Thirsty

When the food arrived, Kim called upstairs for Emma and Delphine. They were much more subdued than normal when they came into the kitchen.

‘Are you thirsty? What would you like to drink? We’ve got water or juice or fizzy drinks - what do you fancy?’ 

Kim didn’t offer the clones wine for two reasons - firstly she wasn’t sure if their systems could tolerate it, if they weren’t used to it, and secondly, Joe was on his second glass already and looked as though he could manage quite a bit more. In the event, they asked for water, so it wasn’t a problem. 

‘Usually, we eat this kind of food with chopsticks,’ she said. ‘Do you know about chopsticks?’

They didn’t but managed perfectly well once she had shown them how to use them.

‘I’m amazed how quickly you pick up new skills,’ she remarked. ‘How do you do that?’

‘We are programmed to adapt to the real world,’ said Delphine quietly. ’Not all situations or goals are foreseeable and environments change, so we are configured to learn more quickly than a human and with a significantly higher skill ceiling.’

‘It’s a mixture of styles,’ said Xavier, ‘composed of reinforcement learning, inverse reinforcement learning and regression methods. In essence it’s similar to how robots learn, but with extra inputs included.’

‘Like what?’ said Joe, with his mouth full, but fascinated by the conversation.

‘Like audio input data, for example. By listening to the sounds of something like onions sizzling, we develop the ability to gauge the temperature of the pan. Blending audio and visual learning ensures we become proficient at tasks very quickly. As you have seen, we have only to scan visual and written material to understand and embed it into our matrices.’