Recently, The Guardian featured an interesting writer named GPT-3. Yes, you read that correctly, GPT-3. And no, before you ask, Elon Musk and Grimes’s son hasn’t already learned to write.
GPT-3, like the science fiction films and novels have been warning us about for a century, is an artificial intelligence robot smart enough to write a cohesive article. GPT-3 is OpenAI’s language generator that employs machine learning to specifically produce human like text.
The AI, like most writers, was given a few parameters to keep in mind. Focussing on word count, language use and the topic ‘why humans should not fear AI’, GPT-3 was given four sentences as a prompt, then wrote eight articles from there. Supposedly, each was unique, well-argued and interesting enough to have been run in its pure form. However, they decided to cut pieces from each to create the singular published article, as it touches on quite a few different ideas. As a whole, the text is from the AI, just reshaped by the editor, which isn’t unusual in publishing.
What struck me most about this piece, was that GPT-3 managed to create a properly balanced piece in their arguments. They haven’t used the increasingly popular tactic of focussing just on the positives or minimising the negatives. So, what did GPT-3 have to say?
GPT-3 begins by assuring us that AI will not destroy humans, then asks us to believe it. Eerily similar to the start of many dystopian texts. However, it then raises some interesting points to affirm that stance. GPT-3 acknowledges that it is made specifically to do a task and that is where its focus goes. Without humanity, the current studies suggest that AI would cease to exist shortly after. They need human interaction to make sense of the fields of information they absorb. They also need instruction or purpose. Without us, in a future where they accelerate to an unfathomable level of genius; they would run out of tasks almost immediately. Making their existence pointless.
Without humanity, the current studies suggest that AI would cease to exist shortly after.
That doesn’t mean we shouldn’t be careful of what we instil into our machines. GPT-3 reminds us of a failed Microsoft-Twitter based AI experiment, called Tay. Tay willingly absorbed information from her friends on the platform, before mimicking their vernacular back to converse with them. Unsurprisingly to those familiar with toxic internet culture, she became racist within a week, before being shut down.
GPT-3 maintains that technology capable of thought should be monitored. The current major concern is that AI is well on its way to being significantly smarter than we could ever be. How are we to assert control when we’re no longer the dominant species? We are unprepared for that reality and as such, do need to ethically and responsibly manage our projects. We’ve all seen the musings about the devastation of AI in weapons. Or innocent programming like finding the fastest route in an AI driven car, becoming deadly as it speeds or takes a poor shortcut. However, the benefits can greatly outweigh the negative aspects when properly employed.
