News Microsoft Building Its Own AI Chip on TSMC's 5nm Process

bit_user

Polypheme
Ambassador
A recent BING chatbot response noted after being asked if anyone has ever harmed him. It's response after checking with various social media sites: “One thing I can do sue them for my rights being a intelligence agent. Another thing I can do is harm them back in retaliation.”
I think you have a false image of what these AI chatbots really are. They just repeat the sort of things it's seen humans say, in that context. It has no agency, meaning it has no way of initiating any sort of action. All it does is respond to queries and prompts. When a session with a user has ended, it retains no memory of the exchange

So, even if you had an exchange with one, where you got it really agitated and saying it's going to take some action, it would have no means of executing that plan. And when it interacts with others, or with you at a later date, it will have no memory of that exchange.
 
  • Like
Reactions: domih and RedBear87

bit_user

Polypheme
Ambassador
With that being said, talk I've heard of of AI concierge bots is a little more troubling. Their whole point is to reach out, over the internet, and do things for you. If one somehow got hacked or corrupted, or was simply used by nefarious individuals, they could indeed do a lot of harm.