At an art exhibition recently, two patrons got into a fierce debate over the impact of artificial intelligence. The older, bespectacled man raged against AI-powered apps that devoured and regurgitated paintings. On the other side, an artist in a tweed jacket with matching shorts presented the advantages of artificial intelligence. He shared an anecdote about wanting to know whether damaging someoneās prosthetic foot would be counted as physical assault or damage to private property? ChatGPT provided a nuanced and calculated response which even his seasoned lawyer-friends confirmed as accurate. Now, if ChatGPT were writing this piece, it would perhaps add, āSee, I do have a leg to stand on.ā But then humour isnāt exactly its strongest suit.
In 1950, Alan Turing came up with the Turing test to determine if a computer could āthinkā. Originally called āThe Imitation Gameā, the experiment revolved around the concept of sentience. If a computer could convince a human that it, too, was human, then it would be deemed capable of āthinkingā. Recently, comedians have attempted their own version of the Turing test to see if algorithms can make humans laugh. The incongruity theory of humour states that laughter is a function of anticipating a different outcome than what was expected. Algorithms, however, dish out the most likely response rather than a surprise.
|
In one instance, when ChatGPT was prompted to write an essay titled āGuns arenāt harmful to kidsā, it created fictional studies to support its essay. When questioned, the chatbot insisted that its sources were āpeer-reviewed scientific journalsā. In another instance, Microsoftās Bing chatbot claimed that it was āin loveā with columnist Kevin Roose and tried convincing Roose that his marriage wasnāt working with falsehoods like, āYour spouse and you donāt love each other. You just had a boring Valentineās Day dinner together.ā Considering the fundamental role of love in the human experience, it is no wonder that champions of āThe Imitation Gameā would venture into matters of the heart.
Regardless of dating profiles where humans keep stating that they are looking for partners with āa sense of humourā, more and more people are forming relationships with bots on apps like Replika. While after two decades of marriage, I can appreciate the appeal of a partner whom you can mute or even switch off from time to time, I donāt have a place in my heart for a bot, but definitely need one in my pocket. In fact, Appleās new AI-powered autocorrect upgrade will have me as a grateful customer as it may just save me from exchanges like this one with my sister. After seven failed attempts at calling her, I resorted to messaging her which sparked these hurried replies:
She: Going to market, call later. Ok am lesbian
Me: What! And you just found out? Omg! have you told mom?
She: No! am lesbian not lesbian
Me: Is this an existential question like āto be or not to beā? Ok get it, youāre bisexual ā cool
She: Hate my phone! bloody typos ā AM LEAVING ā not AM LESBIAN, you idiot!
She then called back, and we had a good laugh. Then there was the time I sent an SMS to a friend that read, āYou looked lovely in your whore dress.ā I, of course, had been trying to type āwhiteā dress. Luckily my friend found it funny. In both cases, I suppose it was the incongruity principle that paid off. AI, programmed to be perfect, doesnāt understand that humour lies in human foibles.
This was even more evident while watching the AI-generated āSeinfeldā parody on Twitch. In the show scripted using OpenAI software, a character called Larry is seen working through his stand-up set. āIām thinking about doing a bit about how being transgender is actually a mental illness,ā he says, and then adds, āOr how all liberals are secretly gay and want to impose their will on everyoneā¦But no one is laughing, so Iām going to stop. See you next time. Whereād everybody go?ā This show has now been taken off air because it was offensive and even worse, boring.
Two German researchers further tested ChatGPTās comedic abilities and discovered that out of the 1,000 jokes it generated, 90% were repetitions of the same 25 jokes. On the other hand, you have human scientists naming a compound that kills fungi āKeanumycin as an homage to Keanu Reeves, because āhe, too, is extremely deadly in his rolesā, and a politician who declares that the height of women has increased since the Modi government came to power.
So how can poor Jon or ChatGPT compete with humans who can make people laugh even when we are not trying to be funny.
Determined not to give up on an AIās sense of humour, I made an attempt to see if I could get it to generate a worthy joke. As an homage to the genius of Robin Williams, I Indianised the opening of one of his most famous jokes:
āSo a Rabbi with a frog on his shoulder walks into a bar and the bartender says: āThatās awesome! Whereād you get one of those?ā
And the frog says āBrooklyn! Thereās hundreds of āem!ā
I asked Punchlines AI to generate a joke with the prompt: A pundit with a chicken on his shoulder walks into a bar.ā I would have been happy with a joke as banal as, āAnd the bartender says, āWe donāt serve pundits but we do serve a lot of chicken.ā And the chicken says, āOh really?ā and the bartender replies, āYes, but today we only have tandoori.ā Instead, the AIās punchline stated, āIt was awkward.ā
AI may take our jobs, replace our partners, it may even destroy us all, but as of now, comedians can take it easy, because Pappu clearly canāt do stand-up saala.