Falling ChatGPT Usage Shows AI Is Misunderstood

Usage is tailing off because people don't know what ChatGPT and products like it are for.

ChatGPT drew 590 million visits from 100 million unique visitors in January. That’s the fastest ramp up in the history of the consumer internet, according to some analysts. But how many of those people came away disappointed?

A lot, the data suggests. Mobile and desktop traffic to ChatGPT’s website worldwide fell 9.7% in June compared to a month earlier, according to figures from Similarweb.

Some of this is about privacy, restrictions, and regulation. Colleges across India, Australia and the US have restricted ChatGPT on internal networks and devices. The European Union will likely pass AI legislation soon. Scores of large companies have curbed employees using the technology on privacy fears. All of this has ‘shaken the faith in the AI revolution,’ Gerrit De Vynck of the Washington Post wrote in July.

It’s reasonable then to question whether generative AI is just another fad. From NFTs to blockchain, technologies that we were told would become widespread have so far failed to live up to their early promise.

But generative AI is different, at least partly because it’s already so embedded. Many of the technologies that we already use, from Gmail to Power BI, are powered by machine learning. Chatbots and Co-pilots are just a new iteration of something that already forms an integral part of the software on your computer.

Recent scepticism boils down to communication. People were expecting big things and found a chatbot that lied, didn’t give useful information, or found that the quality of responses actually declined over time.

GPT3 was probably released too early. Open AI and other large tech companies were ill-prepared to explain the potential. Usage is falling off because people don't know what ChatGPT and products like it are for – that they can learn to reason and share insights about the world, like what makes a good property development proposal or a tasty tart, rather than a templated back and forth conversation.

Several large companies have recognised the potential. Firms including property giant JLL have introduced their own ‘GPT’ products and, though we don’t think siloed products run by market participants are the right approach, it does demonstrate that the corporate world has a better understanding of how generative AI can transform its abilities. Employees can now leave the boring-but-necessary tasks to the Co-Pilot, while simultaneously using it to spot market trends or gather insights in a way that constitutes a creative superpower.

The technology is still nascent, so it naturally takes time for the best use cases to emerge. Where many view ChatGPT and other bots as a conversational version of Google search, the Fifth Dimension AI Co-Pilot has abilities that users are only just beginning to grasp. Imagine an investor that receives fifty potential opportunities a week. The Co-Pilot will draw up a shortlist of the most suitable opportunities by Monday morning. By extracting the structured data and applying it to the right investment criteria, it will also weight and score each opportunity. Investors can receive detailed reasoning as to the choices the Co-Pilot made, plus advice on how to approach due diligence. It’ll keep getting better, too.

This is a narrow example, but I think it illustrates the gap between what’s possible and how ChatGPT has been utilised so far. It’s also why we at Fifth Dimension AI and others like us need to make more noise, because confusion as to what this transformative technology really is will keep a lid on adoption, but not for long.