https://futureofcoding.org/ logo
#of-ai
Title
# of-ai
g

guitarvydas

06/05/2023, 2:05 PM
• humans have cognitive limits, • e.g. the Magical Number Seven, Plus or Minus Two • https://en.wikipedia.org/wiki/The_Magical_Number_Seven,_Plus_or_Minus_Two • does GPT have similar cognitive limits? • is this a Golden Number? a universal truth? a universal constant? • is it necessary to arrange information in a hierarchy? • or, does GPT’s perfect memory change this pattern?
i

Ivan Lugo

06/06/2023, 12:45 AM
I don’t think the idea of token prediction has the same cognitive limit, I think it’s a function of the language itself. I think we’re discovering that our language has become a massive, global, flexing and morphing neural net itself. It encodes context, history, meaning. We’re slowly discovering that over the course of a few thousand years, humans have encoded into their language and symbolic representations the very data itself it’s describing. The fact that science is coming out so easily in these LLMs doesn’t surprise it, but it’s beautiful nonetheless. Our scientific discourse is by itself consistently self referential, and both accepts and rejects itself with additional language. Papers that are cited heavily and appear more often in token inference are a ‘core’ of that token-space. None of this is surprising again. We wrote dictionaries wherein all the words point to each other. An encyclopedia is a dictionary with ordering. An LLM is an encyclopedia with statistic relationships. IMO.
w

wtaysom

06/06/2023, 6:58 AM
Counting, considering alternative mental models are the main place where I encounter limitations in Chat. So any sort of basic SAT problem is going to leave it confused.
Given the nature of some recent updates I've noticed, I wonder if Chat is a little better now. Let's find out...
(Dialog follows.)
Me: I saw three people get on an empty blue bus. Then I saw five children get on the yellow school bus. A mom ran up carrying her boy. She climbed into on the school bus. Then she ran to catch the blue bus before it left. How many people are on the blue bus and ow [sic] many people are on the yellow bus?
Chat: Let's break down the information given: • Three people got on the empty blue bus. • Five children got on the yellow school bus. • A mom ran up carrying her boy and climbed onto the school bus. • Then, she ran to catch the blue bus before it left. Based on this information, we can determine the following: • Initially, there were 3 people on the blue bus. • The yellow school bus had 5 children and the mom with her boy. However, it is unclear whether the mom and her boy stayed on the school bus or transferred to the blue bus. The given information suggests that the mom ran to catch the blue bus, implying she might have left the school bus. So, we can conclude that there are: • 3 people on the blue bus. • 5 children on the yellow school bus. As for the total number of people involved, we have accounted for 3 people on the blue bus and 5 children on the yellow school bus. However, we don't have enough information to determine the number of people on the blue bus after the mom ran to catch it.
Me: Just to clarify... The mom dropped off her boy in the yellow bus, and she did make it onto the blue bus, and each of the buses has one driver. So how many people are on each bus in the end?
Chat: Apologies for the confusion. Based on the clarification you provided, here's the revised breakdown: • Initially, there were 3 people on the blue bus (excluding the driver). • The yellow school bus had 5 children (excluding the driver) and the mom with her boy. • The mom dropped off her boy on the yellow school bus and then boarded the blue bus. • Assuming each bus has one driver, we need to include them as well. So, after the mom boarded the blue bus: • Blue Bus: 3 people (initially) + 1 mom + 1 driver = 5 people. • Yellow School Bus: 5 children + 1 mom (who dropped off her boy) + 1 driver = 7 people. Therefore, there are 5 people on the blue bus and 7 people on the yellow school bus in the end.
So it definitely looks like Chat has been prompted to take apart these questions step by step. You'll also notice it is more cautious about making assumptions.
d

Denny Vrandečić

06/08/2023, 12:03 AM
LLMs have context windows, and they are a limitation of current systems
m

Matthew Carr

06/09/2023, 9:32 PM
@Denny Vrandečić Not sure how well this performs in its current state but this company just recently came out saying they created an LLM with a context window of 5 million tokens. https://twitter.com/magicailabs/status/1666116935904292869?s=46
d

Denny Vrandečić

06/09/2023, 10:12 PM
Wow!
7 Views