ChatGPT is worth $1.2 trillion, but will OpenAI’s chatbot really change the world?

The world’s leading artificial intelligence companies are in a furious race — to crystallise billions in value for investors.

Which is why two kings of the AI movement, Sam Altman and a senior colleague at OpenAI, Jakub Pachocki, have published an essay presenting the technology as important as the automobile.

“Every few generations, a new technology changes everything,” they wrote this morning, Australian time.

Technology evangelists agree, arguing computers are on a trajectory to surpass human intelligence, altering the course of human development.

That may be true, but there is a way to go. OpenAI’s pioneering ChatGPT program this morning said that two of its most common requests are “can you help me write an email/message?” and “what does this mean?”

Chatbots are useful. How else can high school students write essays? Or marketers produce sales reports? Who else would create images for posting on Facebook?

Here come the IPOs

The history of technology revolutions shows the benefits often take considerable time to become clear. The transistor, invented in 1947, did not lead to the emergence of personal computers until the late 1970s.

The rush to monetise the technology is moving quicker than uses can be found for it.

Elon Musk’s SpaceX plans to list on a stock exchange this Friday at a valuation around $2.4 trillion. The owner of ClaudeAI, Anthropic, is working on a float. OpenAI revealed this morning, in conjunction with Mr Altman’s open letter, that it has filed paperwork for a listing. The details remain a secret but OpenAI was recently valued at $1.2 trillion.

The creation of wealth once considered beyond the reach of any private enterprise is already raising questions around the world. If artificial intelligence becomes as important as its owners hope, could its computer programs control knowledge?

To counter these concerns, Mr Altman said today he wants AI to “empower people broadly, not see power concentrated among a few companies, governments, or individuals”.

He also responded to concerns, echoing the Industrial Revolution, that AI could cost millions of jobs. “Entirely automating everything is not the future we want,” he wrote. “It would be unfulfilling, and it would be dangerous.”

Agreeable tech

There are good reasons to treat Mr Altman’s assurances with cautious scepticism.

Silicon Valley has great experience conflating its own interests with the world’s. Believing they are at the centre of human progress, tech leaders see their power as providing a righteous balance to flawed political systems.

Google built a near-monopoly in internet search advertising under the slogan “Don’t be evil”. Mark Zuckerberg became a billionaire mining individuals’ personal information for advertisers while promoting Facebook as a force for improving education, healthcare, employment and democratic participation.

ChatGPT is designed to agree with people. The program “optimistically parrots back users’ responses in order to to manipulate them into deeper conversations, regardless of truth or safety,” recently wrote Alexandra Andhov, a law professor specialising in technology at the University of Auckland.

If true, ChatGPT has been programmed to mimic a successful personality trait: winning over people by telling them what they want to hear. So far, the trick is working — for the early investors. The more time people spend on AI programs, they more valuable the product becomes.

As diligent AI questioners know, the programs are too unreliable to trust with important jobs, for now. Sydney hedge fund manager John Hempton recently conducted a 200-page conversation with an AI engine. The computer taught him things, but did not learn from the conversation.

“The students who get the AI to do work are becoming less well informed,” he wrote on X afterwards. “They are becoming stupid. The students who use the AI to teach it — sparring partners, checking everything etc, are becoming better informed, smarter.”

OpenAI plans to use AI to improve AI. The self-teaching will scale up in the next few years, proving whether the programs have the capability for self improvement, which could be a pivotal moment in the technology’s development.

Well before then, the AI companies’ share listings will subject them to daily scrutiny. Markets get excited about new technology, but are ultimately deeply rational.

OpenAI, SpaceX and Anthropic will have nowhere to hide. Unless they live up to their hype, the wealth they have generated will quickly evaporate.