ChatGPT celebrates 1st anniversary: A look at pathbreaking features, controversies, and more

As the fastest application to cross 100 million users, celebrates its first anniversary. Here is a look at the high and low points of this landmark technology.

OpenAI and the launch of ChatGPT

On this day, OpenAI unveiled ChatGPT to the world. The newly launched AI chatbot could produce human-like responses to almost any question. ChatGPT is generative AI – in simple words, it is an AI that can generate content rather than just analyse or act on existing data. In just five days, OpenAI’s latest product amassed close to 1 million users. In the run-up to the launch of ChatGPT, generative AI was the most talked-about tech in Silicon Valley, with OpenAI at the centre of all the action. OpenAI became a known force after launching its DALL-E, a text-to-image model, in January 2021. The company was founded in 2015 by Sam Altman and Elon Musk, who stepped down to focus on Tesla in February 2018. However, Musk remained a donor and advisor for some time.

At the time of OpenAI’s launch, the company’s goal was to “advance digital intelligence in the way that is most likely to benefit humanity as a whole.” Essentially, OpenAI is an AI research business that has shifted from being a non-profit to a ‘capped’ for-profit model. Taking a cue from the failure of Microsoft’s Tay chatbot and Meta’s Blender Bot, which gave controversial responses and false information, OpenAI employed a Moderation API, a system that helps in gauging things when they go against the company’s content moderation policies. OpenAI has described ChatGPT as a model that interacts in a conversational manner. Upon its launch, ChatGPT was limited to data before September 2021. Subsequent iterations saw this limit being reduced.

How the world responded to ChatGPT?

ChatGPT’s incredible capabilities stunned the world. In our assessment, while there were looming concerns about the loss of employment, the chatbot has largely been received well, with many seeing it as an efficient assistant for human tasks. For many, ChatGPT redefined the concept of machines’ ability to learn, proving that machines can indeed learn something as complex as human languages and the ability to interact.

Immediately after the launch, social media exploded with users sharing the various capabilities of the chatbot. From writing impeccable essays and assisting with travel planning to offering creative ideas and writing code, ChatGPT’s abilities were among the most talked-about topics on the internet for days. In February, ChatGPT made history by reaching 100 million users in just two months. To understand the chatbot’s soaring popularity, one should note that Facebook achieved 100 million users in four years, Snapchat and Myspace in three years, Instagram in two years, and almost a year for Google.

Big breakthroughs

In 2018, OpenAI introduced the first iteration of the GPT series, the GPT-1. It consisted of 117 million parameters and showed the ability of unsupervised learning. In February 2018, GPT-2 was introduced with 1.5 billion parameters and significant improvements in its text generation capabilities. However, this version was not released to the public. OpenAI later launched an upgraded version in November 2018. GPT-3 was unveiled in 2020 and was trained on 175 billion parameters; this version showed some groundbreaking capabilities and was introduced as ChatGPT in November 2022.

ChatGPT was also used for generating essays and even writing exams. It went on to pass several competitive exams. The chatbot cracked the Wharton MBA Exam, US Medical licensing exam, law school exams for programs at the University of Minnesota. The bot also failed a few exams, such as the UPSC where it failed to pass the 84 percent cut off in the general category.

On the other hand, GPT-4 was introduced on March 14, 2023. It is seen as a significant advancement over its predecessors as it could adapt to specific tones, emotions, and genres of writing. It could generate code, process images, and interpret various foreign languages. At launch, it was announced that GPT-4 came with support for 26 languages.

The phenomenal success of ChatGPT led to OpenAI launching the iOS app in May 2023, and in July the Android version was rolled out. In February 2023, the company introduced its pilot subscription plan ChatGPT Plus at $20 per month.

When ChatGPT made headlines

Misinformation and hallucination: This has been perhaps the most talked-about concern about the chatbot ever since its launch. ChatGPT could inadvertently gather misinformation from its training data and present it as fact – this phenomenon is known as hallucination. Several experts also raised concerns about the lack of nuance in understanding complex issues and its propensity to respond with dangerous instructions or showcase biassed behaviour. It should be noted that with every iteration, OpenAI has been working to reduce hallucinations in its chatbot.

Questions on academic integrity: Reportedly, in January 2023, schools in New York City banned ChatGPT over safety and accuracy concerns regarding the content produced by the chatbot. Educationists expressed concern that the technology could impact the learning abilities of students and hinder the development of critical thinking and problem-solving skills. Around this time, teachers from around the world had expressed concerns about students using ChatGPT for their assignments. This also led to the use of AI content detectors by institutions to prevent academic dishonesty.

On ethical implications: The potential for biassed or incomplete training data and questions about the transparency of the information the system uses was a recurring theme throughout the year. OpenAI reportedly also established an ethical AI team to address these issues.

Underpaying Kenyan workers: In January, OpenAI was severely criticised after a report in Time stated how poorly it treated its workforce in Kenya. The report revealed that the company paid as low as $2 an hour to outsourced labourers whose work included labelling toxic and harmful content.

Companies that banned ChatGPT: In July 2023, Apple reportedly banned ChatGPT from its corporate devices, according to an internal document reviewed by The Wall Street Journal. Similarly, in August 2023, JPMorgan Chase banned its employees from using ChatGPT for work purposes, citing concerns about the chatbot’s potential to generate inaccurate or misleading information. Other companies like Spotify and Amazon too had expressed concerns over the chatbot. It is to be noted all the organisations that banned the chatbot have restored it at different points in time.

Countries ban ChatGPT: Seven countries have banned ChatGPT: Russia, China, North Korea, Cuba, Italy, Iran, and Syria. However, Italy later restored access. The reasons for the bans vary, but they are mostly based on concerns about its potential for misuse, such as misinformation or the generation of harmful content. Some countries have also raised privacy concerns about the chatbot’s data collection practices.

Some notable features over the year

Expanded context window and internet browsing: As per the big reveal of OpenAI DevDay, the context window size has been expanded to 128,000 tokens. In March, OpenAI announced that its set of plugins will allow ChatGPT to perform a variety of tasks, including browsing the internet. As of today, the GPT-4, which is limited to ChatGPT Plus subscribers, has the ability to browse the internet.

DALL-E 3 in ChatGPT: In October, the company announced that it is integrating its text-to-image model DALL-E 3 into ChatGPT, allowing its users to create unique images from simple conversations. It is currently available for Plus and Enterprise users.

Build your own GPT: At OpenAI’s first-ever developer conference, Sam Altman announced that the chatbot will have the Custom GPT feature, which will allow anyone to create their own GPT applications. Users simply need to feed specific information to create a GPT that offers customised outputs. The feature has allowed millions of users to create their own chatbots for a variety of use cases.

Code Interpreter: In July, OpenAI made its in-house plugin Code Interpreter available for all ChatGPT Plus users. With Code Interpreter, ChatGPT can run code, create charts, analyse data, perform maths, edit files, and much more.

ChatGPT with Voice: In September, OpenAI launched voice chats for ChatGPT and in November, it was extended to both Android and iOS platforms. Earlier it was only limited to Plus and Enterprise subscribers. The feature makes ChatGPT conversationally adept as it offers human-like voice responses to questions from users.

We have attempted to capture key milestones and developments, there are inarguably many more achievements and challenges related to ChatGPT that have come to the fore in the year. However, all of that cannot be contained in a single piece, hence we have highlighted the significant elements for the sake of brevity. Considering the rapid pace of research and development, we can expect OpenAI to continue on its path to bring some novel innovations in AI.