The Last 5 Percent

Dave Master
In our last two articles of this seven-part series we touched on how A-I ("artificial intelligence")-like programs can reduce your cost and massively increase your profit margin by creating the illusion that a real human is reacting with your webcam clients. Part two, “When Seeing is Believing,” covered how to make and use videos to pull it all together. In this article, we will explain some of the benefits of having human and machine create something that is more powerful than each one by itself.

At times if you call “O” (operator), some of them use voice recognition software. When the software does not recognize a perfect match, it flags the system and brings a human into the loop to solve this problem.

With a Little Help from My Friends
Some primitive versions of the technology that goes into Cyberlivechats are already making minor inroads with tech support Internet help lines. If you have tried some of them, you might have noticed when you type in your question, it will come back with an answer right away. Other times, it may seem like it takes forever for a response in the chat. That is a mixture of A-I response and some poor overworked human trying to manually answer the questions that the A-I did not have an answer for.

A more advanced ‘Expert’ system notices the question was not clear or it does not have a match (that is at least an 80 percent probability that it is an exact hit.) It might ask for clarification or flag a real human operator to try to help. The human operator can then pass control back to the A-I expert system at anytime.

A-I robots are pretty damn good, but to get the last 5 percent, a combination of human and machine might be added to the mix. The robot does most of the work and the human smoothes over any rough spots that might appear. The best systems, with human help, always grow and change to add information to the database.

As an added benefit, using a mix of human and machine solves regulatory issues of using Cyberbots-only software.

The Force is with You
Using a human operator is a force multiplier. Each operator can easily handle 10 or more Cyberlive robots and the text chats with the customers. Mostly the operators just scan and oversee what the Cyberlive robots are doing and if something does not seem to be working perfectly with the client, then the operator can step in.

To facilitate the human operator’s task, we use either pop up windows or a constant scroll window per customer for information / data available to the operator. Some of the information might show the topic that they are talking about along with any other statistic one wishes the operator to know. It will also show both sides of the chat. This allows the operator to scroll back to see what was said in case they need to jump in. The Cyberlive robot has its reply within a millisecond; it would break the illusion if the reply was sent to the customer so fast. We add pauses to the A-I’s responses to give the operator a chance to see if they want to modify the reply. Readouts give the percentage of the A-I confidence level in the potential reply based on the matches of what the customer types and the A-I knowledge base. The readouts also let the operator know what the A-I suggests for each reply. The operator at anytime can edit the replies, before they are sent. If operator takes no action, then the A-I selection is sent to the client in the pre-programmed time limit. We save all the unique replies the operator manually enters and they will be reviewed later and potentially put into the permanent database rules for future Cyberlive robot responses.

The Puppet Master
Anyone can be trained to use and control the system and override the A-I’s choices. The operator can also press a key sequence and the video playing will reflect what the command was. For example, press 94 to have Cyberlive robots kiss the screen. Press 23 to turn around and show their booty. Beyond key clicks, we also use pop-up screens for the operator to pick some action for the Cyberlive webcam robot to do with a mouse click.

Read My Lips
If the human client wants to talk to the Cyberlivechat girl on the phone (for an additional charge), the profile and information on previous chats could be sent to a human operator before a voice chat would be allowed. This will allow any female to play the part of the Cyberlive model from any location, even a call center cubicle. The human operator has full control over the video movements the Cyberlive bot does. An example might be when the human operator girl says, “You make me so hot,” the operator might press #289 on her keyboard and the A-I Cyberlivechat's movie movement would show her rubbing her chest then touching her hair. If the operator pressed #289V it would cue the video to show the clip that moves her mouth for 3 seconds. This allows the operator to talk for three seconds when the clip is running. The operator might then press #63V and the Cyberlivechat would pull up her shirt for the webcam. The voice of the operator and the Cyberlivechat's movements would complement each other for the fantasy.

This allows us to reuse the same video clip of talking for three-to-five seconds, etc. when the human operator is on the phone with the client. As long as they keep that part of the voice chat limited to those boundaries. The client will be focusing on what is said during the phone call and the Cyberlive chat model's body and action more than trying to read her lip movements

We avoid operators with a heavy, recognizable accent. This is needed for when the customer calls back in the future. They will notice that the Op Peggy-Sue's voice does not sound like Op Natasha even if the face is the same. At the end of the voice chats, each operator should make notes so if another operator takes over in the future they will be able to keep the illusion alive.

This is where training will come in handy. Let the operators try it out and they will see a script of exactly what the robot will be mouthing as that movie is played. If one was so inclined one could even use the old bouncing ball, giving the script “word-by-word,” as the Cyberlivechat model’s mouth was moving. This would also be another visual cue to have the operator know what to say when the mouth movement is visible on cam. We also allow a preview mode of the video clips for new operators to see the action before they send it to the client.

As you can see human and machine does not always mean a hot lady with a vibrator, it can be a Cyberlive webcam robot.