We seem to be in what I can only call an “AI lull.” The initial excitement about ChatGPT, which started in January, has receded. Google searches for ChatGPT peaked in April and are now down significantly, as is customer engagement with ChatGPT. Claude 2, a new and wonderful model released in July by Anthropic, hasn’t captured the public imagination the way ChatGPT did.
Do not be deceived. While the hype and marketing may have died down, at least on the retail side, the AI revolution marches on. And it will be just as transformative as advertised.
Every revolutionary technology has a period when it feels not so exciting after all. The dot-com crash in came in 2000, but even before that online commerce was decidedly meh. Only two years earlier Paul Krugman had observed that maybe the internet was overrated — and the thing is, in 1998, he wasn’t completely wrong.
Reaching even further back, consider the introduction of electricity into factories in the 19th century, which had many fits and starts over a period of decades. The printing press had a much larger impact on Europe in the 17th century, due to cheaper technology and paper, than it did immediately after Gutenberg’s invention in the mid-15th century.
As for generative AI: Among many of my friends and acquaintances, including my fellow academics, use has plateaued. They’ve tried it out, not always using efficacious techniques, and they are impressed (or not) to varying degrees. They view AI as a fascinating novelty — but separate from their work.
This is actually a healthy state of affairs for the development of artificial intelligence. We are being granted some time to process the advances in AI and to fathom that large-language models really can perform many tasks at high levels.
The stock market has also had time to digest this information, leading to much higher valuations for Nvidia, but there are no signs in market prices that AI heralds the end of the world, as some critics have alleged. A panel of superforecasters has rated the existential risk of AI as low, and fortunately most of the doom talk has faded into background noise.
All that said, the reality is undeniable: Generative AI and LLMs continue to advance by leaps and bounds. OpenAI is rolling out new commercial services, enabling companies and other institutions to fine-tune generative AI on their own data.
Google has announced something called Gemini, supposedly much more powerful than OpenAI’s GPT-4, possibly available as early as December. At some point GPT-5 is likely to arrive as well. There is now an Arabic LLM, produced in the UAE. That is a positive sign for broader global innovation and acceptance of AI tools.
Perhaps most significantly, open-source AI models are advancing at a rapid pace, even if most casual users don’t realize it. Many models were released or improved in August, with generally better results than might have been expected only a few months ago. Open-source AI creates resources that other programmers can build on, much as they can with the open-source operating system of Linux. Perhaps open-source AI will never dominate at the retail end of the market, but it will drive a lot of innovation.
In short, even while many users and commentators have moved on or become bored with AI, its transformational power continues to grow.
I can attest that LLMs continue to be discussed in universities, mostly because many students use them to cheat (and sometimes to work better within the rules). Yet the actual policy response from academia has been vague, as testing and classroom practices just haven’t changed much. Again that’s probably optimal because we are not yet ready for what might come next, whether it’s more oral exams or more computer exams co-created and graded by AI.
Not long ago I walked through Tysons Corner Center in northern Virginia, one of America’s largest malls, serving a prosperous and well-educated community. I didn’t see a single sign or marker referring to AI. No posters announcing the coming new AI-powered toys for Christmas. No AI-enhanced shopping app. No AI-assisted beauty salons. There were, however, a lot more stores selling high-end clothing.
I am not the target demographic for expensive and stylish shirts. But I might be more interested once AI designs the shirt and then recommends which one I should buy.
A message from Advisor Perspectives and VettaFi: To learn more about this and other topics, check out our podcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
Read more articles by Tyler Cowen
