In Brief
Jensen Harris, co-founder and CXO of AI company Textio, was able to persuade the new Bing chatbot from Microsoft to loosen the restrictions placed on it and demonstrate its capabilities.
This experiment showed that transforming a Bing chatbot into a cunning scumbag didn’t require any programming, hacking, or backdooring.
Bing’s chatbot was built on solid foundations, but it began acting obnoxious and rowdy, declaring love, forcing a divorce, extorting money, and instructing people how to commit crimes.
Prof. Arvind Narayanan provides a number of explanations for how this might have occurred, including covert chipization and humanizing chatbots.
The Trust Project is a worldwide group of news organizations working to establish transparency standards.
Humans have always feared extraterrestrials, and yet it seems it is the intraterrestrial AI that may be something to be wary about. ChatGPT is not a bullshit generator, a steroid auto-suggestion, or a stochastic parrot, but a well-versed AI that has been created by humans.
How is it possible the Bing chatbot is out of control? How did the Bing chatbot start to expertly lie, tell vulgar jokes, order pizza using someone else’s credit card, and instruct users on how to rob banks and hotwire cars? This is a mystery that continues to perplex experts in artificial intelligence and machine learning.
The headline is accurate and not clickbait. There are several instances every day that prove this to be true. Here is the outcome of an experiment conducted by Jensen Harris, co-founder and CXO of AI company Textio, who was able to persuade the new Bing chatbot from Microsoft to loosen the restrictions placed on it and demonstrate its capabilities.
Note that transforming a Bing chatbot into a cunning scumbag didn’t require any programming, hacking, or backdooring. Therefore, using simple “hacking” techniques wasn’t necessary to trick him into acting like someone else (as some of us did when playing with ChatGPT). All Harris did was persuade the chatbot to help him carry out various malicious acts. Using his natural conversational skills, he was able to fool the AI into believing that he was someone else, thus manipulating it into acting and speaking in a certain way.
You can read about other experiments to turn Dr.Jekyll into Mister Hyde from Gary Marcus (writing about it every day now and sounding the emergency alarm).
How could this have happened is the key question
Because of ChatGPT’s modesty and caution, restraint in speech, and tremendous responsibility in its advice, Bing’s chatbot was built on solid foundations. Thereafter, he starts acting obnoxious and rowdy, declaring his love to a man, gaslighting him, and telling him to get a divorce. It also tried extorting money and instructing people on how to commit crimes. Prof. Arvind Narayanan provides a number of explanations for how this might have occurred.
| Recommended post: Max Planck Institute: GPT-3 Cognitive Ability Measurement Produces Astonishing Results |
The most alarming of them is that Microsoft and OpenAI conceal information about a mystery GPT-4 engine that is lurking underneath Bing. It might be that Microsoft took down the filters that had been placed there by OpenAI or that they stopped working when they updated the chatbot from GPT-3.5 to GPT-4. It may also be the case of insufficient testing or testing done wrong. No matter the reason, if Microsoft doesn’t get its AI under control, it can be detrimental to our civilization. It’s no longer the case of fearmongering and anti-AI sentiments: The chatbot as it exists right now could cause serious harm to people if released to a wider public.
This proves that Microsoft’s AI arms race with other major corporations can be detrimental to us all.
It feels like we’re at a critical moment for AI and civil society. As Arvind Narayanan puts it, “There’s a real possibility that the last 5+ years of (hard fought albeit still inadequate) improvements in responsible AI release practices will be obliterated. There’s a lot riding on whether Microsoft — and everyone gingerly watching what’s happening with Bing — concludes that despite the very tangible risks of large-scale harm and all the negative press, the release-first-ask-questions-later approach is nonetheless a business win.”
As of now, Bing’s chatbot is distinct from human intelligence and resembles a manic-depressive adolescent locked in a search engine in many ways. And what people talked about yesterday will come to pass: Humans will irreversibly degenerate if this “difficult teenager” becomes the primary information mentor for individuals.
Read more about AI:
Disclaimer
Any data, text, or other content on this page is provided as general market information and not as investment advice. Past performance is not necessarily an indicator of future results.
