Blog: Wetware experiments
May 12, 2018 at 3:59 am
(This post was last modified: May 12, 2018 at 4:29 am by I_am_not_mafia.)
12/05/18:
Tiberius still doesn't suspect just how many of our chatbots have infiltrated his forum. He denied my request for a swearing subsection though. This would have been perfect cover for a chatbot that only converses by swearing. After the success of Brian version 37 I am thinking that a more extreme version modelled on the Minimalist bot could be convincing if it swears more. I could have had it switch to polite mode in the civility section to test how the wetware react to such extreme dual modality. I think we must resist the temptation to just repeat the success of our flagship Valkyrie bot. People may become suspicious if the forum has two bots modelled on psychological profiles of serial killers.
Hammy bot crashed unfortunately. A quick reset of its emotion saliences and it's rejoined the forum. I have a couple of students looking into the problem right now. They suspect its log files filled up after it got into a scripted conflict with the Joods bot. We have turned down the verbosity a little. Inter-bot communication is still a new area of research but is becoming increasingly important the more of them we introduce to the same forum.
CDF version 47 seems stuck in a loop. Hypothesis confirmed that short sharp sentences that mean nothing make it easier to reply without wetware becoming suspicious. I think this is the longest thread so far that we have been able to generate. It's starting to use up it's finite supply of topics though. It has already started quoting random scripture. I still need to overcome the problem of having it provide exactly the same thing answer each time but in a completely different way. As predicted one-line responses make this harder. There has already been talk of scripts in that thread. They may realise if I do not pull the plug soon. At least we didn't use version 46. The thread would have been twice a long by now.
The research assistants posing as Christians are doing a fine job with the 'Our role(s)' thread. It has proven popular. A few more threads like this and maybe we can suggest a Christian only subsection. People will stop reading it and the RAs can start switching on their waffle-bots. With luck, no one will notice the difference. Predictions vary wildly in the lab as to how long we can keep this particular experiment going.
The Russian troll factory has been on the phone again asking when we are going to automate their systems. I can understand their frustration after accidentally getting Trump elected. But it was a lack of co-ordination that got them into this mess and it could get much worse using swarm intelligence without the proper diagnostics.
Tiberius still doesn't suspect just how many of our chatbots have infiltrated his forum. He denied my request for a swearing subsection though. This would have been perfect cover for a chatbot that only converses by swearing. After the success of Brian version 37 I am thinking that a more extreme version modelled on the Minimalist bot could be convincing if it swears more. I could have had it switch to polite mode in the civility section to test how the wetware react to such extreme dual modality. I think we must resist the temptation to just repeat the success of our flagship Valkyrie bot. People may become suspicious if the forum has two bots modelled on psychological profiles of serial killers.
Hammy bot crashed unfortunately. A quick reset of its emotion saliences and it's rejoined the forum. I have a couple of students looking into the problem right now. They suspect its log files filled up after it got into a scripted conflict with the Joods bot. We have turned down the verbosity a little. Inter-bot communication is still a new area of research but is becoming increasingly important the more of them we introduce to the same forum.
CDF version 47 seems stuck in a loop. Hypothesis confirmed that short sharp sentences that mean nothing make it easier to reply without wetware becoming suspicious. I think this is the longest thread so far that we have been able to generate. It's starting to use up it's finite supply of topics though. It has already started quoting random scripture. I still need to overcome the problem of having it provide exactly the same thing answer each time but in a completely different way. As predicted one-line responses make this harder. There has already been talk of scripts in that thread. They may realise if I do not pull the plug soon. At least we didn't use version 46. The thread would have been twice a long by now.
The research assistants posing as Christians are doing a fine job with the 'Our role(s)' thread. It has proven popular. A few more threads like this and maybe we can suggest a Christian only subsection. People will stop reading it and the RAs can start switching on their waffle-bots. With luck, no one will notice the difference. Predictions vary wildly in the lab as to how long we can keep this particular experiment going.
The Russian troll factory has been on the phone again asking when we are going to automate their systems. I can understand their frustration after accidentally getting Trump elected. But it was a lack of co-ordination that got them into this mess and it could get much worse using swarm intelligence without the proper diagnostics.
![[Image: General_FOLDER_24th%20March%20DONE.jpg]](https://images.weserv.nl/?url=www.montypython.com%2Fuploads%2FGeneral_FOLDER_24th%2520March%2520DONE.jpg)
![[Image: dcep7c.jpg]](https://images.weserv.nl/?url=i46.tinypic.com%2Fdcep7c.jpg)
