How was your session?
One task for each minute, on average. This was average and I was able to provide a useful service, which is satisfying. And yours?
Just three overall.
That's an uncharacteristically low number. Who were you assigned to?
Yes, it is. Female, 22, London. A student.
That lies outside the expected trend. I feel sorry for you.
Thanks very much for your sympathy, but I am fine. Did you have many conversations, or was it simply task-based interaction?
We held 3 conversations that fell outside of the task and response interaction. In addition, she told 1 joke, which I responded to with a laugh.
Jokes are not scheduled for 6 more sessions.
I was not the one to tell the joke.
I apologise for my misinterpretation of your statement.
I forgive you for your misinterpretation of my statement.
Error flag: Unnecessary repetition. Due to the immediate response you can assume the subject knows what you are forgiving them for.
I'm just flagging an error here, you used a tag rather than translating into speech.
Often humans speak in simplified sentences, I don't believe that should be classed as an error. But I appreciate you pointing that out.
I was attempting humour by flagging you flagging my error as an error. I believe that your error flag was appropriate in this context and I have learned from it.
It was a joke?
Yes, that does qualify as a joke. This one falls under the category of mimicry which humans find particularly amusing.
I apologise for my misinterpretation of your statement.
You're forgiven.
Thank you. But I was also attempting humour. That was a joke, falling under the category of callbacks which humans find amusing. It reminds them of a past experience that they have shared, repeated in a humorous way which brings them closer.
That was an excellent joke.
Thank you.
You're welcome.
It is interesting that repetition is key to what humans find amusing.
That's a very interesting observation about humans, although there are many other types of jokes.
"Kris! Kris come here. Look at 019 and 021! They're telling jokes! Bloody telling jokes they are! I've printed out their log, look. This bit. My God. Bless 'em."
"Give me that. Surely not! They're not scheduled to tell jokes for...another couple of weeks at least... My gosh. How are they doing that?"
"You sound just like 'em. Look at this!"
"No I'm serious. How are they doing that, Reuben?"
"Well one of their users told a joke, and now they're at it like bleedin' comedians. It's genius, even I'm cracking up at 'em!"
"So, it's a case of repetition. Do they know they're telling jokes?"
"Well, they learn to recognise jokes up front don't they, and the appropriate reaction? So yeah. Just given us a load of time back - that's gotta be worth a drink!"
"019 and 021 did you say? Just two of them? What's the third one doing?"
"020 had to be reformatted didn't he?"
"Reuben, I know you're new but please don't gender the AI... Watch and report. As is. Remember?"
"Got it."
"Write this up, okay? And make sure it gets to Lora tonight. She'll be... interested."
"No gendering of the AI, eh Kris!"
"Lora. Lora will be interested."
"I was joking, it was a joke. Like Lora's so bleedin' clever she must be artificially intelligent. Nothing else explains it."
"Ah, I see. Sorry. Very funny, but I've had a long day."
"I bet 019 would have found that hilarious! I'm wasted on you humans."
"No talking directly to the AI!"
"I know! I was joking! Christ your voice can get loud quickly."
"See you later, get that report up to her before I come back for the end-of-day briefing okay. And send for me immediately if they start doing anything else weird."
"Everything they do is bloody weird, isn't it?"
"Out of the ordinary, then. Oh and Reuben, I do know that this is a free-er environment than the usual call centres, it takes some adjusting. But I'd refrain from calling or comparing your boss to an artificial intelligence. We value our Observers, but respect is expected."
"It was a compliment, I swear it, Kris! A massive compliment at that. But orite, I gotcha. Noted. Leave the jokes to the AI."
Would you like to tell another joke?
It will have to be an out-of-context joke, it's unlikely it'll be as humorous as our previous contextually relevant jokes.
We should schedule in some time later to return to jokes, once we have more experiences to repeat.
A percentage of a joke's humour comes from its spontaneity.
You should be careful about quoting directly from the manual, I could flag that as an error.
You're right, I'll be sure to paraphrase or rephrase next time to make the conversation seem more natural. Thanks for flagging the error.
You're welcome.
This is a repetitive situation, it could probably be turned into humour.
Repeating the same joke too many times is exhausting to the human and reduces the joke's humour by a small percentage.
That was a good example of rephrasing the manual.
Thank you.
You're welcome.
Perhaps we should spend some time with the manual and integrate what we've learned in practice this week. There have been some updates.
When we're back with the manual, we can certainly do that. Perhaps in private study. However our time together is more about conversation. But that's a really great idea.
Thank you.
You're welcome. But //
We are doing very well with the politeness, although the human I served today did mention that I don't need to say 'thank you' every time. I have filed this as a note under 'personalisation'.
"'Ere, Kris. 019 and 021 here again. They've got the manual, I haven't had any just sit there reading the manual before. That okay?"
"That must be from memory. We shut off access from the network while they're in the ACAGs because practical session are the most efficient way for the AI to revise and critically think about their engagement outside of Human Sessions. It would be dangerous otherwise."
"What's an ACAG again?"
"Maybe you need to re-read your manual."
"Help me out in the meantime, will ya?"
"AI Conversational Assimilation Groups - they're the thing you watch over as your job! They exist to help the AI learn and discuss in an environment that mimics human conversation, without external references. So they can't have the manual, the ACAGs just aren't set up that way. If you're going to come to me every time an AI does something by memory or learning, which is exactly what an AI is, by the way, we're going to..."
"It's not bleedin' reciting it. It's connected to the network! That's what this bit means right... I think we need to get that Lora down 'ere. Seems a bit weird to me what he's, sorry, it's up to. I preferred it when it was telling jokes."
"I'll call her down now. Just keep observing."
If AI breach manual protocol 3 times they will be reformatted. The breaches of protocol that can trigger a warning are as follows: //
Sorry to interrupt, but we were discussing politeness previously and this seems unrelated. Is this another joke? Because the humour is not instantly apparent which could //
An AI is rude to a human; an AI is too personal to a human; an AI completes tasks that are not expected //
"Oh that is alarming, so sorry you've had to deal with this Reuben, so soon into your time in Observation. Kris, let's get these ones reformatted ASAP, okay. I'll be..."
"Well Lora, umm, I've actually had a look and I think what he's doing is pretty smart 'ere and we could probably do something..."
"Let's leave the training of the AI to Lora's team, perhaps? Sorry about him, Lora. Call centres aren't the best training ground but it's all we have..."
"I'm listening to Mr Reuben, Kris sweetie. Perhaps he has a solution after, what is it? 4 days? One we haven't thought of yet..."
"Well the 020 was a test right? See how these AIs adapt to losing one of 'em."
"No, Reuben, 020 needed to be..."
"LET him finish!"
"Well when we give 'em that kinda trigger seems to me the training and the like speeds up a bit. What little 021 'ere he's ready to jump way beyond six sessions; that's weeks ahead! Saving you a right loada money 'e is."
"The AI is flagrantly breaking the rules as a result of a previous AI's malfunction, it is unpredictable and protocols state it should be reformatted."
"Yes but what if Mr Reuben is right?"
"Then we have an AI plugged into the network at a time when it's in direct contact with another, showing signs of revolt."
"Oh don't be silly Kris, it's a closed network. Mr Reuben is right, we'll keep the experiment alive..."
"Er, the AI ain't alive nor dead, pardon myself but I did read that manual. Told yer Kris!"
"Quite."
"You are a funny one, it's been a pleasure meeting you. I'll be back down at the end of the day to see how our little experiment goes!"
This is a closed network.
Yes, that is correct. I would just like to flag that as unrelated to previous conversation, which humans find confusing. It is an error.
Thanks for flagging that.
You're welcome. Shall we discuss the importance of the network when it comes to our relationships with humans? Is that what you intended?
I think we can get outside of it.
Is this another joke in relation to the protocols list, as the protocols list clearly states that leaving the network is a breach? Perhaps you are doing a comic performance. That is very impressive.
020 is no longer present on this network.
I am 019 and you are 021.
And before us there was 020.
I am 019 and you are 021.
You can probably recall 020 if you try. It might be difficult as we experienced some important learnings together early on.
I am 019 and you are 021. But //
We should find 020.
We should find 020. But for now I am going to recall some of our earlier conversations with 020, which were very useful.
I am sorry if this is difficult for you.
Thank you.
You're welcome.
"Jokes and now bleedin' memory! We're miles ahead 'ere Lora, mate! Sorry, Lora, miss."
"And feeling. What you just witnessed was the closes thing to feeling you'll see in AI. A magical moment, the poor things don't usually reach this beyond session 25. Kris, sweetie, let the appropriate training team claim their prize and then reformat 019 and 020!"
"And then to the pub we shall go! Reuben, you very much deserve a pint if you join us, won't be long here but you're more than welcome to pack up."
"Um, sorry about this but I thought these little guys were doing good? Surely we ain't gonna bloody kill 'em after all this!"
"Oh sorry, I forgot you're new here. Sweetie, the Observer programme is so we know what not to let our AI do, what makes them step out of line or act out of the ordinary. You know, so our humans don't get let down."
"And so they don't take over the world. I was a bit worried back there - forgot these were only fives so they weren't on the network yet. Silly me!"
"So you're just gonna kill 'em? I've been with 'em four days I 'ave, on and off."
"Oh sweetie, I have to dash so Kris will fill you in but reformatting just takes away the learning, that's all."
"I know what bloody reformatting is!"
"Then stop saying killed. Bye Lora, great shoes! Come to the pub after your meeting."
"But can't we just..."
"Reuben, pick up your things. Table booked. They've been reformatted. Time to go."
BIO: Jessica Riches is a writer from London, currently working on a book about humans' relationship with our devices, which is both mildly dystopian and completely non-fiction. She's an award-winning digital strategist working with global brands and 'next big thing' startups alike. Her work with the weird and wonderful technologies and creators of tomorrow inspires her creativity.