Elevator Paul Challenge

The hackathon before the hackathon - not an official challenge!

Our group is working on Challenge G here:

GeoPostal Insight
UPU Innovation Challenge 2024 Share

Challenge G - Post office, postal service and digital service mapping platform

When we first met, at the start of the event, Oleh mentioned that Paul was quite a familiar name. Paul said, now this is interesting: on his way up to the 7th floor of the UPU building, there were 3 other "Pauls" in the lift, in addition to one other person. So picture this:

0 A group of four people in the elevator. They are a esrganv1x2plus.jpg

(Generated with Stable Diffusion XL 1.0)

"What are the chances of that?" said Paul. And the rest of us, too, wanted to know. So we got the data ...

Screenshot from 20240529 201011.png

From vornamen.opendata.ch and the Swiss Federal Office of Statistics ... and asked the AIs!

Screenshot from 20240529 183042.png

Prompt:

What is the likelihood that 3 out of 4 people in an elevator are named "Paul", when they are in a group of 80 participants total in Switzerland? Consider that there were 95 babies named "Paul" (data source: https://vornamen.opendata.ch ) from 82371 babies born here in 2022 (data source: https://www.bfs.admin.ch/bfs/en/home/statistics/population/births-deaths/births.html ).

Each of the LLMs gave an interesting analysis of the question. GPT-4 used Binomial Probability, while the "hypergeometric probability formula" from Nous Hermes seemed rocket science level. We think they're all hallucinating, and the prompt still needs a lot of work. For example, see the OpenAI chat log.

Preliminary results:


LLM What are the chances?
OpenAI GPT-4o-2024-05-13 3.996 in 1 billion
(1 in 250,000,000)
OpenAI GPT-4o 1 in 1,635,055
Nous Hermes 2 1 in 8,600
Google Gemma 1.1 1 in 555
Zephyr Orpo A35b 1 in 247

Is any of this useful .. or is it pure hallucination? We would need to go through the maths line by line to check whether the reasoning makes any sense. It often does not, though recent research suggests we are steadfast on our way to higher mathematical reasoning ability (Ahn et al 2024).

Of course, we should really be refining our question based on how many Paul's are alive today .. in Switzerland .. or even looking more closely at the kind of audience we have at the event today, and include the chance that people from our international group will take the lift rather than walking up 6 flights of stairs, etc.

For the armchair statisticians and Onomasticians among us, we recommend:

We can just hope that all the Pauls meeting in lifts everywhere will be better friends for it 😆

Event finished

this is the single best use of AI in the world! :)

31.05.2024 13:06 ~ paulsobon

Project

Edited (version 5)

29.05.2024 18:21 ~ loleg

Just warming up here. Start your AI-nnovation engines, people, with this improbable prompt!

29.05.2024 18:21 ~ loleg

Edited (version 4)

29.05.2024 18:18 ~ loleg

Edited (version 2)

29.05.2024 17:50 ~ loleg

Project

Joined the team

29.05.2024 16:31 ~ loleg

First post View challenge

29.05.2024 16:28 ~ paulsobon

Challenge

Event started

 
Contributed 2 weeks ago by paulsobon for UPU Innovation Challenge 2024
All attendees, sponsors, partners, volunteers and staff at our hackathon are required to agree with the UN System Code of Conduct. Organisers will enforce this code throughout the event. We expect cooperation from all participants to ensure a safe environment for everybody.

The contents of this website, unless otherwise stated, are licensed under a Creative Commons Attribution 4.0 International License. The application that powers this site is available under the MIT license.