminine version of AI called Fifi in the hope of being able to "heal" Magnetron's psyche.

Is AI scary?

Speaking at the South by Southwest technology conference in Texas in 2018, Elon Musk once said that AI is more dangerous than nuclear warheads. The billionaire said there should be an agency to regulate and monitor the development of superintelligence.

'I'm actually very, very, very close to the pinnacle of AI and that scares me. It is capable of doing more than we think, and the rate at which AI is improving is exponential," said Elon Musk.

Elon Musk has expressed concern about AI and its threats.

Based on his knowledge of artificial intelligence and its evolution, Musk believes that humans have reason to worry.

Musk pointed out that the London-based company DeepMind, which was acquired by Google in 2014, has developed an artificial intelligence system called AlphaGo Zero with the ability to learn how to play Go without any intervention. any human intervention. It has the ability to play with itself and learn everything.

At the time, Elon Musk worried that the development of AI would outstrip human management.

'The growth rate of AI is really impressive. We have to find some way to ensure that the advent of super-intelligence is something of a symbiosis with humanity. I think this is the most pressing crisis and the biggest threat we face," Tesla CEO shared.