One year ago, there was no ChatGPT, no easy way to ask questions or get quick, detailed answers. People couldn’t discuss the new AI that kids used for homework or the one making headlines, promising to change the world. There was no hype around GPT, short for generative pre-trained transformers, reaching new heights.
Now, as we celebrate the launch of ChatGPT, let’s reflect on the past 12 months. It’s been a year that brought significant changes, both positive and negative, including recent dramatic events at OpenAI. I’ll be looking back at VentureBeat’s coverage throughout the year to understand these developments.
November 2022, OpenAI Introduced ChatGPT and the GPT-3.5 Series
On November 30, 2022, during the NeurIPS machine learning conference in New Orleans, there were widespread rumors about GPT-4. Attendees speculated that the conference might unveil details about GPT-4.
However, OpenAI surprised everyone by introducing a new model within the GPT-3 family, known as text-davinci-003. This model, part of the ‘GPT-3.5 series,’ was said to outperform its predecessors by handling more complex instructions and generating higher-quality, longer-form content.
Simultaneously, almost as an afterthought, the company released an “early demo” of ChatGPT, another component of the GPT-3.5 series. Described as an “interactive, conversational model,” ChatGPT’s dialogue format allowed it to answer follow-up questions, acknowledge mistakes, challenge incorrect assumptions, and reject inappropriate requests. Surprisingly, ChatGPT quickly gained immense popularity, becoming the fastest-growing consumer application in history.
December 2022, OpenAI CEO Admits ChatGPT Risks. What Now?
Two weeks after the release of ChatGPT, it became the talk of the town, generating massive hype along with a wave of criticism. People called it everything from a “sensation” and “the most disruptive technology since [fill in the blank]” to a “world-class bull**** artist.” Some likened it to that person at the bar who talks endlessly, mixing engaging facts with random nonsense, but whom you’d never want to introduce to your parents.
OpenAI CEO Sam Altman suddenly entered the Twitter conversation (it was still Twitter back then), urging caution. He pointed out that ChatGPT is incredibly limited and, while good at certain things, could create a misleading impression of greatness. Altman emphasized that relying on it for anything important was a mistake at that time. He described it as a preview of progress, highlighting the need for further work on robustness and truthfulness.
January 2023, Top AI Conference Bans ChatGPT in Paper Submissions (and Why It Matters)
In January, a heated discussion had emerged regarding how businesses, organizations, and institutions would handle the emergence of large language models capable of aiding communication or, from different perspectives, borrowing, expanding, or even plagiarizing ideas.
An early instance of this debate unfolded at a machine learning conference. The International Conference on Machine Learning, while discussing the use of machine learning, explicitly stated in its call for paper submissions that “papers containing text generated from a large-scale language model (LLM) like ChatGPT are not allowed unless the generated text is presented as part of the paper’s experimental analysis.”
February 2023, Google Puts $300 Million Into Anthropic as the Competition With ChatGPT
In February, the competition to rival ChatGPT was in full swing. Google joined the race by investing $300 million in Anthropic, one of OpenAI’s most talked-about competitors. Anthropic had recently unveiled its own generative AI model named Claude. Financial Times reported that Google’s investment would secure around a 10% stake in Anthropic, valuing the company at approximately $5 billion.
Founded in 2021 by researchers who had left OpenAI, Anthropic gained significant attention in April 2022. In less than a year, the company surprised many by announcing a substantial $580 million in funding. Notably, a significant portion of this funding came from Sam Bankman-Fried and individuals associated with FTX, the cryptocurrency platform that faced accusations of fraud and subsequently went bankrupt.
March 2023, OpenAI Releases Highly-Anticipated GPT-4 Model in Surprise Announcement
Just four months after the launch of ChatGPT, OpenAI surprised everyone with the release of the much-anticipated GPT-4 model. This update enhanced the technology powering the popular chatbot, ChatGPT, and OpenAI touted GPT-4 as its “most advanced system, producing safer and more useful responses.”
GPT-4 improved the core technology of ChatGPT, enabling the chat software to tackle more challenging problems with increased accuracy. This was made possible by its broader general knowledge and enhanced problem-solving abilities.
Additionally, GPT-4 introduced new features, such as the ability to process images as inputs and generate captions, classifications, and analyses. With a capacity to handle over 25,000 words of text, GPT-4 opened up possibilities for long-form content creation, extended conversations, and document search and analysis.
April 2023, As Critics Circle, Sam Altman Hits the Road to Hype OpenAI
OpenAI experienced a splendid spring, and CEO Sam Altman capitalized on it by initiating a global spring tour. It kicked off with an in-person meeting in April with Japan’s prime minister, where Altman hinted at potential plans to establish an OpenAI office and expand services in the country.
The tour, covering 17 cities, aimed to promote OpenAI and included stops in Toronto, Washington D.C., Rio De Janeiro, Lagos, Madrid, Brussels, Munich, London, Paris, Tel Aviv, Dubai, New Delhi, Singapore, Jakarta, Seoul, Tokyo, and Melbourne.
However, the tour coincided with challenges for OpenAI on various fronts. Elon Musk, Steve Wozniak, and thousands of others signed a contentious open letter advocating for an AI ‘pause.’ Italy declared a ban on OpenAI’s ChatGPT due to data privacy concerns, and there were complaints that GPT-4 violated FTC rules. Additionally, a ChatGPT bug exposed security vulnerabilities during this period.
May 2023, OpenAI is Introducing ChatgGPT Plugins for Plus Subscribers, Here’s Why That’s a Big Deal
In May, the positive developments at OpenAI continued with the announcement, detailed in a release note, that ChatGPT Plugins would be introduced for ChatGPT Plus subscribers. The beta release enables ChatGPT to “access the internet and utilize 70+ third-party plugins.”
OpenAI mentioned that ChatGPT Plus users would have early access to experimental features, subject to potential changes during development. These features would be accessible through a new beta panel in the settings, gradually rolling out to all Plus users over the next week.
Back in March, OpenAI had officially introduced 11 third-party plugins, featuring branded offerings from Instacart, Kayak, and Zapier. The expanded offering of over 70 third-party plugins now included functionalities for chess play, recipe-finding, live soccer updates, and nutrition. Developers interested in creating ChatGPT plugins could join a waitlist for access.
June 2023, More Details Emerge About Google Deepmind Efforts to Surpass ChatGPT
Come summer, the competition in Large Language Models (LLM) was once again intensifying. A few months after the collaboration of Google Brain and DeepMind under the banner of Google DeepMind, aiming to counter the competitive challenge posed by OpenAI and its transformative ChatGPT, DeepMind CEO Demis Hassabis revealed ongoing efforts on a new system called Gemini. Teased at Google I/O in May, Gemini was said to draw on techniques that contributed to AlphaGo’s victory over a Go champion in 2016.
As per reports, Hassabis mentioned that the Gemini system would integrate LLM technology with reinforcement learning techniques from AlphaGo, with the objective of imbuing it with enhanced planning and problem-solving capabilities. However, by November, there were reports of a delay in the Gemini project.
July 2023, OpenAI Product Leader Denies Claims GPT-4 Has Gotten ‘Lazier and Dumber’
During the summer, OpenAI encountered additional challenges. The Federal Trade Commission initiated an investigation into the leading generative AI company for potential violations of consumer protection laws. Simultaneously, comedian and author Sarah Silverman filed a lawsuit against OpenAI and Meta, alleging copyright infringement of her humorous memoir, “The Bedwetter: Stories of Courage, Redemption, and Pee,” published in 2010.
However, the most significant response came in the wake of a report suggesting that OpenAI’s GPT-4 model, the powerhouse behind ChatGPT, had become “lazier and dumber” due to a “radical redesign.” OpenAI’s product team addressed this claim, with Peter Welinder, the VP of product at OpenAI, countering it on Twitter.
He clarified that not only had the company not made GPT-4 less intelligent, but each new version was actually smarter than its predecessor. Welinder hypothesized that with increased usage, users might notice issues that were not apparent before and encouraged users to report any perceived regressions for investigation.
August 2023, OpenAI Launches Long-awaited ChatGPT for Enterprise but is It Playing Catch-up?
OpenAI’s move into the enterprise sector had been highly anticipated. However, when ChatGPT for Enterprise was finally launched in August, some questioned whether the company was trying to catch up.
Numerous other companies were already vying for attention in the enterprise business audience with generative AI. Cohere provided tailored Large Language Model (LLM) solutions, Anthropic partnered with Scale AI for enterprise targeting, and even Microsoft Azure had its own OpenAI service. Additionally, open-source players were in the competition, with Meta’s LLaMA 2, for example, being available for commercial use.
September 2023, OpenAI Grants ChatGPT Access to the Entire Internet
From the beginning, one of ChatGPT’s major limitations was its lack of knowledge beyond September 2021.
However, in September, OpenAI made a significant announcement. They revealed that ChatGPT can now access the internet to provide current and authoritative information, including direct links to sources. This capability was achieved through integration with Microsoft’s Bing search engine.
This development resembled the earlier introduction of ChatGPT third-party plugins. Alongside that announcement, OpenAI introduced two of its own plugins: Code Interpreter (now renamed “Advanced Data Analysis,” enabling ChatGPT to handle uploaded files) and “Browsing,” utilizing the Microsoft Bing API and a text-based browser to search the web and summarize information for users.
The reintroduction of web browsing for public, non-paywalled sites was enthusiastically shared by the company’s leadership on their personal X accounts, with CEO Sam Altman expressing excitement and CTO Mira Murati echoing the sentiment.
October 2023, OpenAI Brings DALL-E 3 Image Generator to ChatGPT for Enterprise, Teases Classifier
During October, OpenAI introduced access to its latest image-generating AI model, DALL-E 3, for users of both the ChatGPT Plus subscription service and ChatGPT for Enterprise.
In a blog post unveiling this development, OpenAI stated, “In comparison to its predecessor, DALL-E 3 produces images that are not only more visually striking but also sharper in detail. DALL-E 3 excels in rendering intricate details such as text, hands, and faces reliably. Moreover, it demonstrates proficiency in responding to extensive, detailed prompts and supports both landscape and portrait aspect ratios.”
November 2023, CEO Controversies
In November, OpenAI, and consequently ChatGPT, kicked off the month on a positive note with a series of announcements during the company’s inaugural developer conference, Dev Day. The unveiling of custom GPTs, the introduction of the new GPT-4 Turbo, and the launch of the Assistants API created a massive wave of hype and excitement that swept over Silicon Valley and beyond.
However, on November 17, OpenAI made a significant move by firing CEO Sam Altman, less than two weeks re-hired before ChatGPT’s first anniversary. The resulting drama hinted that the excitement around OpenAI might never be quite the same.
Nevertheless, there’s no denying that ChatGPT’s impact on mainstream usage has been transformative over the past year. As we look ahead, we’ll have to wait and see what 2024 has in store for this chatbot, which, it’s important to note, is fundamentally “a large language model.”
Summary
In just a year, ChatGPT went from being non-existent to changing the way we talk and sparking discussions. OpenAI had its ups and downs, from ChatGPT’s launch to introducing GPT-4 and facing controversies. Reflecting on this year, uncertainties linger, especially after recent CEO issues. Yet, ChatGPT’s undeniable impact on everyday communication hints at a new era in AI interaction. Looking ahead, the fate of this “large language model” is unclear, and 2024 may reveal the next steps in its evolution.
