GPT-3: Top use cases & their likelihood
GPT-3, also standing for Generative Pre-trained Transformer 3, is one of the latest AI models by OpenAI. Many say that it will revolutionize the world of Artificial Intelligence, and while this is a slight over-exenteration, the truth is that GPT-3 indeed does poses some impressive abilities.
In this week’s article, we’ll take a look at some of the most promising features of GPT-3 and how likely they are to persist in the future.
#1. Code assistant
Recently, Microsoft demonstrated the use of GPT-3 for generating Power FX formulas. In this particular use-case, GPT-3 was able to transform natural language into a database-selection query:
There are several other examples on the internet where GPT-3 has been used to write code of one type or another. The AI even managed to create another AI for a computer vision task:
Software engineers know that writing the code for design is one of the most tedious tasks out there. GPT-3 has been tested in this scenario as well:
The Limitation
While the above-mentioned code examples certainly look impressive, GPT-3 has been shown to not be very effective at solving more complex tasks — mainly because our “natural language” is full of ambiguity, unlike a structured programming language. Describing complex tasks with a language that can be interpreted in multiple ways (our “natural language”), not surprisingly, introduces a lot of uncertainty & often inaccurate results.
As it stands, GPT-3 will not be replacing the competent programmer in the near future.
The Future
Not perfect at completing complex tasks does not mean unable to complete any tasks. The demanding job of a programmer is here to stay, but GPT-3 might be able to make it a little bit easier.
While the model fails to produce longer & more complex pieces of code, this does not apply to design & basic algorithmic challenges. This might not seem like much, but layout & design work are among the biggest time-consumers when it comes to product creation.
One of the best GPT-3 uses is a code assistant tool: able to write short pieces of code for independent tasks in design & algorithms. In fact, GitHub & OpenAI are already working on just that — The Copilot!
#2. Automatic Writer
One of the areas to which generative pre-trained transformers are applied is language generation. Essentially, given a topic and a desired length, GPT-3 is able to create a cohesive piece of text.
For example, GPT-3 wrote a #1 blog in Hackernoon, generated clever answers to difficult physics questions, and even wrote a short philosophical essay by itself. Here’s a snippet from it:
Consider that what you’re reading above has been entirely generated by a computer program! Safe to say, this technology has a lot of potential.
The Limitation
Generating basic philosophical essays might be impressive to me & you, but college Professors do not regard it as highly. In fact, GPT-3 only managed to score an average of C while writing university-level papers. Pretty good for a computer program, but still far from optimal. Safe to say that GPT-3 won’t be writing your assignments any time soon.
The Future
In a text-generating sense, GPT-3 can be a good starting point. Real-world text-related tasks are by all likelihood too complicated for optimal performance, but GPT can be used for some basic structure never-the-less.
Among many others, one of GPT-3’s more prominent uses will likely be as a “writer starting point” tool. In fact, Writesonic has already created a GPT-3-based product to do just that — write pretty decent text for you. There is an entire cohort of GPT-3 Startups focused on content writing!
#3 Demo designer
Explaining one’s ideas to others can be a complicated task, especially if the ideas are highly visual. One of the most cited uses of GPT-3 that I am particularly excited about is its use in generating application demos:
The same principle through which those applications have been created can be applied to generating websites, landing pages & general layout design. A recently introduced online platform — Debuild — also focuses on building demo apps lightning fast — with pure language!
The Limitation
While impressive to the eye, App & Website designs created by GPT-3 lack a lot of main features that real-world products have: a good About Page, a User-database, a Logo, etc.
The produced apps are certainly a good means of producing a demonstration, but in their current form, are not immediately usable. Safe to say that GPT-3 will not be replacing the app developer any time soon.
The Future
That being said, the visual aspect of the produced applications is very impressive — especially considering that the entire process takes less than 10 seconds. Such basic design sketches can serve as effective inspiration for designers & entrepreneurs — to visualize their ideas better both to themselves and to an audience, in a time-efficient manner. Designer & Debuild.co are paving the way to making this possible with GPT-3! There is an entire cohort of GPT-3-based products focused on making various designs — including CopySmith AI, who might save marketing teams many hours in the near future!
Conclusions
The uses cases above will likely make their way into the workflow of professionals in the years to come. As we already said, while impressive, these tools will be far from perfect to begin with. However, the time-saving potential they have is incredible — even for the basic tasks they’re currently good at!
Design & basic programming work might require a lot fewer skills in the future. Do you think this is a good thing?
Let me know in the comments below!
StAI smart!
