On GPT-3, OpenAI, and APIs

A new beta release from OpenAI has thrown the API world into a speculative frenzy of the potential and implications of AI and APIs.

Consider the following:

“OpenAI, a non-profit artificial intelligence research company backed by Peter Thiel, Elon Musk, Reid Hoffman, Marc Benioff, Sam Altman and others, released its third generation of language prediction model (GPT–3) into the open-source wild. Language models allow computers to produce random-ish sentences of approximately the same length and grammatical structure as those in a given body of text.”

Nothing too special, right? Just some boilerplate PR-driven tech journalism. Except it’s written entirely by an AI.

This article about GPT–3 posted on developer Manuel Araoz’s blog was written entirely using GPT–3. GPT–3 stands for Generative Pretrained Transformer version 3, and its creators, OpenAI, have set the tech world ablaze with its performance and its implications.

GPT–3 is a neural-network-powered language model. Language models use probability to fill in text. They’re basically Google’s predictive text on a grand scale. It’s not a new or even particularly novel technology. What is striking about OpenAI’s GPT–3 is its size and scope. The language model features over 175 billion parameters. This is over 10x as many as the previous incarnation, GPT–2.

Both APIs and AI have been around for a while. They might not be as headline-grabbing as they once were, but with the arrival of GPT–3, that looks as if it’s about to change. AI working in conjunction with APIs, are a powerful combination. Together, we may be standing on the brink of the augmented future we’ve been waiting for.

Meet GPT–3

It’s not often that a somewhat academic tech release generates such a flurry of hype, worry, and scorn. GPT–3, the model from OpenAI, has been generating thinkpieces aplenty alongside a cavalcade of ecstatic social media posts from those fortunate enough to Beta test the new API.

Consider this tweet from developer Sharif Shameem, who’s working on a better way to build web apps:

In it, Shameem illustrates how GPT–3 allows him to describe an intended layout, using common language, and it returns the resulting JSX code.

GPT–3 is already generating workable code, and it’s still in its Beta stage.

Machine Learning architect Ayush Patel demonstrates GPT–3 can generate SQL code, as well:

GPT–3 has also already been used to generate creative fiction and even write poetry.

GPT–3 brings us that much closer to Artificial General Intelligence, or AGI, which is the whole reason OpenAI was created in the first place. Its founders – who include some of Silicon Valley’s most powerful and influential thinkers, including Elon Musk – realize AGI is inevitable. It’s imperative that people with ethics get there first; from deepfakes to political sabotage, we’ve already seen what kind of damage even rudimentary Machine Learning can wreak on an unsuspecting public.

GPT–3 is just the most recent example of an AI-powered by an API. Let’s take a look at some other innovative ways that AIs and APIs are working together. First, we’re going to look at how APIs can fuel an AI, and vice-versa.

How APIs Empower AI

Machine learning networks are only as good as the data they’re trained with. It’s relatively easy to create a primitive AI bot that can query a database and return with a simple answer. It’s a challenge of a whole different magnitude to create a neural network that can effectively think, reason, and logic.

Combining APIs and AI makes sense. APIs provide a non-stop torrent of real-time information, so the network is continually learning, adapting, and evolving. Even better, AIs and APIs create a powerful symbiotic relationship, where an API helps train the AI and then the AI feeds the API, in turn.

To illustrate, consider this case study of how Prolife Foods, a retailer in Australia and New Zealand, cleverly employed a mixture of AI and APIs to streamline their operation in a number of surprising ways.

Prolife Foods: A Case Study

Machine learning is tailor-made for retail scenarios. After all, it’s what drives suggestive selling and product recommendations. Of course, retailers will investigate how they can integrate Machine Learning to make their retail operations more competitive and profitable.

Prolife Foods collaborated with a hardware manufacturer named Aopen and meldCX, a developer web-based API solutions, to integrate Machine Learning. MeldCX uses a mixture of computer vision and data analysis to assess objects at the POS. The software then analyzes the data, stores the behavior, and then makes a business decision.

Seems relatively simple, right? Even this relatively rudimentary Machine Learning model can have some surprising implications.

This case study from Insight.Tech delves into some of the potential implications of this system. APIs working with AI can help reduce shrinkage, improve accuracy, improve productivity, and make for a better customer experience across the board.

Other Applications Of AI and APIs

AIs often have their own APIs, making their services available to developers. Consider the Google Prediction service, which is a lot like GPT–3’s younger sibling. Google Prediction does more than simply providing predictive text, however. It can also use Machine Learning for pattern recognition, natural language processing (NLP), as well as provide a recommendation engine. Applications that use the Google Prediction API can be used to perform sentiment analysis, classify documents, or predict purchases, as well.

[Wit.ai[(https://labs.wit.ai/demo/index.html) is another NLP network that allows developers to add speech functionality to mobile apps. The Wit.ai API can be used to add speech functionality to home automation, smart TV, or wearable tech.

There are dozens of AI-related APIs out there. Virtually every virtual assistant offers an API, for instance. The Alexa Skill Management API is one example of how AI-fuelled APIs can be employed in any number of innovative ways. The Alexa Skill Management API lets you integrate the API to learn new skills. The Summarize Bot API quickly summarizes data from any number of sources, and hints at some of the ways that AI-fuelled virtual assistants might take advantage of the API format.

Conclusion: AI and APIs

Artificial Intelligence and APIs go hand in hand. APIs essentially function as the nervous system, while AIs and Machine Learning make up the brain. With the Beta launch of GPT–3, we’re starting to see the actual potential of Neural Networks, with Machine Learning being trained with enough data they can actually think.

APIs need AI and Machine Learning, as well. APIs can leave organizations and networks vulnerable to cyber-attacks, for instance. APIs are likely going to have to integrate some kind of AI or Machine Learning, going forward, to ensure their security.

Programs like OpenAI are going to keep being increasingly important as AI gets closer and closer to attaining sentience. They’re ensuring that AI is implemented fairly and justly. It hints at the potential for a truly augmented future, with Humans and machines living in harmony.