The advent of artificial intelligence products hitting the market has led many companies, commentators and investors alike to proclaim that we are on the cusp of a new chapter of humanity. The revolutionary technological progress promised by artificial intelligence has been compared with the printing press, the steam engine, electricity, computing and the Internet.
However, this utopian vision is completely at odds with the capabilities of the AI products currently being sold to corporations and consumers. Sam Altman’s ChatGPT, the most prominent and renowned AI product, displays no signs of intelligence, answering basic questions incorrectly which elementary school students would be able to get right.
I asked the latest version of OpenAI’s ChatGPT-4o to provide me with a list of US states ending in the letter ‘Y’, to which the chatbot responded that New York, Pennsylvania, Tennessee were all valid answers:
What is perhaps even more concerning is that when ChatGPT was asked the same question ten different times, it gave ten different answers, only one of which was correct. It appears to have a mind of its own, and not in the sense that Silicon Valley executives have proclaimed will usher in a new era of history.
When asked again, ChatGPT stated that ‘Wyoming’ was an acceptable answer to the prompt:
On another occasion, ChatGPT provided New York as an acceptable answer:
We asked ChatGPT again. Once again, we were provided with an incorrect list of states:
In ChatGPT’s fifth attempt, it answered with a much longer list of states supposedly ending in the letter ‘Y’, containing yet more wrong answers, including New York, Pennsylvania, Tennessee and Wyoming:
On its sixth attempt, ChatGPT correctly provided two states ending in the letter ‘Y’.
When asked for the seventh time, the chatbot produced yet another incorrect – and different – answer:
Sam Altman’s ChatGPT cannot decide which incorrect answer it wants to give for simple questions, and offers a different response each time. How can someone release a product this bad, while simultaneously claiming that AI will have a transformative positive impact on society?
Artificial “Intelligence” products on the market not only lack intelligence but are actually stupid. Who in their right mind would roll-out a product for public release when it is this bad? Only Elon Musk’s defective self-driving artificial intelligence software, Tesla Full Self-Driving, rivals the levels of stupidity displayed by Large Language Models.
Silicon Valley’s honeymoon period with AI will not last long. After hundreds of billions of dollars of investment and years of development, ChatGPT still cannot correctly answer a simple question that children will get right. How can we ever rely on AI to do anything important if we can’t rely on it, and everything must be checked by a human, then what’s the point? Why waste so many billions?