We have unfair expectations to blame for ChatGPT’s let-down

livemint.com:

ChatGPT was released in November 2022, it shot up to 100 million users in 2 months. The hype around it, however, shot up faster and higher. It was considered to be as big as the internet, search and the iPhone, and pundits compared it to electricity, even fire.

The New York Times called it “our Promethean moment," and one of its reporters felt that “we are not ready" for something so big. The early arrival of Artificial General Intelligence (AGI) was considered a given, with experts saying that there was a “non-zero chance" of achieving it within five years. The imminent demise of Google was blithely predicted, as a tweet by Dan Grover summing up this wild exultation: “I can’t get enough of this ChatGPT thing...

It feels like the first iPhone. The moon landing. Staring into the abyss.

I am not ready for this." Almost a year hence, the narrative has sobered down. Two recent articles, one by Gary Marcus and the other by John Thornhill, are especially illuminating. Marcus, to his credit, has been remarkably consistent in his views, and was perhaps the lone dissenter during the peak of ChatGPT’s hype.

In his recent piece (bit.ly/45KQDtC), he refers to an S&P survey where enterprises struggle with the “high costs and confusion" to deploy it. Consumers seem to be tiring of it too. Stack Overflow reports a drop in ChatGPT traffic, and the GPT4-infused Bing is still stuck with a minuscule 3% market share.

Legal battles are looming large, with a set of authors suing OpenAI and others for plagiarizing their content “at industrial scale". NYT is considering doing the same to make OpenAI pay for each infringement. James Thornhill’s FT article (bit.ly/44LWiyu) continues the scepticism, talking about ChatGPT’s hallucinations and