1696: AI Research
AI Research |
Title text: Lambda calculus? More like SHAMbda calculus, amirite? |
Explanation
Developing Artificial Intelligence (AI) has been a challenge for a long time. Even to develop one with the intelligence of a six-year-old child would be a great milestone, and presumably a stepping stone on the path to making one with the intelligence of an adult human.
In this comic, Randall/Cueball jokingly suggests that in order to accomplish this goal, one can give him an AI that's already as smart as an adult, and let him teach it childish and stupid things. He is shown teaching it dumb jokes, much like the ones a sassy six-year-old would make, as the first "fart" joke where artificial is changed to fartifical.
The humor in the comic is that Randall is essentially accomplishing the present goal of a six-year-old AI by starting with the final goal, which is a full human intelligence, and making it dumber, just by teaching it poor humor. This is not unlike the old joke, "The easiest way to make a small fortune on Wall Street [or similar] is to start with a large one."
The specific situation may also be a reference to Tay, a Microsoft chatbot that was taught to troll within hours of its exposure to the public.
"Updog" refers to a light-hearted practical joke in which the perpetrator casually uses the neologism 'updog' in a sentence ("Hey, I'm going to get some updog, you want any?"). The unsuspecting listener is expected to be curious about the meaning of the neologism and ask the perpetrator its meaning, specifically in the format "What's 'updog'?", inadvertently invoking the highly casual greeting of "What's up, dawg?". The perpetrator then draws attention to this by replying along the lines of "Not much, you?", causing the target to realize the foolish thing they just said. Other neologisms used in the context of this joke include 'updoc', 'snoo', and 'samatta' ("What's up, doc?", "What's new?", and "What's the matter?", respectively). Updog is mainly an American joke not particularly well known in other English-speaking countries.
In the title text there is a joke on lambda calculus, where lambda is changed to SHAMbda. Lambda calculus is an area of mathematical logic and theoretical computer science. It is a formal language which can express computation and evaluation. It is Turing Complete, which means it can do any computation which can be executed by a computer. However, it is very simple, consisting only of two primitive notions: abstraction, which is forming a function and application which is applying a function to an input value. For example, a function which squares a given number can be written λx.x². Here the λ indicates an abstraction (hence the name lambda calculus), the x is the input value and the output is x². As an example of application, if we apply this function to 5, we get (λx.x²)(5) = 5² = 25. The title text makes fun of this by inserting the word "Sham" into the phrase, a word used to describe a trick or con; essentially, it denies that such calculus is useful or valid.
The title text finishes with amirite, short for am I right? which is often used to finish sentences on web forums, to prevent anyone saying you are wrong. Not very mature to use in a serious discussion, so very fit to use for a AI that tries to emulate the intelligence of a six-year-old.
The AI tip in the caption relates to the Protip category.
Transcript
- [Cueball (representing Randall, or at least the person who wrote the caption) is standing in front of a computer console displaying AI, talking to it. The computers reply is indicated to come from the console with a zigzag line, rather than the straight lines for Cueball.]
- Cueball: Then you say "More like fartificial intelligence!"
- Computer: Understood.
- Cueball: Great! Now let me teach you about "updog".
- [Caption below the panel:]
- AI tip: To develop a computer with the intelligence of a six-year-old child, start with one as smart as an adult and let me teach it stuff.
Discussion
I'm Australian an I don't know about the "updog" Thing, just sayin' 162.158.2.139 05:15, 20 June 2016 (UTC)
- I'm British and I haven't heard of "updog" either. 141.101.98.18 08:29, 20 June 2016 (UTC)
- Never heard of that in the two years that I lived in England and South Wales. 141.101.93.49 09:06, 20 June 2016 (UTC)
- Saw "updoc" (same joke) on an episode of Scrubs once. 141.101.98.137 11:34, 20 June 2016 (UTC)
- I think we've pretty much confirmed that "updog" and similar is not well known in most English-speaking countries. Yay! I was correct after all. (I wrote the first draft for this explanation, which included the question of whether it was or not.) 108.162.219.59 07:01, 22 June 2016 (UTC)
Ok. Added a note about how developed AI tech. is nowadays. That one needs a citation. Also I believe a paragraph about the difference between AI and how computers generally work is called for. AI is built on neural networks to mimic the way human brains work. Computers have a more simplistic design, although it works wonders for number crunching and following programs. This contrasted with AI, which can figure out things on its own (learn) and not having to be told everything. Todor (talk) 17:17, 20 June 2016 (UTC)
- Those stupid chatbots that have existed for decades now are *not real* AI. But you could train those to respond in certain ways. It is quite possible this comic makes fun of this. It is also perceivable that an actual AI without sufficient higher-order reasoning would also easily be fooled, although by trial-and-error learning (that is characteristic for AI) it ought to *eventually* figure out that you are bullshitting it, and adjust its behaviour accordingly. Todor (talk) 18:00, 20 June 2016 (UTC)
- I thought it might be a reference to teaching the Urban Dictionary to IBM's Watson 108.162.250.158 (talk) (please sign your comments with ~~~~)
- Oh gods — please don't even mention the Urban Dictionary. A classic example of a good idea gone bad because of little or no moderation. RAGBRAIvet (talk) 09:03, 21 June 2016 (UTC)
- Part of the joke is that 6 year olds are actually really smart compared to computers. Perhaps it would be easier to pass the Turing test as an adult-like AI then as a child-like AI.162.158.83.180 09:20, 21 June 2016 (UTC)
Another similar joke was used in the cartoon feature movie "South Park: bigger, longer and uncut", with the word "buttfor". "What's the butt for?" - "For pooping, silly". 108.162.244.73 05:29, 22 June 2016 (UTC)
Is it unimaginable that Randall is hoping for people (who don't know about the joke) to google "What is updog?"? 141.101.104.139 12:11, 22 June 2016 (UTC)
I don’t really like the lambda calculus example. Usually you don’t grant yourself the luxury of natural numbers (you use Church numbers instead), let alone a square operator. Introducing them contradicts the previous point about Lambda calculus’ simplicity. 162.158.83.156 12:58, 12 August 2016 (UTC)
Has anyone tried doing this with ChatGPT? 108.162.242.105 13:56, 29 January 2024 (UTC)
- I think ChatGPT does count as an AI trained to be like an adult AND then you need to prompt-convince ChatGPT to act like a bratty kid.
- oh wait
- there's vedal987's "NeuroSama" 172.70.233.187 (talk) 22:05, 23 October 2024 (please sign your comments with ~~~~)