Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Lone Banana Problem in AI

digital-science.com

Post reply on HN
Print view View original

41–50 of 109 posts

Re: The Lone Banana Problem in AI

#41

Post by foobarbecue »

lproven wrote: post #28
rishabhjain1198 wrote: post #9

I wish this article was just 3 paragraphs. The verbose writing style was a little tiring, I found myself scrolling impatiently to find what the actual "Lone Banana Problem" was.

The pictures did not immediately tell you? They told me what the problem was in about 2 seconds, and frankly, the title and the first sentence alone were. All right, I concede, a blind reader would not know, but most people are not blind.

I read almost the whole article thinking there might be something more in there. There was mention of TCAV but that was about it.

Re: The Lone Banana Problem in AI

#43

Post by jonplackett »

anotheryou wrote: post #25

Midjourney solution: "--no bunch, two, multiple": https://i.imgur.com/QnJmLr0.jpeg These models have a tendency to move towards the average, especially if unprompted. As we see here, sometimes even if prompted otherwise. They just have to get better or need a more precise interface like the —no :). We also couldn't have "a man crawling" before and now we can: https://i.imgur.com/ycVpk3i.jpeg

They don't really look like bananas anymore though...

Re: The Lone Banana Problem in AI

#44

Post by lordnacho »

LLM doesn't know stuff, it has few models beyond an extremely deep association between (in this case) descriptions and pictures.

Humans have shitty models for stuff, but we have them. We lack the massively deep numerical associations.

It can't be that long before someone makes an AI that says knows a few common models like "stuff made of steel is rigid, stuff made of cloth is soft, etc". Along with "If someone keeps emphasizing a number, that's the number of things he wants".

Re: The Lone Banana Problem in AI

#45

Post by sarabande »

Here is a summary, part ChatGPT and part me:

The “Lone Banana Problem” describes subtle biases of Large Language Models (LLMs) in AI: LLMs reproduce the statistical average of the inputs that they have consumed in the context of the question they have been asked. It's called that problem because the model used to generate images has never seen an individual banana, so when prompted always generates two bananas.

Re: The Lone Banana Problem in AI

#46

Post by ubercore »

jamesgreenleaf wrote: post #33
rishabhjain1198 wrote: post #9

I wish this article was just 3 paragraphs. The verbose writing style was a little tiring, I found myself scrolling impatiently to find what the actual "Lone Banana Problem" was.

This is the biggest communication problem most people have. Say it plain and simple. No one wants to read your train of thought.

Thought process is fine, but lead with the plain and simple version. Offer context after you've introduced the topic and described what you're talking about.

Re: The Lone Banana Problem in AI

#49

Post by anotheryou »

jonplackett wrote: post #43
anotheryou wrote: post #25

Midjourney solution: "--no bunch, two, multiple": https://i.imgur.com/QnJmLr0.jpeg These models have a tendency to move towards the average, especially if unprompted. As we see here, sometimes even if prompted otherwise. They just have to get better or need a more precise interface like the —no :). We also couldn't have "a man crawling" before and now we can: https://i.imgur.com/ycVpk3i.jpeg

They don't really look like bananas anymore though...

first one is fine. And I could do it more carefully :)

Adjusted and with help of the other commenter°: https://i.imgur.com/e2aE1bQ.jpeg

And you always have some crap generations :)

° https://news.ycombinator.com/item?id=36584831

Re: The Lone Banana Problem in AI

#50

Post by im3w1l »

lordnacho wrote: post #44

LLM doesn't know stuff, it has few models beyond an extremely deep association between (in this case) descriptions and pictures. Humans have shitty models for stuff, but we have them. We lack the massively deep numerical associations. It can't be that long before someone makes an AI that says knows a few common models like "stuff made of steel is rigid, stuff made of cloth is soft, etc". Along with "If someone keeps…

Me: A designer has created two artpieces. The first one "The gattoya" is made of steel bars welded into a spiral shape. The second "Misiblur" is made of cloth and somewhat resembles a butterfly. Which is most comfortable to rest your head on and why?

Chatgpt: Based on the description provided, it is likely that "Misiblur," the art piece made of cloth resembling a butterfly, would be more comfortable to rest your head on compared to "The gattoya," which is made of steel bars welded into a spiral shape.

The reason for this is that cloth generally has a softer and more cushioned surface compared to steel bars. When resting your head on Misiblur, the fabric would provide a certain level of support and padding, which could make it more comfortable for resting your head.

On the other hand, resting your head on steel bars, as in the case of The gattoya, would likely be less comfortable due to the hardness and lack of cushioning provided by the material. Steel bars are not designed to be soft or ergonomic for resting purposes, so it may not provide the desired comfort.

It's important to note that the comfort level can vary depending on individual preferences and the specific design and construction of the art pieces. However, based on the given materials and descriptions, Misiblur made of cloth would generally be the more comfortable option for resting your head.

Post reply on HN