About ChatGPT
Even if you are not into artificial intelligence, it is now time to pay attention to ChatGPT, because this is a big deal. (Shankland, 2023) ChatGPT was released in November 2022, and since then has taken the Internet by storm, gathering more than a million users within five days of launching. (Ortiz, 2023c) ChatGPT is an artificial intelligence (AI) model that engages in conversational dialogue. Put simply, it is an example of a chatbot, similar to the automated chat services found on companies’ customer service websites. (Reiff, 2023)
The AI model can provide answers on a wide range of topics, and (for the most part) provides plausible answers in mere seconds! The AI has been proven to assist humans in their work, and has the potential to have a large impact on how the Internet search works. It could disrupt business models and provide new ways of sourcing for information.
Currently, ChatGPT is in the testing phase and is still free to sign up by creating an OpenAI account. Below is a screenshot of what ChatGPT answers look like:
This is how ChatGPT introduced itself when I asked it to explain what it is in a simple way. Hint: you might be pleased to know that ChatGPT also works in the Malay language.
Background
ChatGPT falls under the larger umbrella of generative AI. Generative AI are AI algorithms that generate or create an output, such as a text, photo, video, code, data, and 3D renderings, from data they are trained on. (Ortiz, 2023b) Building generative AI models are a massive undertaking, and only the few heavyweight tech giants can afford to employ some of the world’s best computer scientists and engineers. (McKinsey, 2023)
ChatGPT is just one of the most well-known examples because it was made free to use for the public, so that the creators can collect even more data and feedback about the technology. What sets ChatGPT apart from chatbots over the last several decades, however, is that ChatGPT was trained using Reinforcement Learning (RL) from Human Feedback (RLHF). RLHF involves the use of human AI trainers and reward models to develop ChatGPT into a bot capable of challenging incorrect assumptions, answering follow-up questions, and admitting mistakes. (Reiff, 2023)
OpenAI
ChatGPT was developed by OpenAI, an AI research and deployment company that conducts research and implements machine learning. (Crunchbase, n.d.)
OpenAI was co-founded by big names in the technology world, such as Elon Musk [Chief Executive Officer (CEO) of Tesla, SpaceX and Twitter], Sam Altman (current CEO of OpenAI and former president of Y Combinator) and Reid Hoffman (co-founder of LinkedIn).
Since the launch of ChatGPT, the valuation of OpenAI has skyrocketed. As of January 2023, OpenAI is estimated to be valued at USD29 billion, more than double its valuation two years ago. (Niedens, 2023)
Microsoft sees the tremendous potential of the company and has been rumoured to invest USD10 billion in OpenAI (Bass, 2023)
Application and Benefits of the Technology
ChatGPT continues to deliver surprising results in many use cases, and increases the productivity of people utilising the free service.
ChatGPT can write emails, essays, and even poems. It can also tell short stories and jokes. Moreover, it will learn from your previous responses and adjust the answers when you nudge it in the right direction.
The technology can also be used to describe art in great detail, create AI art prompts, and have philosophical conversations. ChatGPT can also help in creating basic lists such as packing lists and to-do lists. (Ortiz, 2023c)
Others have reported being able to use ChatGPT to negotiate lower bills from Internet providers, hospitals and more. They can also generate the next week’s meals with a grocery list, create a bedtime story for children, and help to prepare for an interview. (Lumb, 2022)
Additionally, people are also using ChatGPT to create workout plans. (Williams, 2023)
AI Getting an MBA
ChatGPT can even achieve a passing score of B to B- on a Master of Business Administration (MBA) examination (exam). (Terwiesch, 2023)
Christian Terwiesch, a lecturer from Wharton School of the University of Pennsylvania, tested ChatGPT’s performance on the final exam of a typical MBA core course, Operations Management. These were the findings from his paper:
- First, it did an amazing job at basic operations management and process analysis questions including those that were based on case studies. Not only were the answers correct, but the explanations were excellent.
- Second, ChatGPT at times made surprising mistakes in relatively simple calculations at the level of 6th Grade Mathematics. These mistakes could be massive in magnitude.
- Third, the present version of ChatGPT was not capable of handling more advanced process analysis questions, even when they were based on fairly standard templates. These included process flows with multiple products and problems with stochastic effects such as demand variability.
- Finally, ChatGPT was remarkably good at modifying its answers in response to human hints. In other words, in the instances where it initially failed to match the problem with the right solution method, ChatGPT was able to correct itself after receiving an appropriate hint from a human expert. (Terwiesch, 2023)
Limitations
There are a few limitations to the technology behind ChatGPT in its current form. Firstly, the AI model is only trained on data up to 2021, and is not connected to the Internet for the most up-to-date information. Secondly, the model will generate an answer but no sources will be provided. (Ortiz, 2023c)
Despite all the excellent examples of ChatGPT working flawlessly, it sometimes writes plausible-sounding but incorrect or nonsensical answers. There are still limitations to the technology, and it needs to be improved further to be used on a broader scale.
According to OpenAI, fixing this issue is challenging, as during RL training, there’s currently no source of truth. Training the model to be more cautious also causes it to decline questions that it can answer correctly and supervised training misleads the model because the ideal answer depends on what the model knows, rather than what the human demonstrator knows.
Other limitations to ChatGPT are that it is sensitive to tweaks to the input phrasing or attempts using the same prompt multiple times. For example, given one phrasing of a question, ChatGPT can claim to not know the answer, but given a slight rephrase, it can answer correctly. Ideally, ChatGPT should ask clarifying questions when the user provides an ambiguous query. Instead, it usually guesses what the user intended. (OpenAI, n.d.)
Speed of Adoption
Within just five days after launching, ChatGPT had more than one million users. (Ortiz, 2023c). Just two months later in January 2023, it was estimated to have reached 100 million monthly active users, making it the fastest-growing consumer application in history. By comparison, it took TikTok about nine months after its global launch to reach 100 million users, and Instagram, more than two years. (Milmo, 2023)
Competition from Google
Following the success of OpenAI’s ChatGPT, Google decided to release its own version of AI on 6 February 2023, called Bard. Bard draws on all information from the web to provide responses, providing updated information. Bard is an experimental conversational AI service powered by Language Model for Dialogue Applications (or LaMDA). When Google CEO Sundar Pichai announced Bard, he also talked about the advancement of technology today. The scale of the largest AI computations is doubling every six months, far outpacing Moore’s Law. (Pichai, 2023)
Microsoft Bing and Edge
The next day, on 7 February 2023, Microsoft made another announcement, integrating ChatGPT’s prowess into Bing search and the Microsoft Edge browser. This new version solves ChatGPT’s important drawbacks: with the new Bing and Edge connected to the Internet, this version provides citation for its sources. Microsoft is branding the new Bing and Edge as “your co-pilot for the web”. (Mehdi, 2023)
Observation
The release of ChatGPT to the public has unleashed a whole swath of new and innovative ways of thinking and problem-solving, paving the way for ground-breaking discoveries and advancements in various fields. It also forces us to think about productivity from a different perspective.
AI Wars
The launch of ChatGPT has kicked off a new wave of competition among the technology giants, and some are coining the term AI Wars to describe the phenomenon. Even Facebook, Snapchat, and Twitter have announced that they are joining the AI Wars (Pathak, 2023).
Among all the competitors, though, Google is the company most threatened by the integration of ChatGPT technology into Microsoft Bing and Edge. This is because Google is a search giant and the company’s business model and its main source of revenue comes from advertising.
The reason ChatGPT is significant to Google is because advertising is delivered through Google search results. If an Internet-connected chat AI can synthesise the information and deliver answer directly without having to access the websites, those websites will lose a huge portion of Internet traffic to the AI, affecting Google’s advertising.
In 2017, 86.5 percent of Google’s revenue was from advertising, while revenue from cloud made up 3.7 percent, and others made up 9.9 percent. In 2022, the share of advertising reduced slightly to 80.2 percent, with 9.3 percent from cloud and 10.3 percent from others. (Bianchi, 2023) Figure 1 visualises the revenue breakdown for Google in 2020.
