GPT-3 is an AI from OpenAI that generates content, using reasoning to refer to the information they fed it in order to create new written content. GPT-3 is short for “Generative Pre-Trained Transformer” and its 2nd and 3rd iteration has grabbed attention across the internet with headlines hinting at consciousness. The 3rd edition released in May 2020 and has 175 billion parameters for constructing text. The size of the data set it works from makes it two orders of magnitude more powerful than GPT-2.
What does it do
Working With AI: GPT-3 And The Future Of Content Marketing
San Francisco based OpenAI developed GPT-3, starting with GPT in 2018, and released it as an API that is still not fully available to the public. It’s task agnostic and requires minimal tuning in order to operate aside from trial-and-error training. This means its operation varies. It depends on what command it is given You can communicate all commands in human-readable, plain text.
What This Means for Humans:
A strong marketing strategy always includes creating content. In the United States, there’s a bias toward hiring content writers to write article content that’s easily accessible. GPT-3 promises high-quality text, but OpenAI strongly encourages hiring a human to edit the machine’s output. This may mean a shift in demand to increase for editors.
The initial hype around GPT-3’s release sounded like science fiction to some writers. But robots aren’t replacing human writers just yet. AI isn’t capable of replacing human creativity or the ability to generate emotional appeal. Content writing by machine is often formulaic, and there’s minimal variety in sentence type or length.
A New Tool
Humans have always learned to use new machines, especially those they can benefit from using. Writers are not going to be replaced. However, some content writing may shift to AI generation followed by human editing. The future will be interesting to witness as this technology develops, as the use of such a tool could increase the average writer’s output. The generative and summation tools can be used to generate content in virtually any field.
Impact on Marketing Industry
GPT-3 is limited in its ability to make sense of complicated information. For example financial data or business analytics. It’s excellent at creating content that is grammatically correct (but that’s not exactly revolutionary), however, it still produces errors in reasoning since it is unable to comprehend. This means that content writers aren’t going anywhere.
Companies may find themselves relying on editors and creatives to use the tool results in great content just like before. But opening roles for curating and developing the content created by the AI. This is a boon for the marketing industry as a content generator, but the quality of the content requires hiring of creative writers and editors to make it worth publishing.
What’s Wrong With Robotic Content?
If GPT-3 just needs to be edited for foul and biased content, where is skill involved? Part of the answer lies with more robots. Google’s robots, to be precise. Google’s SEO practices are studied religiously by content writers to ensure content that makes it to the front page. There are a lot of regulations and updates Google has enacted over the last decade to keep unique and creative content at the fingertips of those who use its search engine.
In contrast, you can easily detect GPT-3’s content. One of the most ironic things about GPT-3 is that it’s great at detecting its own content, even when a human might struggle.
How do I get it?
Currently, you will have to join a waitlist to get GPT-3. Potential users fill out a form including a field where they can describe how they would use the API. They also ask that you provide risks and benefits of using such a powerful AI.
Their safety-first mentality is not without merit. Critics of the GPT project cite cases where a mental health advocate program was tested. GPT-3 encouraged a patient’s desire for suicide. In most cases, especially for content intended for publication, editing by a human is encouraged. If your business staffs content writers, it may be worth applying for the waitlist for access to GPT-3.
Benefits of GPT-3
While custom article writing might not be the best use of GPT-3’s features, it does promise value for short-form content like meta descriptions, alt text, footer copy or product descriptions. There are also benefits to using GPT-3 to provide a first draft for content writers. These tools could encourage diversification for your staff writers. They could be providing the bandwidth for them to take on additional projects requiring research and time they don’t currently have.
What Doesn’t It Do?
The time savings of having a content generation machine are not stronger than the benefits of professionally written content. Especially when the content positions your business as a leading expert in your industry. Copywriting still requires a human touch as well, since emotional plays and convincing often require comprehension and reasoning abilities that are not yet programmable.
The Future of Content Marketing
At this time, most companies don’t have access to GPT-3. But as time goes on and we see more iterations of the GPT project with more lifelike results. Ethical implications around harmful bias mean OpenAI limits who has access to GPT-3 and provides useful guidelines to avoid harmful effects. You can read how and why OpenAI has chosen not to provide open-ended applications that would allow long-form text generation.
OpenAI is currently working with Middlebury Institute, University of Washington, and Allen Institute for AI to research applications focused on representation research and fairness in order to prevent safety issues.
They write that “Our goal is to continue to develop our understanding of the API’s potential harms in each context of use, and continually improve our tools and processes to help minimize them.”