Last week, EY Chief Global Innovation Officer Jeff Wong joined Barron’s Editor at Large Andy Serwer for a discussion about the future of artificial intelligence (AI). At EY, Wong advises CEOs and EY’s own employees on how to understand emerging technologies and implement them into business processes. He shared some of his thoughts — and some interesting AI-generated videos — with the audience. 

“What we actually see in generative AI is they’re improving by about 10 times every 18 months. So, for everybody who’s used one of these generative AI systems — maybe ChatGPT — each release that comes out is far for remarkable and something that we haven’t seen before.” 

OpenAI’s large language model chatbot, ChatGPT, was launched in November of 2022 and quickly became the fastest consumer launch of a product in recent history.  Wong noted that immediately after it came out, “we [EY] sent a dozen accounting and tax tests through the system. The first version that came out with scored one out of twelve.” Six months later, OpenAI released a second version of ChatGPT, and Wong tried the test again. “The second version scored eight out of twelve. It passed a whole series of exams that we ask our people to pass [just] by us feeding [it] and it answering. These things are continuously improving in the back end,” he shared. 

While ChatGPT has had this early lead in the market for AI chatbots in the West, many Chinese AI companies have developed their own large language models. “Out in Asia, in China, they have to have their own series of large language model providers. While they didn’t come out early with it — those investments were more based in the US — they’re certainly racing along the way,” says Wong. Given the regulatory landscape in China, Chinese companies have to adopt Chinese language models into their business practices instead of American ones, which has given them a huge advantage in China’s growing market. 

Wong noted that while this kind of AI, which more and more companies are trying to integrate into workplaces, isn’t perfect — it is much more humanlike than people think. “One of the interesting interactions that we will all have with these new software systems is that we’re used to computers being directive, distinctive, having a certainty to its answer. We think of computers mostly as calculators...These systems, the large language models, are probabilistic, so it will give you different answers every time.”  

Among the greater challenges that AI presents is its inability to exclude bias from decision making. Wong is on the advisory board of a nonprofit called AI for All that is working towards having a diverse set of leaders who are writing AI algorithms and shaping policy around it. “It is very, very easy to miss the bias and some of the problems associated with [AI] if you don’t have those leaders in the room,” says Wong.  

Front and center in the conversation about AI is the ongoing New York Times lawsuit against Open AI, who the paper is accusing of using their work to train ChatGPT. While Andy Serwer initially thought that the case would be a “slam-dunk” for the Times, he noted that he understood OpenAI’s argument after a friend explained this: it would be illogical for Andy to be inspired by someone else’s work and then be sued for copyright infringement after writing a piece that involved no quotes or plagiarism. Wong followed up on this, stating that situations like the New York Times lawsuit incite some of the “interesting policy, philosophical, legal, moral questions that we have to ask ourselves” when it comes to the use of AI.  

Questions around the risks of AI have slowed its adaptation in the business world. A couple years ago, EY themselves rolled back the integration of AI into human resources and hiring decision making after realizing it didn’t account for bias. Wong warns that the challenges AI produce are significant, but he’s hopeful that policymakers and “organizations like the Asia Society” will keep up with the technology to ensure AI is integrated in productive and non-reckless ways. “I look to see communities like ours. Business community leaders, academic leaders getting together and helping put up some guard rails for ourselves and being thoughtful about how we use this technology.” 

While AI has its challenges, the demand for generative AI is higher than ever, and companies like OpenAI are struggling to find the processing power needed to meet demand while keeping usage costs low. “The more that we can make this available to more people, more companies, it’s the better,” says Wong. “You know the faster we can make sure this is in the hands of everybody, the better, because that will allow everybody to participate in some of the upside and intervention involved with this.” 

 You can watch the full conversation between Serwer and Wong here.