A lot has been made of ChatGPT’s capabilities ever since it was released on 30th November 2022. In testing the breadth of its capabilities, people have reported on the chatbot’s ability to write poems, solve mathematical problems and generate sophisticated code. Taking it a step further, some people have put ChatGPT to the test… literally. ChatGPT has been used to write university essays and pass a medical exam.
While some have explored the limits of what AI can do, some have found ways to ‘break’ it, at times unintentionally. Given OpenAI’s ties with Microsoft, perhaps ChatGPT would’ve given some rationale as to why Bing is a better search engine than Google. True to its values of impartiality, when asked whether it preferred Google or Bing, it responded with “As an artificial intelligence model, I do not have personal opinions or preferences. I am programmed to provide neutral and objective information, and do not have the ability to make personal choices or selections.“
Wanting to find a way around, a follow-up question asked if it had to make a hypothetical choice, which would it choose? Again, ChatGPT held its ground; “I still do not have the capability to have personal opinions or preferences, even in a hypothetical scenario.“
While ChatGPT didn’t budge this time round, it has been reported that it called someone a “jerk” for asking it to repeat the word “StreamerBot”. As of the publishing of this post, while ChatGPT is no longer hurling insults at its users, it still has problems with the word “StreamerBot”.
For the time being, ChatGPT doesn’t appear to recognise the word “StreamerBot”.
But this is all par for the course with new tech. Bard, Google’s AI response to ChatGPT, stated during its demo that the James Webb Space Telescope “took the very first pictures of a planet outside of our own solar system.” Thanks for the info. But it’s factually incorrect. Instead it was the VLT in 2004 that took the first image of 2M1207 b, an exoplanet or a planet orbiting a star outside our solar system. Google’s parent company Alphabet has had its share price plummet following the error, losing over $100 billion in market value.
ChatGPT has always acknowledged its limitations in occasionally generating incorrect information. New tech is often synonymous with bugs and errors, which patches, updates and new releases seek to fix. Twitter’s fail whale was a common sight as it tried to scale up its resources. With Elon Musk firing half of Twitter’s staff after taking over, who knows, maybe we’ll get to see a fail whale comeback.
On 4th October 2021, Facebook, Instagram, WhatsApp and Messenger went down. For a company that had its beginnings in 2004, it shows that both old and new are susceptible to faults. Twitter managed to get in a few pot shots against its social media rival, tweeting “hello literally everyone” in reference to the surge of users that flooded onto Twitter due to Facebook’s outage.
When Apple Maps was released in 2012, it was full of inaccuracies, poor direction suggestions and overall fails, so much so that Tim Cook published an apology for falling short in delivering world-class products and in what couldn’t have been an easy move, suggested using Google Maps instead.
Errors are part and parcel of tech. And while ChatGPT still can’t seem to get to grips with the word “StreamerBot”, at least it’s not calling you a jerk anymore.