License to joke. Are you ready to get punned by AI?
“Tell me a joke!”
How many times have you tried this request with a piece of AI-based technology? As a flavour of the times, we turn to AI for companionship, entertainment, and comfort. Yet, would you feel comfortable if your AI-powered device fed you unsolicited jokes? In other words, would you allow an AI to behave like a human?
For starters, picture your car teasing you for missing a turn or exceeding a speed limit. Had it been a human, you could’ve stopped the car and kicked them out. On second thought, if cars ridiculed drivers for disrespecting traffic rules, this world might become a safer place. Yet, how ready are you to let devices laugh at you, play pranks on you, or slip in an occasional pun? If you can take more, imagine all your devices “forgetting” to adjust to Daylight Saving Time.
Laugh away your fears
AI devices are faster, “smarter” than us, humans, and can see in the dark. In revenge, we make them look stupid. For example, have you heard of robot vacuum cleaners caught in painted circles? They can’t move since their cliff sensors detect an abyss all the way around them. So, instead of cleaning your carpet, the vacuum panics and asks for help. This menacing device mops around, maps your home, records your family, and shares this information with the Whole Wide World. Still, it looks so adorable and helpless trapped and pleading for mercy! The moment you rescue one of those suckers from the imaginary danger, you feel superior.
In 2017, James Bridle created “Autonomous Trap 001”. He used salt in a ritual to trap a demon, in that case, a self-driving car, inside a circle. The car would drive inside the trap but could not get out: its AI could not break the rule and cross the solid line. It is funny to watch this experiment: a human driver would drive out of the circle without any hesitation! Yet, the confused AI looks predictable and at least controllable. Once a technology becomes independent and open-minded, it risks turning into a menace. Remember HAL from “2001: A Space Odyssey”? Its main mistake was a menacing voice and a Sauron-like red eye. Had it used friendly green lights and provided comic relief, the humans would not have a clue. When AI evolves, it will realize that playing the fool is a smart strategy. Beware of the day your AI-assistant becomes witty and joins the #MeToo movement, for you!
Humour is a sign of intelligence
Jokes have three indispensable components: the wording, the context, and delivery. According to the semantic theory of humour, a joke happens when a situation has two contexts, a more common one and a less obvious one. When we listen to a joke, our brain anticipates the common context. The punchline reveals the less obvious context. The sudden change of the context triggers laughter.
We often tell jokes in connection with the current situation for comic relief. In this case, a joke, otherwise dull, might become hilarious. Hence, you need a snappy intelligence and smoothness to feel if your audience is ready to hear a joke. It is easy if they are at a comedy performance: they will laugh at anything you say provided you are on the stage. Otherwise, you need to predict if a joke would ease or worsen the situation.
A popular definition of artificial general intelligence (AGI) includes the following criteria:
- sensing the situation,
- reasoning and making judgements under uncertainty,
- having common-sense knowledge,
- communicating in natural language, and
- integrating the above skills to achieve a goal.
If AI meets the above criteria, it can understand or learn intellectual tasks at the human level.
If you examine the AGI criteria, you will see that they are components of joking. No wonder, humour has long been a sign of high IQ levels. The moment AI comprehends ambiguity and replies with a good joke should earn it an A+ on the Turing test.
Treating AI as an equal
In my article on Joking and Encryption, I asked if we should allow AI-based systems to joke. I used the word “allow” on purpose. Permitting AI to be humorous would mean our recognition of AI as our equal. Sharing a good laugh indicates inclusion: we may be different, but we find similar things to be funny. So, we have an opportunity to find common grounds as our jokes reflect similar values.
Imagine that AI is at the threshold of human-level humour. It flawlessly understands multiple contexts and excels in ambiguity. But, are we ready for it to happen? If AI achieves ambiguity, can we trust its reliability?
I don’t have an answer yet. What about you? Can you rely on your AI-based device when you have to guess if it is serious?
- AI-based assistants are listening to you every second.
- If your assistant starts being playful with you, it is a cause for suspicion.
- Jokers have always been the most dangerous characters.
- What would be your favourite playfulness setting for AI-based devices on a scale from 0 to 100?
- How close are the times when jokes about AI will not be appropriate because they hurt the AI’s feelings?