Generative artificial intelligence (AI) has shaken up the world order and created a sensation globally. ChatGPT has set a world record for a consumer application, reaching 100 million users in under two months. TikTok took nine months and Instagram took two years to reach that milestone.
The chatbot era started less than a decade ago, and many digital solutions saw the possibilities. The ability for a machine to converse with a user with minimal programming was expected to be groundbreaking. This innovation kicked off a chain reaction (a series of events or actions that are triggered by an initial cause or event) of AI research.
Fast forward to 2023 and the generative AI race has dwarfed the chatbot hype. Generative AI efforts were largely in stealth mode lurking in the shadows the past few years until OpenAI launched ChatGPT.
OpenAI is a research firm co-founded by tech industry luminaries, including Elon Musk and Peter Thiel from the PayPal mafia; Sam Altman, an entrepreneur associated with Y Combinator and Reddit; and Reid Hoffman, the co-founder of LinkedIn.
With the goal of creating AI models that are ethical and safe, the team — which started its work around 2015 — launched its breakthrough product, ChatGPT 3.5, in May 2022. The rise of the product since then and its potential to add efficiencies and disrupt the digital value chain have been exciting new developments.
What is so intriguing about these generative algorithms? How are they different from other AI innovations of the past? This article explores these questions.
What is ChatGPT?
ChatGPT is one of the pinnacle generative AI models, and it is ruling the roost as far as AI is concerned. To understand ChatGPT and its prowess, it's important to first understand generative AI and how it is different from other AI algorithms. Generative AI algorithms are a class of computer programs that have the ability to artificially think, extrapolate and answer questions based on data on which they are trained.
This is fundamentally different from traditional AI algorithms, which are limited to extrapolating from the dataset they are trained on. This means traditional AI algorithms can answer solely from the dataset they are trained on alone. The new generative algorithms are in a class above their AI predecessors in this regard.
ChatGPT is capable of generating incredibly human-like responses to a wide range of inputs. Given different types of inputs, words, phrases and language, this natural language processing (NLP) model can help give precise responses to a wide range of input.
The tasks that ChatGPT can accomplish include language translation, sentiment analysis, text summarization and more. However, one of the most striking features, apart from the above-mentioned NLP tasks, is its ability to write in a diverse tone and voice, making it a powerful model.
As the world was digesting the rise of this AI tool, Microsoft announced a $10 billion investment in OpenAI in January 2023. Microsoft also announced its plans to integrate the new AI tool with Bing, Office 365 and its browser, Edge.
As of September 2021, ChatGPT had 175 billion parameters. While that detail was publicly available, the next version of the tool, GPT-4, was rumored to have over 100 trillion parameters. Parameters are configuration variables that help the model with having human-like conversation.
What is Google Bard?
Google Bard, like ChatGPT, is a natural language processing tool developed by Google. The project was under development for a considerable amount of time before ChatGPT was launched and has kickstarted the AI war between Google and Microsoft. While Bard is also an AI NLP model that uses generative algorithms, it is touted to be more creative and emotive than ChatGPT, making it more humane in conversation.
The name “Bard” comes from the traditional term for a poet or storyteller in Celtic cultures, highlighting the model’s focus on creative writing. Building on the success of Google’s earlier language models like BERT, Google Bard is a continuation of the company’s work in natural language processing and artificial intelligence.
Although Google Bard has received plaudits for its amazing capabilities, there are also worries about the possible abuse of text produced by artificial intelligence. The ability to discriminate between human- and machine-generated content may become more elusive as technology develops, raising concerns about issues like authorship, plagiarism and even the veracity of historical texts.
Key differences between Google’s Bard and OpenAI’s ChatGPT
Both models are fundamentally different in many ways, including but not limited to corpus trained on, intended way to express, understanding, etc.
The greatest difference between ChatGPT and Bard is the respective datasets the models were trained on and the relevancy of the datasets. ChatGPT as a model was trained on datasets up until 2021, while Google Bard used Google’s own data sources.
Furthermore, ChatGPT has a monetary component attached to it. People can pay $20 per month and avail its service on demand, while Google Bard is free to use in countries in which it has been deployed. Finally, ChatGPT as a product is considered much more adept at writing longform content that is more analytical by nature. While Bard can give similar results, ChatGPT seems to have its nose ahead on this front.
The Google vs Microsoft Battle
What happens when two titans clash against each other in the market? One of the most intriguing events of 2022 was the launch of ChatGPT. Not only did it disrupt the ecosystem, but it also elicited an intense response from Google. For the first time in years, Google sensed intense competition, so much so that it launched Bard in a rush in February 2023.
With a hurried launch, Bard’s first interactions with humans caused concerns, leading to parent company Alphabet losing nearly $170 billion in market value. This was due to investors’ fears that Bard wasn’t competitive enough for ChatGPT, and it forced Google CEO Sundar Pichai to seek the help of all employees at Google to train Bard through conversation.
Furthermore, from a monetary and business-penetration perspective, OpenAI has roped in one of the “Big Three” consulting firms, Bain, to come up with a growth strategy for ChatGPT. With the help of Bain and its own initiatives, ChatGPT has found many premium clients for itself. Salesforce, Snapchat, Duolingo, Coca-Cola and Slack are a few of the leading names using ChatGPT.
In comparison, Google is yet to embark on an aggressive “go-to-market” strategy for Bard. It is understandable, as many tout it as the next evolution of Google’s search engine, transforming the way people navigate the web.
Finally, Microsoft has been able to grab users and turn people’s attention toward Bing, its own search engine, through the integration of ChatGPT. After struggling for many years, Microsoft has been able to make a small dent in Google’s market share by growing the user base for Bing, a feat that is commendable given Google’s dominance.
Governments, regulations and the way forward
AI regulation is one of the most heated debates in the innovation economy. With the advent of generative AI models such as ChatGPT and Bard, how are governments poised to react?
Recently, Italy banned ChatGPT in landmark legislation, citing privacy concerns. The ban was revoked shortly after the OpenAI team revealed its data-processing practices and implemented age-gating measures.
However, ChatGPT and other generative AI products could have more trouble with regulators, particularly in Europe, where data privacy is considered critical. This is particularly true in critical sectors like health care and financial services, where AI products are expected to gain a foothold.
Generative AI’s ability to rapidly answer questions and to generate answers at near-human accuracy can be frightening to the authorities. Any data that is given to these models is stored with it for training. Even if the deletion of this data can be enforced by regulation and user demand, the intelligence obtained by the machine from this data can’t be undone.
As a result, delivering sensitive information to these tools might prove problematic in regulated industries. Also, ChatGPT, when administered standardized tests — such as the GMAT, GRE and law exams — aced them compared with the performance of human counterparts.
Despite all these challenges, the world is moving toward a place where these products will penetrate consumer lifestyles, corporate decision-making and perhaps even governments’ data analysis.