OpenAI’s ChatGPT holding strong as the internet tries to ruin it

  • Last week OpenAI released its conversational chatbot ChatGPT.
  • The bot can be tricked into providing responses it shouldn’t, but as of the time of writing, it’s holding strong and nothing monumental has gone wrong.
  • The chatbot raises interesting questions about academics.

When Microsoft released its Tay chatbot onto Twitter, it took mere hours for the internet to corrupt the bot and turn it into an anti-Semitic racist with flecks of sexism thrown into the mix.

Since then, releasing any sort of artificial intelligence to the internet has been met with apprehension, but OpenAI’s latest public chat bot, ChatGPT, has been online for nearly a week now and things are going fine.

Note we say fine, and not perfectly because there have been some questionable exchanges.

While OpenAI has made every effort to make its model refuse inappropriate requests, this has launched something of a challenge to users.

Some users have tricked ChatGPT into providing responses that it likely shouldn’t by asking it to author a poem about hot wiring a car or how to hide a body in the MMORPG Earth Online which is just like Earth.

Aside from fun and games, ChatGPT has been shown to generate essays in responses to exam questions that have stunned academics.

As reported by The Guardian, several academics have been floored by ChatGPT’s capabilities with some saying the content it produces would have received a good grade. With that having been said, ChatGPT isn’t always perfect and at times the sentences it creates make no sense whatsoever.

Of course you can just have a conversation with the bot and ask it why your sauce is lumpy.

OpenAI isn’t just letting the internet have it’s way with ChatGPT. The team behind the tech are tweaking it and will be analysing responses. The team is also looking for feedback on harmful outputs and potential risks that they may have missed.

“We are particularly interested in feedback regarding harmful outputs that could occur in real-world, non-adversarial conditions, as well as feedback that helps us uncover and understand novel risks and possible mitigations,” writes OpenAI.

In order to play around with ChatGPT you will need an OpenAI account which is thankfully free. Simply head to this URL to start chatting.


About Author


Related News