Until now OpenAI's amazingly effective natural language generation neural network GPT-3 has been in private beta with a limited number of users. As yet there is no date for general public access, but beta users have been notified of charges from the beginning of October, which indicates what GPT-3 might cost.

GPT-3 (Generative Pretrained Transformer 3) doesn't just predict the next word in a sentence, it can write complete stories given just a few words. If you don't know about it, read The Unreasonable Effectiveness Of GPT-3, which explains how its success is achieved in the context of AI research since the 1980s.

Unlike its predecessor GPT-2, which is open source under a modified MIT licence, GPT-3 is a commercial product. Moreover,  OpenAI, which started out as a non-profit before splitting to became a for-profit company in March 2019, intends to use it as a source of revenue, stating in the FAQ to a blog post in June 2020:

We see developing commercial products as one of the ways to make sure we have enough funding to succeed.

We also believe that safely deploying powerful AI systems in the world will be hard to get right. In releasing the API, we are working closely with our partners to see what challenges arise when AI systems are used in the real world. This will help guide our efforts to understand how deploying future AI systems will go, and what we need to do to make sure they are safe and beneficial for everyone.

OpenAI offers three main reasons for releasing an API instead of open sourcing the models used by GPT-3:

1. Commercializing the technology helps pay for ongoing AI research, safety, and policy efforts.
   
   
2. Many of the models underlying the API are very large, taking a lot of expertise to develop and deploy and making them very expensive to run. This makes it hard for anyone except larger companies to benefit from the underlying technology. We’re hopeful that the API will make powerful AI systems more accessible to smaller businesses and organizations.
   
   
3. The API model allows OpenAI to more easily respond to misuse of the technology. Clarifying this with:

Since it is hard to predict the downstream use cases of our models, it feels inherently safer to release them via an API and broaden access over time, rather than release an open source model where access cannot be adjusted if it turns out to have harmful applications.

Given that AI is so commercially valuable, and potentially very harmful if misused, making GPT-3 a proprietary technology seems justified and having Microsoft as OpenAI's preferred partner also seems sound. 

The pricing notified to current beta testers is subject to change, but provides an indication of the general order of magnitude of charges that OpenAI is exploring.

Starting 1 October 2020:

The FAQ answers "What does 2M tokens equal in terms of number of documents/books/etc" with:

This is roughly equivalent to 3,000 pages of text. As a point of reference, Shakespeare’s entire collection is ~900,000 words or 1.2M tokens.

So for $100 per month you might be able to generate the works of Shakespeare without a team of typing monkeys. On the other hand if you need Scale you probably do need deep coffers.

More Information

Language Models are Few-Shot Learners

Related Articles

AI Goes Open Source To The Tune Of $1 Billion 

OpenAI Recruiting Fellows 

Elon Musk Leaves OpenAI Over Conflict of Interest

OpenAI Five Dota 2 Bots Beat Top Human Players

OpenAI Bot Triumphant Playing Dota 2

Open AI And Microsoft Exciting Times For AI 

OpenAI Universe - New Way of Training AIs

To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin.

Comments

Make a Comment or View Existing Comments Using Disqus

or email your comment to: comments@i-programmer.info