Email Writing Apps: What Can They Really Do?
Isadora Teich wrote this article
Currently, while Gmail does offer short and sweet automated replies, they are not quite comprehensive enough to replace the act of sending and writing emails entirely.
However, in the future, eventually, most professional emails may be drafted by AI. Currently, several tools are in the works.
Let’s take a look.
Compose is just one of several brand new automated writing tools that may change the game. This family of apps and tools are all built using the exciting text-generation technology known as GPT-3. You may remember this, as it went viral after being revealed over the summer by the artificial intelligence research institute AI.
People across the world were floored as it created a wide range of content with alarming accuracy. This included memes, blog posts, and even Harry Potter fanfiction. On the downside, it also can be used to craft hate speech. This is because of how GPT-3 learns.
Its algorithm creates text based on digesting vast amounts of internet content.
Some entrepreneurs are already making use of GPT-3 to do real work. This includes writing emails and creating marketing copy.
This new tool is meant to automate the writing process for web copy.
It generates words for web pages and Google ads using basic information about a brand or campaign. Marketer Chris Frantz, a Snazzy cofounder, says the service streamlines the process of content creation by removing the work of having to come up with the ideas in the first place.
“The goal is to offload the somewhat monotonous job of writing the copy, and move to the editing part.” He told WIRED.
How Does AI Compete With Humans?
Every time a new innovation is introduced, people seem to either automatically fear and revile it, or claim that it will quite literally change everything.
Remember when people were obsessed with 5G? Some people claimed that it was part of a global conspiracy and even causing coronavirus outbreaks. On the flipside, countless outlets were pouring out articles like this one by Inc, called “ How 5G Will Fundamentally Change Everything You Know About Mobile Computing–From Farms to Phones.”
Months on, it turns out the actual reality is a lot less interesting. It looks like that while 5G has a lot of potential, there are a lot of kinks to be worked out before we get there.
So, outside of hype that sounds like the plot of a sci-fi movie, how does GPT-3 really hold up against human effort? The answer is surprising.
VWO is a company that helps companies quantify the performance of their marketing content. They put AI-created copy to the test. GPT-3 was pitted against human-written material for clients, including travel site Booking.com.
Of six tests with statistically significant results, AI-generated copy gained more clicks or interactions twice. The human-created copy only performed better once. All of the remaining three matchups were tied. While more tests are ongoing, so far; it looks as if AI-generated copy can definitely hold its own against copy written by people.
However, it’s not just about quality.
VWO founder Paras Chopra believes marketers will gravitate to auto-generated material for other reasons. These tools speed experimentation, which gives marketers more opportunities for growth. “The more you can test, the higher the likelihood you end up impacting your business metrics.”
A Penchant For Nonsense
With any tool, especially an emerging one, there are some aspects that need to be refined. As it stands now, one major con of GPT-3 technology is that it can churn out a lot of irrelevant nonsense, which then needs to be sifted through.
This is because apps built on GPT-3 send snippets of text dubbed “prompts” to OpenAI’s cloud servers. Then, GPT-3 sends back new text it calculates seamlessly from the input. This is all based on statistical patterns it saw in online text.
It is an unusual way of interacting with a computer.
It is fascinating and entertaining due to all of the possibilities it presents. However, this can be a bit difficult when you want to achieve a specific result. GPT-3 broadly digests anything on the internet that could be relevant based on its algorithm, without any grounding in reality. This means that a lot of what it generates is irrelevant.
OthersideAI is another email generation app powered by GPT-3.
According to their website, the app empowers users to “Generate full-length emails in your unique style by simply typing an outline of your key points. Watch your perfect email write itself in front of your eyes and get through your inbox faster than ever.”
According to Matt Shumer, a cofounder of OthersideAI, they are working to iron out a lot of the kinks.
Apparently, an early version of their app was “too creative,” which made it unreliable to use. It would correctly interpret prompts such as “2 pm meeting” but then tack on all kinds of fabricated events. This has been fixed, and tweaks are now being made to empower the app to express user’s unique writing styles.
WIRED’s Test on the Magic Email App
WIRED used the GPT-3-powered Magic Email App and found that while the app could handle basic emails very well, things got a little bit weird with more complex messages.
When given simple bullet points, the app could create a more professional and warm-sounding email surrounding them.
However, when it comes to more complex messages, more corrections were often required. This raises an important question. Is it faster to go back and edit AI-crafted emails for errors, or just write them yourself in the first place?
AI Struggles With Context
Also, in an effort to curb hate speech, it appears that Magic Email has made it impossible for users to include certain words, regardless of their context.
This raises interesting questions about censorship and what exactly AI can do for us. While, eventually, they may be able to correctly interpret nuances of context in conversation, at the moment, they cannot.
This is a big problem when AI essentially scans the entire web for content and parrots it back. Studies have shown that GPT-3 has already absorbed white nationalist terrorist content. It has been almost 5 years since Microsoft’s chatbot Tay, which had been designed to mimic a teenage American girl, inexplicably started tweeting about 9/11 conspiracies and in support of Hitler.
To this day, researchers are still not entirely sure of how to approach this problem.
What Do You Think?
At the moment, it is unlikely that all of our emails will be automated by next week. However, in the future, increasing automation is likely, especially for simple messages. If you choose to use this first-generation of AI-generated email apps right now, it is likely you will have to do some editing anyway.
Are you in favor of automated emails and want to get in on this innovation early, or do you not feel like it’s necessary?
Talk to me.
About ChopDawg.com: Since 2009, we have helped create 350+ next-generation apps for startups, Fortune 500s, growing businesses, and non-profits from around the globe. Think Partner, Not Agency.
Follow us on Twitter
Like us on Facebook
Double tap us at Instagram
Connect with us on LinkedIn
Find us on social at #MakeItAppn®