A few days ago, US Senator Chris Murphy tweeted: “ChatGPT taught itself how to do advanced chemistry. This was not incorporated into the model. No one programmed him to learn complicated chemistry. He decided to teach himself, then made his knowledge available to anyone who asked. Something is brewing. We are not ready. Worrying words from an important leader of one of the greatest countries in the world. This understandably scared people, and some wondered if we had finally developed Skynet and were going to be wiped out by machines. Fortunately, that’s not the case and probably never will be, because ChatGPT’s AI is designed as a tool.
RELIEF: Apparently ChatGPT fixed the ending of Game of Thrones
What is ChatGPT and how does it work?
ChatGPT is a language tool that uses AI technology to allow you to have a human-like conversation with the chatbot. It uses a Generative Adversarial Network (GAN), which is two software systems that work together to process data: the generator and the classifier.
The generator looks at a ridiculous amount of data before processing the AI results. Then the classifier uses its own input data to determine whether the output data makes sense.
RELIEF: Could ChatGPT put Google out of business?
What is the Turing test?
The Turing test determines whether a computer is intelligent, like a human being, by asking someone to communicate with a computer. If the person cannot tell that they are talking to a computer, then, theoretically, the computer has reached a human level of intelligence.
This test was developed by Alan Turing, one of the founding figures of computer science. When Turing developed his test, the idea behind it was that the machine should be able to understand what was being said, not just recognize key words and generate a response based on that. This means that, technically, a machine would have to have conscious thought to truly pass the Turing test.
This is not the case for Chat GPT, but it still passed the Turing test.
RELIEF: Will ChatGPT one day replace programmers? Many fear this is not possible
Why it doesn’t matter that ChatGPT passed the Turing test
Many chatbots can pass the Turing Test, but chatbots are single-use tools, not artificial minds with intelligence equal to that of a human. ChatGPT has become exceptionally good at imitating language patterns. But the same goes for parrots, and no one claims that a brightly colored naughty bird, amused by its reflection, understands the meaning of the words it repeats.
ChatGPT cannot make its own decisions, which would be the mark of human-level intelligence. It is, however, capable of following its developers' instructions, and that's exactly what happened when US Senator Murphy assumed that ChatGPT taught itself chemistry.
This does not mean that AI is becoming intelligent. All this means is that the Turing Test is no longer a valid test for artificial intelligence.
AI may be getting more sophisticated, but we're not heading toward a Skynet anytime soon. So you can stop stockpiling emergency bunkers for now.
RELATED: Adobe's new AI lets users type commands to edit images
TL;DR
- ChatGPT passed the Turing test, which freaked a few people out.
- Chatbot AI has not yet developed conscious thought or human-level intelligence; it has simply become more sophisticated than we thought.
- The Turing test is no longer a valid test for AI.