FADE IN
Act 1
INT. BATHROOM - 17:00
Having fully undressed, PETER gets back his smart phone to type on the screen.
PETER (V.O.): Let me ask a language model's analysis about the reliability of a bank in Silicon Valley.
In a second, we see a screen-sized text appearing. Peter reads it.
PETER (V.O.) (Cont'd): A reasonable answer. It's generated via datasets before the year 2022.
Thinking.
PETER (V.O.) (Cont'd): It's only in the past few days that I've heard about the bank. And I've been following the developments timely.
Thinking.
PETER (V.O.) (Cont'd): Otherwise, I'd have no evidence to judge the answer.
Peter begins to take a hot shower.
Act 2
FLASHBACK
INT. STUDY - DAY (DEC 2022)
Peter is checking the Quora app on his phone.
PETER (V.O.): A Quoran asks me about the impact of ChatGPT...What is it?
Peter googles "ChatGPT" and then watches a video clip.
PETER (V.O.) (Cont'd): It's an AI-assisted natural language model. An expert predicts that there'll be more such apps.
Thinking.
PETER (V.O.) (Cont'd): How can AI beat human intelligence?
Peter types his answer on screen.
INT. COUNCIL CHAMBERS, HKUST - DAY (JAN 2023)
After a meeting, PROFESSOR#1 shows to Peter some texts on his phone screen.
PROFESSOR#1: I've got some possible lyrics in both English and Chinese. They're generated by ChatGPT.
Peter reads them.
PETER: Quite good. How did you get them? People living in Hong Kong can't access to the app.
PROFESSOR#1: With a little help from a friend overseas.
INT. RESTAURANT, HKUST (GZ) - DAY (26 FEB 2023)
As they chat over a meal, PROFESSOR#2 shows to Peter some Chinese texts on this phone screen.
PROFESSOR#2: Via VPN, I first asked ChatGPT to generate a simple love poem, then a sophisticated one, and then to review the two poems.
Peter reads the poems and the review.
PETER: Good works plus a fair comment. There's Chinese proficiency too. Can the mainland Chinese ones in the pipeline surpass that?
INT. RESTAURANT, HK JOCKEY CLUB - NIGHT (28 FEB 2023)
Chatting with Peter, a GUEST shows Peter some texts on his phone screen.
GUEST: While betting on horses the other day, I asked Sage to predict the winning horse. It can't do that.
PETER: No sage advice, as its datasets are dated. Incidentally, how can you ask Sage and others?
GUEST: Via poe.com, I can invite you to join the platform.
INT. STUDIO - DAY (DAYS AGO)
Checking the Quora app on his phone screen, Peter is in thoughts.
PETER (V.O.): Since July 2019, I've answered over 11.2K questions but I've never asked anyone any question.
Thinking.
PETER (V.O.) (Cont'd): Challenging incorrect premises and providing solutions, I've to think deep. That's intellectually satisfying.
Pausing.
PETER (V.O.) (Cont'd): With the introduction of language models, I've no inclination to answer questions on Quora anymore.
Pausing.
PETER (V.O.) (Cont'd): Those asking questions should learn how to learn nowadays.
Thinking.
PETER (V.O.) (Cont'd): Is it still essential to go to schools or universities to get the knowledge transfer?
Pausing.
PETER (V.O.) (Cont'd): Can those who make a living by writing be able to write language model algorithms?
RETURN TO PRESENT
Act 3
INT. BATHROOM - CONTINUOUS
PETER (V.O.): The language models provide structured, concise and precise answers, although not updated.
Pausing.
PETER (V.O.) (Cont'd): While they readily admit they don't have personal experience and emotions, they enlighten my blind spots.
Thinking
PETER (V.O.) (Cont'd): As humans, it's now essential that we think deeper than these language models so that we can offer sage advice.
FADE OUT
THE END

Comentarios