WILTSHIRE, ENGLAND – JUNE 21: Revelers watch the sunrise over Stonehenge during celebrations
ChatGPT3 became the new internet sensation last year when it lets users generate text and answer complex questions in a way that feels almost human. But, beyond ChatGPT3’s prowess, the technology’s underlying impact – generative AI – on business is only just beginning to be felt.
ChatGPT3, along with its image generator cousin Dall-E, has the potential to revolutionize the way content is created, from blogs to white papers, from student essays to business correspondence. It gives access to expert-level syntax and grammar to anyone who uses it. But it also raises important ethical questions.
This isn’t the first time the technology has caught the public’s attention. IBM Watson made headlines in 2011 when it won the game show Jeopardy! and that of Amazon
AMZN
Yet these technological solutions, initially hailed as game-changers, have not had the impact that many were hoping for. IBM Watson rose to worldwide fame by triumphing at Jeopardy!, but it did not become the universal problem-solving engine some pundits prophesied. Yet machine learning has since become the most prevalent technology supporting AI at scale. Similarly, Alexa was initially understood as the revolutionary all-purpose personal home assistant and has not fully realized this promise as its underlying technology – deep learning neural networks – has seen massive developments.
ChatGPT3, introduced in November 2022 by private AI research institute OpenAI, is the latest product built off of the institute’s GPT3, the third iteration of its grand Generative Pretrained Transformer language model. The heart of the model is the Transformer Algorithm, introduced in 2017 by a Google Brain team in an article titled “Attention is all you need”.
Transformers are a type of artificial neural network architecture that uses self-attention mechanisms, which allow the model to process variable-length input sequences and learn dependencies between input elements of more flexible way compared to traditional recurrent neural networks (RNNs). This makes transformers particularly well suited to handle long-range dependencies and to parallelize the training process, making them faster and more efficient to train than RNNs.
By scaling the model and training it on more and more data from the Internet, OpenAI’s GPT3 produced surprising results, learning not only the structure of the English language, but also the languages coding he encountered, such as HyperText Markup Language or HTML. Given a prompt, it could write consistent and coherent text and even translate from English to HTML, allowing users to create web pages without knowing how to code.
The first vertical released by OpenAI based on GPT3 was Codex, which translates natural language into code (the basis for a coding auto-completion tool called GitHub CoPilot). ChatGPT3 is the latest spin-off from GPT3.
But, as transformative as ChatGPT3 seems to be, it comes with caveats. The content generated by ChatGPT3 may be biased or based on incorrect sources (eg he told me that Armenia is part of the European Union). This raises the question of how to verify the accuracy of the information it provides.
Another issue concerns intellectual property. ChatGPT3 can generate answers to questions based on a large amount of data from various sources, but unlike Alexa, it does not cite its sources. This raises the question of who should be given the intellectual property rights for the answers he provides.
It’s not yet clear to what extent ChatGPT3 will disrupt the way the general public uses the internet or the way students write their research papers, but the technology behind it has the potential to disrupt industries and processes like never before.
Beyond ChatGPT3, generative AI has already begun to disrupt major industries. In biopharma, generative AI can generate millions of candidate molecules for a certain disease and then test their application, dramatically speeding up R&D cycles. In the supply chain, it can optimize processes by generating scenarios and optimizing specific constraints. In marketing, it can personalize experiences, content, and product recommendations. In finance, it can generate personalized investment recommendations, analyze market data, generate and test different scenarios to propose new trading strategies.
Whatever the future of GPT3, generative AI will be a profound technological revolution that will have a huge impact on a wide range of industries and potentially contribute to some of the world’s most complex problems, such as education, health and climate change.