ChatGPT is arguably the most explosive development in generative AI technology, possibly in the technology field in recent memory. It seems that no one can talk or write about anything else, even my 8-year-old has jumped on the hype bandwagon firing endless questions at me to which I have few answers.
As consumer applications fare, the hype itself is not to be underestimated. The application unveiled on 30 November 2022 gained a million users in the first five days, and in recent analyst statements, the popular chatbot brought to you by OpenAI reached 100 million active users by the end of January averaging a mind-boggling 13 million unique visitors a day. To give you a sense of its exponential power, by comparison it took TikTok nine months to hit 100 million users, and Instagram over two years. It is by all accounts the fastest growing App user base of all time.
So what is ChatGPT? The short answer is an application based on GPT-3 tech (Generative Pre-trained Transformer 3), a large-scale natural language model released in 2020 that uses deep machine learning to sift through, analyse, and process internet data and sources to reference its knowledge memory bank. In turn, the chatbot provides answers to questions mimicking human writing patterns with the algorithm built to understand context and relevancy to generate human-like plausible responses.
What is already clear is that the power of GPT-3 seems to have shaped the future of AI wars between the technology giants. Microsoft, who own a significant stake in the company OpenAI, announced another whopping multi-million-dollar investment including supercomputer development and cloud computing support for the company. The investment is rumoured to be as much as 10 billion dollars. Moreover, in what appears to be a series of timely announcements, Microsoft went further announcing new AI powered updates to its Bing search engine and Edge browser, including some of OpenAI’s GPT-3.5 language technologies to be deployed through Bing.
This is a big deal: betting heavily on ChatGPT signals a significant shift in Microsoft’s business and with the internet overall. As these large language models start to permeate everything we do online, how we engage with the internet will change in ways we can’t really fully predict. The big picture for now is the multi-billion dollar industry segment that is internet search. And stealing ground on Google is clearly driving Microsoft’s carefully planned execution. Google, perhaps under market pressure, allegedly brought forward the launch of its own chatbot known as Bard AI based on Google's language learning model, LaMDA. It further announced a 300 million dollar investment in a firm called Anthropic, which is developing a rival application to ChatGPT. Other tech giants have launched their own proprietary versions: Meta launched AI chatbot Blenderbot in the US last summer and in China, Baidu says an advanced version of its chatbot Ernie will roll out as of March 2023.
All said, the technology is still comparatively in its infancy. And as with all emerging technology, one wonders where the data privacy risks lie. This is a question few seem to be asking so far. Let’s stick with ChatGPT here. The method that OpenAI uses to collect data is still to be disclosed. Though we know, at least here in Europe through past precedents, that obtaining training data by scraping data from websites is an infringement of both the EU and UK GDPR, the e-Privacy directive, and ultimately the EU Charter of fundamental rights. I’ve seen a few comparisons with the now infamous case of Clearview AI which built its facial recognition data base using images scraped from the internet and subsequently found itself on the receiving end of enforcement notices from multiple EU DPAs (including the UK ICO). The CNIL for one, found Clearview AI in breach of no less than six different articles of the GDPR. Sometimes overlooked, it is worth remembering that there are a significant number of articles and recitals within the GDPR that relate to AI and machine learning compliance requirements.
A number of the German data protection regulators have also started to voice their concerns over ChatGPT. The general – and all too familiar – stance being that whoever develops and controls the algorithm determines the results. In the regulator’s view, this can quickly lead to questionable or false results, political influence or racial bias and discrimination. Transparency of algorithms and the call to have advanced AI systems checked by independent third parties through institutional and or government action is now a growing mantra. Moreover, the European Parliament have recently questioned whether chatbots such as ChatGPT should be covered by Annex III (high risk AI systems) as it finalises its position on the EU AI Act. Arguably, and in true European fashion, the application of the ‘precautionary principle’ will continue to dominate legislative and regulatory debates as Europe seeks to identify and assess the perceived risks associated with the technology.
There will be many tomes written on the subject of advanced chatbots in the foreseeable future. Already there are philosophical concerns as to the impact of future usage of such AI applications potentially eroding our creative ability. We should by default be mindful not to underestimate the long-term impact of these new technologies which in themselves are only going to become more performant over time. We have to accept this evolution as a given and find the optimal pathways for society to be augmented by the technology, all the while preserving our humanity and our ability to engage in critical thinking.