Microsoft to invest in OpenAI: A closer look at the company behind ChatGPT and Dall-E

In 2015, its mission read, “Our goal is to advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate a financial return.” OpenAI was pitched as a place to seek “a good outcome for all over its own self-interest,” and to ensure that artificial general intelligence (AGI) does not end up replacing or harming humanity.

OpenAI is headquartered at the Pioneer Building in San Francisco (Image: HaeB/Wikipedia)

The organisation is dedicated to developing AI for the benefit of humanity as a whole. At the very core, its mission is to ensure that artificial general intelligence, which is essentially highly autonomous systems that outperform humans at most economically valuable work, benefits all of humanity. The company says that it will attempt to directly build a safe and beneficial AGI, but will also consider the mission fulfilled if its “work aids others to achieve this outcome.”

In 2019, OpenAI changed its non-profit status to “capped” for-profit, allowing it to rapidly increase investments. At around the same time, the company announced intentions to commercially license its technologies.

OpenAI and Microsoft

Microsoft also made a $1 billion investment in July 2019 as well. It also partnered with the company in 2016 where OpenAI became an early adopter of Microsoft’s “Azure N-Series Virtual Machines.” Keep in mind ChatGPT runs on Microsoft’s Azure cloud services.  As of now, the company is reportedly projecting $200 million in revenue for 2023 and $1 billion in revenue for 2024. But keep in mind that running each query on ChatGPT is also costing the company a lot of money.

Microsoft is also reportedly planning to add ChatGPT AI to Office products like Word and Outlook, and even its search engine Bing.

OpenAI’s work: ChatGPT and Dall-E and more

As one of its very first products, OpenAI released a public beta of “OpenAI Gym” in April 2016, an environment/platform for developing and testing learning agents. It’s essentially an open-source interface to reinforcement learning tasks, helping teach agents tasks like walking to play games.

Then in December 2016, the company released Universe, a software platform built for measuring and training an AI’s general intelligence “the world’s supply of games, websites and other applications.” It allows an AI agent to use a computer the way a human does, which is by looking at the screen and operating a virtual keyboard and mouse.

Microsoft might soon integrate ChatGPT into its Bing search engine (Express photo)

In 2020, OpenAI announced GPT-3, a language model containing 175 billion parameters that use deep learning to produce human-like text. GPT-3 is the base for deep learning models like Dall-E and ChatGPT.

In January 2021, OpenAI released the initial version of Dall-E. It has its origins in the GPT language model, although it uses the same to produce images rather than text, by swapping text for pixels. While the first version was fairly rudimentary and produced distorted, low-res images, April 2022 saw the release of Dall-E 2 with significant upgrades. The model then became capable of generating photorealistic images. The tool was made open to all in September 2022, though the number of credits are limited to 15 per month for a free user. 

A DALL-E 2 output from text description “Toyota Prius in a cyberpunk-themed city” (Express photo-generated via Dall-E)

In August 2021, OpenAI released Codex, a general-purpose programming model that translates your text in English into code. It’s designed to help speed up the work of professional programmers, along with helping amateur programmers to get started with coding. The model is the principal building block of GitHub Copilot.

But perhaps the biggest gains in popularity were made when the company launched ChatGPT, the AI chatbot capable of answering all kinds of queries in a very human-like manner. While the model isn’t yet tethered to the internet, meaning it cannot output real-time information, it’s already being envisioned by some as a replacement for Google.

How OpenAI got Google to declare a “code red”

ChatGPT isn’t yet capable of providing information about real-time or even recent events, but its ability to serve up information in simple sentences has got Google shifting uncomfortably in its chair. The chatbot goes the extra mile in explaining information to the user, which is something Google can’t do right now. Sure, the ChatGPT can be unreliable at times, but its abilities alone have led Google management to reportedly declare a “code red,” according to a New York Times report. Google fears the arrival of a major technology shift with the arrival of chatbots like ChatGPT, which could affect its business.

What lies ahead

At the same time, OpenAI is working toward releasing an updated language model — GPT-4 — in late 2023, possibly with over 1 trillion parameters, which will make it even smarter. The increase in parameters should help ChatGPT to produce more convincing responses at an even faster rate. We should also be seeing similar improvements to Dall-E when it upgrades its base from GPT-3 to GPT-4.