Looking back on 2021, the story that might have the biggest future impact on writers is emergence of large natural-language computer models – like GPT-3 from OpenAI - that learn to write.
GPT-3, currently the largest and most
literate, has been trained with access to most of the internet and thousands
of books. This is a major step forward toward AI being able to both understand
and interact with the world.
So why hasn’t GPT-3 put writers out of business?
- At this stage GPT-3 can only mimic text written by a human.
- Although the text GPT-3 generates can be impressive, it doesn’t
understand what it is writing and much of what it produces is unusable. It can write like a human, but it can't think like a one.
- It doesn't have the capability to be concerned about the tone of its writing or how specific audiences will perceive it.
GPT-3, like all current neural networks, lacks the ability to explain/interpret why certain inputs result in specific outputs.
It has been pre-trained. It does not have an ongoing long-term memory that learns from each interaction.
- The majority of its training comes from the internet, which
is rampant with bias and misinformation.
- It’s a resource hog, requiring an extraordinary amount of computation power, data, and money to train.
So, is GPT-3 going to replace writers in 2022?
I believe it will become an incredibly useful tool for writers -- sort of like Photoshop is a tool for designers -- but replace? No
At least not in 2022.
But don't ignore it.
In fact, you should learn how to use it.
'Cause it's not going anywhere.
Make sure you're not either.
No comments:
Post a Comment