AI chatbots already work in Office

Share

Victor Miller [Archival audio clip]: : He asks what principles are most crucial to you, VIC?

VIC [Archival audio clip]: : For me, the most crucial policies focus on transparency, economic development and innovation.

Leah Feiger: It’s so strange. I have to ask, can the VIC have access to sources of information other than these public records? Let’s say an email from a conspiracy theorist who wants VIC to do something not so good with an election that would not represent its voters.

Vittoria Elliott: Great question. I asked Miller, “Hey, you built this bot on top of ChatGPT. We know that sometimes there are problems or errors in the data used to train these models. Are you concerned that VIC might internalize some of these mistakes or that there might be problems?” He said, “No, I trust OpenAI. I believe in their product.” You’re right. He decided – because of what’s crucial to him, as someone who cares deeply about Cheyenne rule – to feed this robot hundreds and hundreds and hundreds of pages of so-called supporting documents. The type of documents that residents will submit to the city council meeting. Whether it’s a complaint, an email, a zoning issue, or anything else. He handed it to VIC. But you’re right, these chatbots can be trained on different material. He said he actually asked VIC, “What if someone tries to send you spam? What if someone tries to cheat you? Send you emails and stuff.” VIC apparently responded to him by saying, “I’m pretty sure I can tell what’s a real problem and what’s spam and what’s fake.”

Leah Feiger: I would just say that a third of Americans currently do not believe that President Joe Biden legitimately won the 2020 election, but I’m glad that this robot is very, very confident in its ability to decipher the messages of disinformation here.

Vittoria Elliott: Entirely.

Leah Feiger: It was VIC in Wyoming. Tell us a little more about AI Steve in the UK. How is it different from VIC?

Vittoria Elliott: First of all, AI Steve is actually a candidate.

Leah Feiger: What exactly do you mean by “candidate”?

Vittoria Elliott: It’s on the ballot.

Leah Feiger: Oh, OK. No meat puppet?

Vittoria Elliott: There’s a meat puppet and that Steve Endicott. He’s a businessman from Brighton. He describes himself as someone who will attend Parliament and do human things.

Leah Feiger: Brilliant.

Vittoria Elliott: But people who go to vote in the UK next month will actually have the opportunity to vote not for Steve Endicott, but for AI Steve.

Leah Feiger: This is amazing. Oh my God. How it’s working?

Vittoria Elliott: The way they described it to me, Steve Endicott and Jeremy Smith, the creator of AI Steve, the way they described it is a large focus of the community. From the backend, people can chat or call Steve’s AI and can apparently have 10,000 simultaneous conversations at any given time. They may say, “I want to know when garbage collection will be different.” Or, “I’m worried about fiscal policy,” or whatever. These conversations are transcribed by artificial intelligence and transformed into the following policy positions that voters care about. However, to make sure that people are not spamming and trying to cheat, they will implement so-called validators. Brighton is about an hour from London, and many people commute between the two cities. They said, “We want people who are on their way to work to ask them to sign up for these emails as validators.” They will look through and say, “These are the principles that people think are important for AI Steve. Do you, a regular person who actually commutes to work, find this valuable to you?” Anything that gets more than 50% interest or consent or whatever, the real Steve who will be in Parliament will vote on it. They have a second level of control to make sure that everything people say as feedback to the AI ​​is checked by real people. They’re trying to make it a little more tough for them to play the system.

Latest Posts

More News