[ad_1]
When I send an email, it is special. A beautiful, elaborate thing that, who am I kidding, is almost automatic. So why not automate it? OthersideAI is taking this idea (with a starting round of $ 2.6M) beyond autoresponders and smart replies, using OpenAI’s GPT-3 language generation engine to convert bullets into full, custom messages.
Pre-trained generative transformer 3, is of course the latest version of the artificial intelligence model that writes such a compelling copy that everyone under the sun has let him write his column about it, and then tries to surprise readers by revealing the fact at the end. (However, there are usually some indicators.) “Data-reactid =” 13 “> GPT-3, or Pre-trained generative transformer 3, is of course the latest version of the artificial intelligence model that writes such a compelling copy that everyone under the sun has let him write his column about it, and then tries to surprise readers by revealing the fact at the end. (However, there are usually some clues.)
However, access is carefully limited and the OthersideAI team has a welcoming but uncharacteristic relationship with OpenAI. It started when the team was working on their previous project and found that they had more emails than they could handle. At the time, the GPT-3’s predecessor, GPT-2, was all the rage.
OpenAI creates an all-purpose API for its text-based artificial intelligence capabilities
“We created a cold email with him, but then we thought: that might be the business we should be pursuing,” said CEO Matt Shumer. “So we decided to go all out.”
He and his colleagues Jason Kuperberg and Miles Feldstein created a demo that got a bit of attention when they posted it on Twitter and soon gained access to the new version of the GPT engine.
You could say that OpenAI already did the hard part building this amazing language engine, but it’s not as simple as letting it run freely in someone’s inbox. With no restrictions, GPT-3 will chase its own tail down a rabbit hole, churning out really weird stuff, as any AI Dungeon player can attest.
“GPT-3 does an amazing demonstration, but putting it on a product is another story,” Shumer said. “Our job is, in a sense, to tame their creativity.”
The resulting product turns a summary or bullet point into a full email and looks like this in action:
If you don’t like the result, or there is a bug, or you just like to torture AI, you can press the button and it will generate it again, differently. Modify it a bit first and the system will understand that in the future you would prefer the new shape.
GPT systems train on millions of words and phrases, and then generate text inspired by that corpus after receiving input to work. In this case, the system takes as input not only your bullets, but other information from the email chain and past user preferences.
That way, you don’t just get the context – you can say “It was great to sit down for coffee with you” if the coffee is referred to, even if you just wrote “nice to meet” on the panel. And also learn your style, preferring certain words or phrases or learning that you like to close in a certain way.
You can make good guesses about technical and financial details, such as when placing a job offer:
Of course, for something so important, you may ask yourself: why bother letting an AI do it?
Kuperberg said the company, which has nearly 10,000 people waiting to enter its trial version, has seen interest from engineers and developers, as well as sales and support staff. One instantly sees the application in a sales or support scenario where a handful of scheduled questions or answers can be re-generated to be different each time, or slightly adjusted for the person or situation. That avoids the feeling of receiving an “email form” even though it amounts to the same thing.
I mentioned the possibility of helping people who have trouble writing; someone who must write letter-by-letter emails using gaze detection might find this extremely compelling. Shumer said this hadn’t been on their radars to begin with, but that in recent weeks they have seen interest in this direction.
Shumer was careful to ensure that safety comes first and that this is not a data-sucking operation; Obviously, no one would want to use a tool that reads your email and uses that information for nefarious purposes, with the notable exception of Gmail.
They feel secure in their approach, noting that Google seems more interested in selecting the correct answer for the context, and the text generation tools are not robust enough to handle the inputs that Otherside’s GPT-3-based system handles with ease. “
“If you want to do an email in the user’s tone, you can’t guess the details. You need a human. This is not a generated response, it is taking direction,” Shumer said.
The initial round of $ 2.6 million was led by Madrona Venture Group, with Active Capital, Hustle Fund, Chapter One and more participants. Everything points to team building so that the company can build a large-scale product.
Ultimately, they envision this as a small-scale test for a larger system of interlocking AIs that can securely connect to each other, answering questions and providing information in a human-like way, but with minimal human involvement. Obviously that’s a long-term goal, but given all the talk for a decade or so about replacing email has been nothing, maybe it’s time to embrace it, but let someone (or something) more. take over some of the load.
Here are some ways GPT-3 can go wrong
