Tag: chatbots
-
Hackers Simply Asked Meta AI to Give Them Access to High-Profile Instagram Accounts. It Worked
Meta decided to replace a lot of their tech support with a chatbot. Which meant giving that chatbot the power to manipulate data. Which meant, to the surprise of nobody whatsoever, that hackers tricked the chatbot into giving them access to other people's accounts.
-
Humans invented Wikipedia, which made accessing information highly-convenient, at the risk of questions about its authenticity1.
Then humans invented GPTs, which made accessing information even more-convenient2 at the expense of introducing hallucinations that can be even harder to verify and check.
Is humanity's long-term plan to invent something that spews complete nonsense that's simultaneously impossible to conclusively deny?3
Alt
Shonky MSPaint-grade graph showing ease of access increasing as ease of verification decreases, with a trend line going through Wikipedia (2001) through ChatGPT (2022) to an unknown future in 2043.
Footnotes
1 I'm well aware that in many subject areas Wikipedia routinely outranks many other sources for accuracy. But the point remains, because you've no idea what the bias of
randomuser123is; even if you check the sources they cite, you don't know what sources they omitted to include. I love Wikipedia, but I can't deny its weaknesses.2 Sure, ChatGPT and friends aren't always more-convenient. But if you need to summarise information from several sources, you might find them a more-suitable tool than those which came before. Why do I feel the need to add so many footnotes to what should have been a throwaway comment?
3 Actually, now I think about it, I'm confident that I can name some politicians who are ahead of the machines, for now.
-
Intermediary Protocols and Google Duplex
As Google show off their digital assistant's upcoming new feature of being able to make phone calls on your behalf, Dan asks what this means for the way we communicate in the future.