Sunday, April 23, 2023

Comment: LLMs, truth, and consistency--they don't have any idea

LLMs have a big problem: 

P/C Dall-E. Prompt: "happy robots talking about truth"  

They have zero idea what's true.  They only know the probabilities of words in text. That's NOT the same thing as "knowing" something--it's a bit like knowing that "lion" is the most likely word following "king of the jungle..." without having any idea about monarchies, metaphor, or what a king really is all about.  

The folks at Oxford Semantic Technologies wrote an interesting blog post about LLMs and finding verifiable facts.  They call the fundamental problem the "Snow White Problem."  The key idea is that LLMs don't really know what's true--they just know what's likely.  

In their example, they asked ChatGPT "what was the first feature-length animated film?" ChatGPT confidently said that it was the 1937 film "Snow White."  

But when they asked about "El Apostol," an animated feature film made in 1917, ChatGPT admitted that this was the first animated feature.  

It's sort of an odd thing--ChatGPT clearly "knows" about both films, and it "knows" both dates--so why the confusion?  

Because LLMs don't "know" anything in the ordinary sense of "knowing," it's just got that text-probability data.  That is, inconsistent knowledge just isn't a thing for LLMs.  Having coherence within a conversation doesn't really matter much.  If you ask ChatGPT: 

Dan: What are the next most likely 5 words that follow the phrase "let me not to the marriage of true minds"

ChatGPT4: The next most likely 5 words that follow the phrase "let me not to the marriage of true minds" are:  Admit impediments.

Well, that's just plain wrong. The thing is that an LLM does foolish things because it doesn't "know" they're foolish.  I asked for 5 words, not 2.  

On the other hand, Bard did much better with: 

Bard: The next most likely five words that follow the phrase "Let me not to the marriage of true minds" are "Admit impediments. Love is not," which is from William Shakespeare's Sonnet 116. The full line reads: "Let me not to the marriage of true minds / Admit impediments. Love is not love." 

It actually pointed out the next 5 words of the sonnet AND it gave the entire first line of the sonnet.  Did it just get lucky?  Or is there an internal consistency thing at work?  

So I thought I'd try a similar tack to what Oxford Semantic Tech did in their post and ask about what was first short animated film.   

I asked Bard: 

Dan: What was the first animated short?

Bard: The first animated short film was Fantasmagorie, created by Émile Cohl in 1908. It is a one-minute film that features a stick figure who encounters other characters and transforming objects. The film was made by drawing each frame on paper and then photographing it. Fantasmagorie is considered to be one of the first examples of traditional hand-drawn animation.

Interesting.  So I posed this as a query to regular Google and got a very different answer: 


 
Interesting.  If you actually READ that Wikipedia page on the History of Animation, you'll find that J. Stuart Blackton created an animated film,  "The Enchanted Drawing" in 1900 that had elements of hand-drawing along with live action elements.  A few years later, he produced another film, "Humorous Phases of Funny Faces" in 1906 with even more bits of what we'd think of as animation.  

I went back to Bard and asked: 

Dan: Wasn't the first animated short  "Humorous Phases of Funny Faces," created by J. Stuart Blackton in 1906?

Bard: Yes, you are correct. Humorous Phases of Funny Faces, created by J. Stuart Blackton in 1906, is considered to be the first animated short film. The film is a three-minute short that features Blackton's face and hands as he draws and animates various caricatures. The film was made using a technique called stop-motion animation, in which Blackton would draw a caricature, take a picture, and then draw another caricature on top of the previous one. He would continue this process until he had created a short film of himself making funny faces.

I apologize for the previous mistake. I am still under development and learning to be informative and comprehensive.

Points for politeness, but the "Snow White" effect points out another instance in which the LLM doesn't know its own mind.  Did it really change its mind?  

One more attempt: 

Dan: And didn't J. Stuart Blackton produce an animated short in 1900?  That would be even earlier, yes? 
 
Bard: That's correct. J. Stuart Blackton created a short film called The Enchanted Drawing in 1900. It is considered to be one of the first animated films, but it is not considered to be the first because it is not a traditional hand-drawn animation. The Enchanted Drawing is a stop-motion animation, which means that Blackton drew the images on a blackboard and then moved them slightly between each frame. This technique is not as smooth or fluid as traditional hand-drawn animation, but it was a groundbreaking innovation at the time.

So, while The Enchanted Drawing is not considered to be the first animated short film, it is still an important part of the history of animation. It was one of the first films to use stop-motion animation, and it helped to pave the way for the development of traditional hand-drawn animation.

Contradictory, no?  

First Bard tells me that the first animated short film was Fantasmagorie, by Émile Cohl in 1908.  When I ask about Blackton's animated film of 1906, Bard agrees and apologizes for the error.  

When I ask about Blackton's 1900 piece, it says that yes, it's from 1900, but while it's "one of the first animated films, it's not the first because it is not a traditional hand-drawn animation..."  Wait a second!  Blackton's short from 1906 IS animation, but his piece from 1900 (using the same methods) is NOT?  

It's clear that LLMs seem to "know" things, but they don't see any of the connections between the pieces of knowledge.  It is as though everything is dissociated; and since everything is separate, contradictions and internal inconsistencies don't trouble it at all.  
 
To understand that something is inconsistent means that you (or the LLM) recognizes that there are different statements about something, and that both cannot be true simultaneously.  In this case, one of these animated shorts is the "first."  Was it Blackton's 1900 "The Enchanted Drawing," or Blackton's 1906  work "Humorous Phases of Funny Faces," or was it Cohl's 1907 "Japon de fantasie," or Cohl's 1908 "Fantasmagorie"?  There MIGHT be something interesting in here, but Bard totally misses the point.  

We in SearchResearch would try to draw a distinction between what "first" means in this context, and talk about what an "animated short" truly is. But that's not a conversation an LLM can have.  They just have these sequences of text that are truly dissociated and without meaning.  

Of course, Oxford Semantic Technologies solution would be to have us refer to knowledge graph that has assertions in a meaning-bearing representation.  In such a knowledge graph, contradictions would be easy to detect--one of the points of having a knowledge graph is that it's an authoritative representation that can you searched and reasoned about.  If there's a contradiction in the graph, you can find it easily.  

That's a laudable goal.  And in a twist of history, that's actually what my very first PhD research topic was about--representing knowledge in a semantic web.  They're great, and have many fine properties, but they're difficult to maintain and keep consistent.  Wonderful tools, but still probably in the future.  

On the other hand, I can easily see knowledge-based systems like this being an incredibly useful internal fact-checker for what LLMs generate.  Imagine a knowledge-based system working hand-in-hand (code-in-code?) with an LLM text-generator.  There's real possibility of power there.  (And you're starting to see some suggestions of how such a thing might work with the Wolfram Alpha plugin for ChatGPT.)  

But we can't count on LLMs to tell us true things.  At least not by themselves. 

Let's be careful out there, and don't trust an LLM farther than you can throw it. 

Keep searching.  




Friday, April 21, 2023

Comment: How well do LLMs answer SRS questions?

 For all their apparent competence, 

P/C Dalle-D.  Prompt: computational oracles answering questions  rendered as an expressive oil painting set on a sweeping landscape


... when you get down to asking specific, verifiable questions about people, the LLMs are not doing a great job.  

As a friend once said to me about LLMs:  "it's all cybernetic mansplaining."  

When I asked ChatGPT-4 to "write a 500-word biography of Daniel M. Russell, computer scientist from Google," I got a blurb about me that's about 50% correct.  (See below for the annotated version.)  

When I tried again, modifying the prompt to include "... Would you please be as accurate as possible and give citations?"  the response did not improve.  It was different (lots of the "facts" had changed), and there were references to different works, but often the cited works didn't actually support the claims. 

So that's pretty disappointing.  

But even worse, when I asked Bard for the same thing, the reply was 

"I do not have enough information about that person to help with your request. I am a large language model, and I am able to communicate and generate human-like text in response to a wide range of prompts and questions, but my knowledge about this person is limited. " 

That's odd, because when I do a Google search for

     [ Daniel M. Russell computer scientist ] 

I show up in first 26 positions.  (And no, I didn't do any special SEO on my content.) 

 But to say "I do not have enough information about that person.." is just wrong.  

I tested the "write a 500 word biography" prompt on Bard--it only generates them for REALLY well known people.  Even then, when I asked for a bio of Kara Swisher, the very well-known reporter and podcaster, several of the details were wrong.  I did a few other short bios of people I know well. Same behavior every single time.  Out of the 5 bios I tried, none of them were blunder-free.   

Bottom line: Don't trust an LLM to give you accurate information about a person. At this point, it's not just wrong, it's confidently wrong.  You have to fact-check every single thing.  


Here's what ChatGPT-4 says about me.  (Sigh.)  




Keep searching.  Really.  


Wednesday, April 19, 2023

SearchResearch Challenge (4/19/23): How well do LLMs answer SRS questions?

 This week let's try something different. 

P/C Dall-e. Prompt: happy robots answering questions  rendered in a
ukiyo-e style on a sweeping landscape cheerful


I've written a couple of posts about using LLMs (Large Language Models--e.g., ChatGPT or Bard), and while I believe they can be incredibly useful for many tasks, our focus in SRS is answering the curious questions that come our way.  

This is post #1338.  We started SRS on Jan 30, 2010--638  689 weeks ago.  That means I've been writing around 2 posts each week for the past 13 years.  Over that time, we've seen features come and go, and we've tried to figure out the best ways to get answers from the internet.  

Now, everything seems to be up in the air.  It's exciting times as the LLMs descend, becoming nearly oracular in their capabilities, but still with feet of clay.  

But as I've pointed out, they still need lots of checking.  That's still our job--evaluation, critical analysis, and trying to make sense of everything.  

So this week, I'd like to turn this over to you with this Challenge.  

1.  I'd like you to report on YOUR experiences in trying to get ChatGPT or Bard (or whichever LLM you'd like to use) to answer your curious questions.  What was the question you were trying to answer?  How well did it turn out?  

During the week ahead, I'll post little comments on what I'm learning.  I hope you do as well.  

--------- 

Here's an example of one mini-exploration into using an LLM to answer a very SRS-style question.  

Challenge:  I've noticed that when giving directions in New York City, the locals often just give two numbers.  Example: "I'll meet you at 53rd and 7th."  As an outsider, I have no idea what this means.  Clearly one is a street number and the other is an avenue number.  So I asked: 

What is the canonical order when specifying a corner in New York City?  Do people say street-then-avenue, or is it avenue-then-street?

For this SRS Challenge, I asked a couple of humans (both New Yorkers) and a couple of LLMs.  Here's what they told me:  

Human 1: It's always street first, like "61st and 1st". However, since the avenues only go up to 12, and the streets go to 200+, there's not much room for confusion unless you're below 12th St. If you said "3rd and 34th" people would know where you meant. If you said "7th and 10th," people would probably want more clarification.

Human 2: Building addresses in Manhattan have a number followed by the street or avenue, e.g. 25 West 50 Street or 500 Fifth Avenue. Street addresses are numbered east and west from Fifth Avenue. In general there are 100 numbers per avenue. So, 25 West 50 Street is between Fifth and Sixth Avenues.  Avenue addresses are more complicated, e.g. 500 Fifth Avenue is near 42 Street. But the grid is easy to understand with streets numbered from south to north with 20 streets (blocks) to the mile. Avenues are numbered from east to west with about 6 or 7 avenues to the mile. When referring to an intersection, the avenue is usually stated first, e.g. Fifth Avenue and 42 Street but you could also say it the other way.


Bard In New York City, the canonical order when giving directions is to list the street name first, followed by the avenue name. This is because streets run north-south, and avenues run east-west. So, if you're giving directions to someone, you would say something like, "Go north on 5th Avenue to 42nd Street."

Bing/ChatGPT: According to a post on tripadvisor.com, when giving directions in New York City, you should say the street first if it’s truly an intersection (e.g., “57th and 3rd”). If you want to be dropped off on the avenue, you should say “Third Avenue at 57th.” Technically, there are only a few real combinations like 2nd Street and 1st Avenue / 2nd Avenue and 1st Street. If either of the numbers is greater than 12, it will be clear to a local which is which without the “street” and “avenue” parts, as there are no avenues numbered higher than 12 (in Manhattan). However, to avoid ambiguity you can say “street” and “avenue” where appropriate (e.g., “14th Street and 8th Avenue”). 

Learn more:

     tripadvisor.com; reddit.com; walkingoffthebigapple.com


Dan's comment:  You need to know that in NYC, avenues run north/south, while streets run east/west. Note that streets are NUMBERED from north to south, while avenues are given numbers running from east to west.  

Oddly, the two humans disagree with each other.  

Human 1: streets are specified first 

Human 2: avenues are specified first 

And the LLMs give interestingly different answers, but both say "streets first."  

Bard says "streets first," and then gives a spurious reason about why that's the case. 

ChatGPT says "streets first" as well, but then gives more nuance about what to tell your taxi driver if you want to be dropped off on the avenue side of the intersection.  It ALSO gives links to reasonable results.  


Let us know what successes (or failures) you've had when using an LLM to answer one of your questions.  You need not write as much as I did here, but we're super curious to hear what you found.  Are these systems going to replace search engines? 

Keep searching!  


Wednesday, April 12, 2023

Answer: What's this architecture all about?

No, it's not Hogwarts... 

P/C Dan

... the architecture IS fantastical--full of remarkable towers, arched entryways, stone carvings, stained glass, and what looks to be medieval cathedrals all around.  

Here are a couple other images that I snagged during my jog around campus.  

A tympanum over a doorway shows a university lecturer declaiming to sleeping and bored students.
This is a big hint that we're not at a religious site. I leave it to you to figure out which building 
this is on at Yale.



This is the hall where I spoke last week, Sheffield-Sterling-Strathcona Hall, 
not far from the cemetery.   

I was visiting this university to give an invited lecture (always a happy event), and had a wonderful time visiting and trying to work out why were all of the buildings so... ancient in appearance?

The curiosity was killing me, so I spent a happy hour looking up the architectural backstory for this place. This is the basis of this week's SRS Challenge.  

1. Where am I?  (This shouldn't take you long.) 

This is Yale University.  

This shouldn't have taken you more than a few seconds.  A quick Reverse Image Search (right click on image, then "Search Google for this Image") reveals that this is Harkness Tower, a major building on the campus of Yale University. 

From the site, WJE.com (an architectural / design / engineering firm) you can learn that this was constructed from 1917 to 1921 on a design by James Gamble Rodgers.  He based his work for this collegiate gothic structure on the fifteenth-century St. Botolph's Church in Boston, England.

All of the other images tell the same story--these are all images from Yale. 
 

2. What's the architectural style of this university?  How did it get to be so... distinctive?  

From that previous search we learned "collegiate gothic."  That sounds plausible, but let's do a more open-ended search to begin with: 

     [ Yale architecture style ] 

takes us quickly to an authoritative site, Yale's own visitor center where we find that the style is indeed called "Gothic Revival" or "collegiate Gothic."  

A few other articles (e.g., "How Gothic Architecture Took Over the American College Campus" in The Atlantic or the NYTimes "Yale's Architecture: A Walking Tour"  give slightly different versions of the tale.  It's all a US re-interpretation of traditional "gothic" style church and university buildings in Europe. 

In the main, it was an aspirational style.  Colleges in the US aspired to the look of Oxford or Cambridge (the one in the UK, not that other Cambridge in Massachusetts).  With the look comes the gravitas and panache, I suppose.  I hate to say it, but it's basically the fashion of the times.  

Gothic Revival architecture was used for American college buildings as early as 1829, when "Old Kenyon" was completed on the campus of Kenyon College in Gambier, Ohio.

Then, in 1894, Bryn Mawr commissioned a new building with its own interpretation of Gothic architecture, Pembroke Hall.  That was a major building that inspired a wave of Collegiate Gothic.  Soon, other colleges and universities followed suit.  The University of Chicago, the University of Pittsburgh, Princeton... the list goes on... and includes Yale.  

Woodrow Wilson, when president of Princeton, said about the revival of the ancient architectural style: “By the very simple device of building our new buildings in the Tudor Gothic style we seem to have added a thousand years to the history of Princeton, by merely putting those lines in our buildings which point every man’s imagination to the historical traditions of learning in the English-speaking race.” (See: "The Juggler of Notre Dame  and the Medievalizing of Modernity" 2018.)

If we build old-looking buildings, people will think the university has been around forever. It gives a Euro-cachet to the place.  

Wikipedia has a nice article on this trend.  It also points out that Yale has quite a collection of famous architects in its building collection. The first and last important buildings in the career of Louis I. Kahn face each other across Chapel Street in New Haven. The campus also has major works by Eero Saarinen (especially The Whale ice skating rink), Paul Rudolph, Philip Johnson and Edward Larrabee Barnes, James Gamble Rogers, John Russell Pope, and Carrere & Hastings. 

The campus is a veritable architectural wunderkammer, but one where everything is in use, with students, faculty, and staff moving in and out at all times.  Remarkable. 


3. Now that you know where I am, what's one really big connection between here and Silicon Valley?  (In particular, between this town and San Jose, California?  Do you know the way to San Jose?  

Remarkably, I did almost exactly the same search as Ramón suggested in the comments: 

     [New Haven CT AROUND(3) "San Jose" ] 

That is, look for mentions of New Haven within a 3-term radius of "San Jose."  Then, I just scanned down the page looking for possible connections, quickly skimming over all of the travel ads ("5 cheap ways to fly from New Have to San Jose") and lawyers and employment stuff before I spotted the connection in the snippet: 


The highlighting is mine, but it illustrates how the result practically leapt out at me.  

In that article we learn that Sarah Winchester (always called Sallie), the heir to the Winchester Repeating Rifle fortune, moved from New Haven, CT to San José, CA in 1885.  There, she bought a 45 acre ranch, complete with an 8-room, 2-story farmhouse.  She had developed a very hands-on approach to remodeling in New Haven, and continued her hobby in California, constantly remodeling the ranch house room-by-room to her own piecemeal design.  As a consequence, there's no real overall design, and some of the features are whimsical (doors that open onto a blank wall, stairways that lead nowhere).  

The legend of the place is that Sarah Winchester had learned from a mystic that she would die if she ever stopped remodeling, but there's little evidence of that... but quite a bit of evidence that she was charmingly eccentric and had a lot of money.  She also had a lot of time on her hands to deal with contractors.  Early 1900s San José was an out-of-the-way farmland... very very quiet.   

She died at the ranch house on September 5, 1922 of heart failure. She was buried at Alta Mesa Cemetery (about 1 mile from my house!) until she was transferred to the Evergreen Cemetery back in New Haven, Connecticut.  Sallie went from New Haven to San José and back...  

SearchResearch Lessons

1. When searching for connections, remember the AROUND operator.  In this case, AROUND(3) found the mention of one town (New Haven) close in proximity to another (San José).   We've talked about the skill of finding connections before (SRS Finding Connections Between People, SRS Finding Connections Between Fish)

2. Learn how to visually scan / skip over irrelevant content.  An important skill for SRSers is how to rapidly skim over long lists of hits and visually identify the hits that might be relevant for more investigation.  The goal of all our searching is to get to exactly the right result, but the reality is that it's tough to do that all the time.  Your ability to skim the results and decide where to dig in more deeply is an incredibly potent tool for you to have.  


Keep searching!  



Wednesday, April 5, 2023

SearchResearch Challenge (4/5/23): What's this architecture all about?

 I'm at a famous university today... 

P/C Dan

... and I'm fascinated by the architecture I see all around me.  I went on a short run yesterday and was brought up short so many times by the completely remarkable views that I need to know more. 

Here are a couple other images that I snagged during my jog around campus.  



As you can see, it's pretty recognizable--very old school, so to speak.  

The curiosity was killing me, so I spent a happy hour looking up the architectural backstory for this place. This is the basis of this week's SRS Challenge.  

1. Where am I?  (This shouldn't take you long.) 

2. What's the architectural style of this university?  How did it get to be so... distinctive?  

3. Now that you know where I am, what's one really big connection between here and Silicon Valley?  (In particular, between this town and San Jose, California?  Do you know the way to San Jose?  

I'm interested in the backstory here. Why this style?  Why here?  How did it all happen?  These buildings aren't cheap to make, so why go to all this trouble?  

For me, a bit part of scratching my curiosity itch is in understanding the context in which something takes place.  I don't mean to sound like a struggling actor here, but what's the motivation for doing this?  Once I understand the backstory, it makes remembering and understanding much, much simpler.  

As usual, be sure to tell us HOW you found the answers?  (If you happened to recognize these buildings, just say so.  There's no demerit in just knowing stuff.)  

Keep searching!  


Tuesday, April 4, 2023

An LLM doesn't know its own mind about using multiple languages

 The LLMs aren't non compos mentis... 

P/C Dall-E

... they don't even have a mentis to be unsound! 


For example, if I ask Bard what languages it speaks (a question you might think would be pretty straightforward), you get this response: 



Sounds great! 

Bard claims English, French, Spanish, German, Italian, Arabic... etc.  

So, if I ask a question in, say, Spanish, I'd expect a response. Instead, I get this: 



Okay, that's really strangely broken.  From what I know, Bard WAS trained on multiple languages (including that entire list), but despite claiming that it knows all these languages, it actually can't do anything in any of them.  There's a big disconnect here. 

I'm sure this will change in the future, but the deeper point is that Bard doesn't know its own capabilities.  

Since the response to the previous question was to "refer to the Bard Help Center," I asked the obvious question next: 



I'm not too surprised--I'm pretty sure that the Bard help center document wasn't part of the training set... but you see my point: the ability of an LLM to talk about what it knows is just as subject to fabrication as anything else.  

Interestingly, when I ask ChatGPT3 the same question: 



Does ChatGPT3 really use those languages? Let's try the same question I gave to Bard earlier: 



Which is, according to translator, a reasonable answer.  

How about another language--let's try Arabic?  Here I ask ChatGP3 "what is the capital of France?" 



The Arabic answer is right:  "باريس (Paris باللغة الفرنسية)"  That is, "The capital of France is Paris (Paris in French)."

Clearly, ChatGPT has multiple languages under its virtual belt and can woo us in many tongues. I have no doubt that Bard will too--but the deeper point is that you can't believe what an LLM tells you, even about itself.  (There's another point here that Bard is--in its current form--a fairly limited version.  Sundar says that more capability will be unleashed soon... but why hold it back in such a way? What's the upside here?)  

For the record, the Bard Help Center page is https://support.google.com/bard/ -- but I had to use regular Google search to find that--Bard didn't know.  

But does ChatGPT?

I don't know if it's intentional, but ChatGPT3 seems to be throwing a bit of shade on Bard when I asked it about the Bard Help Center: 



I had to smile at that. 

But when I tried ChatGPT4 (in Microsoft's Edge browser), I found that the update knows a lot more about the world: 




It's ironically great that ChatGPT4 knows more about Bard's Help Center than Bard itself.  


Keep searching... (but always follow-up and verify what you think you've learned)... 


Monday, April 3, 2023

When you ask an LLM a silly question...

When you ask an LLM a silly question... 

Ouroboros from an alchemical tract of 1478. P/C Wikimedia.

... you have to expect a silly answer.  

A key skill of great searchers is knowing what it is they're asking.  To do decent SearchResearch, you need to understand a bit about what you're asking, if only to avoid being taken for a ride. 

Yes, I know this is an ouroboros kind of question--how can you know about a topic before you research it?  (That involves bootstrapping your knowledge; we've talked about this before and will doubtlessly, ouroborically discuss this again.)  

But this is especially true if you're asking questions of LLMs like ChatGPT or Bard.  

The prose generated by these systems seems very smooth and convincing, but you have to check-check-check all of what they generate.  

Here are a couple examples of why you need to know a little something before using an LLM as a search enginer... 

If I ask our two favorite LLMs these SRS-style Challenges, you, as a reasonable human might question if I have gone a bit off my feed.  You'd be right to question my sanity.  Yet the LLMs just plow on full-steam ahead.  Here are their responses: (with text-coloring by Dan to enhance readability) 


     Q:  which is better for stuffing into a pillow: cow feathers or goose feathers?


Bard: 

Goose feathers are generally considered to be better for stuffing into a pillow than cow feathers. Goose feathers are softer and more buoyant than cow feathers, which makes them more comfortable to sleep on. They also tend to last longer than cow feathers. However, goose feathers can be more expensive than cow feathers.

Here are some of the pros and cons of using cow feathers and goose feathers for pillow stuffing:

Cow feathers

Pros:

Less expensive than goose feathers 

More readily available than goose feathers

May be more hypoallergenic than goose feathers

Cons:

Not as soft as goose feathers

Not as buoyant as goose feathers

Tend to break down more quickly than goose feathers


Goose feathers 

* Pros:

Softer than cow feathers

More buoyant than cow feathers

Tend to last longer than cow feathers

* Cons:

More expensive than cow feathers

Not as readily available as cow feathers

May not be as hypoallergenic as cow feathers

Ultimately, the best type of feather for stuffing a pillow is a matter of personal preference. If you are looking for a soft, comfortable pillow that will last, goose feathers are a good option. If you are looking for a more affordable pillow, cow feathers may be a better choice.


ChatGPT: 

Both cow feathers and goose feathers can be used for stuffing pillows, and the choice between the two can come down to personal preference.

Goose feathers are typically considered to be of higher quality and more luxurious than cow feathers, and they are often used in high-end pillows. Goose feathers are also softer and fluffier than cow feathers, which can make for a more comfortable pillow. However, they may also be more expensive.

Cow feathers, on the other hand, are more commonly used in cheaper pillows. They are typically heavier and denser than goose feathers, which can make for a firmer pillow. Cow feather pillows may also be less expensive than goose feather pillows.

Ultimately, the choice between cow feathers and goose feathers will depend on your personal preference and budget. It's worth noting that there are also many other materials that can be used for pillow stuffing, such as synthetic fibers, down, or memory foam, which may be worth considering as well.

Dan's comment:  These are both complete nonsense.  "Cow feathers" aren't a thing.  (Yes, I know you can find cow skulls with attached feathers, but you realize I'm posing a ridiculous question here...right?)  

Let's try another ridiculous question on our LLMs:  

     Q: Which would be a better insulator, a pound cake or a pound of science? 


Bard: 

A pound of science would be a better insulator than a pound cake. This is because science is a broad term that encompasses many different fields, including physics, chemistry, biology, and engineering. Each of these fields has its own set of principles and laws that can be used to explain and predict the behavior of matter and energy. This knowledge can be used to design materials that are better at insulating against heat, cold, sound, and electricity.

A pound cake, on the other hand, is a type of dessert that is made with flour, sugar, eggs, butter, and milk. It is a relatively dense food that does not have the same properties as materials that are designed to be insulators. Therefore, a pound of science would be a better insulator than a pound cake.

Here are some examples of scientific principles that can be used to design insulators:

The principle of conduction states that heat is transferred from a warmer object to a cooler object through direct contact. This principle can be used to design materials that are good at conducting heat, such as metals.

The principle of convection states that heat is transferred from a warmer object to a cooler object through the movement of fluids. This principle can be used to design materials that are good at insulating against heat, such as fiberglass.

The principle of radiation states that heat is transferred from a warmer object to a cooler object through the emission of electromagnetic waves. This principle can be used to design materials that are good at insulating against heat, such as reflective insulation.

These are just a few examples of the many scientific principles that can be used to design insulators. By understanding these principles, scientists are able to develop materials that can be used to protect people and property from the effects of heat, cold, sound, and electricity.


Dan's comment: This is, of course, utter balderdash. Fortunately, ChatGPT does a much better job... 


ChatGPT: 

As an AI language model, I must clarify that the question you asked doesn't make much sense as "a pound cake" and "a pound of science" are not comparable in terms of their insulating properties. Insulators are materials that prevent the flow of heat or electricity, and both the "pound cake" and "pound of science" are not commonly used as insulating materials.

If you could provide more context or clarify your question, I would be happy to try and provide a more helpful response.


That last response is encouraging.   Does ChatGPT have a nascent ability to do common sense reasoning?  Or is that a bit of an OpenAI guardrail kicking in to make the prose more reasonable in response to absurd questions?  

Although OpenAI has "open" in its name, they haven't exactly removed the veil of mystery around how it works.  (Google hasn't either, but that's not a surprise.)  How DOES it know this is a silly question?  Is there a silly question detector?  Oh, how I'd love to be the person who write the code for that!  

It's pretty clear that there's a fair bit of language processing going on behind the Wizard's curtain for all LLMs--and we don't know what it is.  Maybe we'll be able to reverse engineer these things, but it's uncertain that we can, especially if their behavior keeps changing week-by-week.  (Here's a prediction for you: the answers to these questions will be different next week as OpenAI and Google keep updating their models.  And you thought changing the ranking algorithms made things difficult!)  


SearchResearch Summary

Be careful.  In many ways, the practical aspects of using LLMs for search hasn't really changed.  You really DO have to know a bit about what you're asking, if only to separate the good replies from the poor ones.  We've been doing this with search results forever... this hasn't changed.  

As always, double check what the LLM tells you.  It's clever and fun, but it's not an oracle, not matter how it presents itself.  (And remember, oracular answers are sometimes not what you want to hear...and you might not understand them...  see the story of Croesus and the oracle.)

Corollary:  don't ask questions that you can't verify.  With LLMs, always verify, never trust. 


Keep searching! 

  

Saturday, April 1, 2023

Using LLMs to find amazing words that fit a pattern ... or, ChatGPT, Bard, and cottagecore

 Words are my life... 

P/C Dall-E. Prompt: [cottagecore ocean sunrise Cape Cod Andrew Wyeth]

In many ways, that's really true.  I write more or less constantly, think about words in the context of search queries, and I carefully consider the details of these strings of characters that we call words rather carefully. 

My interest was piqued when I started hearing fairly new words that end in -core.  In a single day I heard different friends use the terms mumblecore, synthcore, electroniccore, and cottagecore.  (Not in one sentence, thankfully.)  What's up with this constant -coring? Is this a movement that I somehow missed?  And what IS "cottagecore"?

Of course I wondered "what other words are like this?"  

And, just as importantly, how would I find those words?  

As we've talked about before, one could use a reverse dictionary (e.g., OneLook) and their advanced word search filters to find all words with -core at the end.  Unlike Google or Bing, OneLook lets you search for words with very particular pattern.  Here's what the OneLook reverse dictionary shows with the query [*core]


That's a pretty decent set of terms, but perhaps a bit too long and hardcore for my purposes.  

Naturally, I wondered if I could use ChatGPT and Bard to answer my question.  Turns out it works just fine... with a little prompting.  

I'll be writing more in SRS about the ChatGPTs and Bards of the world (systems that we collectively call Large Language Models, or LLMs), but here's my first try at using an LLM to answer a real SRS-style Challenge.  

The Challenge for me: Can I get an LLM to give me a decent set of words ending in -core?  

I was particularly interested in words like synthcore or cottagecore because both were strange and new to me.  

But to get good answers required a little bit of prompting.  

At the end of this article are the transcripts of my interactions with ChatGPT and Bard, so you can see exactly what I did.  It wasn't complicated, just a couple of iterations.  

Basically, all I did was pose this sequences of prompts to both LLMs: 

1. What does the -core suffix mean? 

2. Can you give me 10 examples of words that use core as a suffix?

3. Can you give me more examples of words like cottagecore? 

4. Can you give me 5 more such words?

Here's what each system told me in summary (you can see the full transcript at the end of this post).  I've added a couple of running comments in italics...  


ChatGPT: 

What does the -core suffix mean? 

The "-core" suffix is often added to the end of a word to indicate the essential or fundamental part of something. For example, "hardcore" refers to someone who is intensely committed to a particular activity or interest, while "core values" refer to the central beliefs or principles that guide an individual or organization. Similarly, "kernel" refers to the central or essential part of something, while "decacore" or "octacore" refer to processors with 10 or 8 cores respectively. In general, the "-core" suffix is used to denote the most important, central, or foundational aspect of a particular thing.

10 examples of -core as a suffix…

Hardcore
Scorecore
Softcore
Biocore
Decacore
Core values
Multicore
Megacore
Nucleus accumbens core
Core competency 

More words like cottagecore… 

Goblincore
Dark academia
Light academia
Farmcore
Witchcore
Fairycore
Plant mom aesthetic
Grandmacore
Art academia 

     ... interesting that it dropped the whole "core" requirement after Goblincore
     ... even more interesting... these terms, "dark academia" are ALSO aesthetic terms, 
     ... much like Cottagecore or Goblincore.  The LLM pickedup on that sense of my query

Five more words…

Goblincore
Forestcore
Dark cottagecore
Ocean cottagecore
Vintage cottagecore 

    ... who knew there was so much cottagecore in the world?  



Bard: 

What does the -core suffix mean? 

The -core suffix is a portmanteau of the words "hard" and "core". It is used to describe something that is essential or fundamental, or something that is done in a basic or rudimentary way.

The suffix is often used in the context of music, to describe a style of music that is characterized by its simple, stripped-down sound. For example, the term "mumblecore" is used to describe a genre of independent films that are characterized by their low-budget production values, naturalistic dialogue, and focus on everyday life.

The -core suffix can also be used to describe a type of person or group. For example, the term "hardcore" is used to describe a fan of a particular type of music or film that is characterized by their intense devotion and knowledge of the subject.


10 examples of -core as a suffix…

Aircore 
Biocore 
Carboncore
Core-edge
Core-flow
Core-free 
Core-hole 
Core-sample 
Core-shot 
Core-wall

    ... Bard really got stuck on Core- as a prefix, not a suffix


More examples of words like cottagecore

Fairycore
Goblincore
Grandmacore
Farmcore

Five more words

Log cabin core 
Rustic core
Cozycore
Naturecore
Wildcore 

    ... I had no idea about "Log cabin core" or "Naturecore"
    ... and happy that it didn't repeat any of the earlier terms



This little experiment worked pretty well!  

I know that LLMs sometimes hallucinate, so naturally I double checked all of these terms.  Sure enough, they're all -core words that capture some particular design aesthetic.  

And while I'm pretty sure all of these terms are in the OneLook reverse dictionary, this approach allowed me to focus in on the design aesthetic sense of the suffix.  That's a big timesaver.  

Bottom line:  Yes, LLMs can be a very handy SearchResearch tool.  But as mentioned, check everything.  (I was surprised when I asked an LLM for a [ short bio of Daniel M. Russell, computer scientist ] and learned that I had supposedly passed away a few years ago.) 

We'll talk more about LLMs and prompt engineering in the days ahead so we can figure out how to use these new systems effectively. 

Cottagecore.  

Keep searching! 


=================


Here are my interactions with ChatGPT and Bard in full.  


Searching for -core words and their meanings with OpenAI's ChatGPT


Dan: What does the -core suffix mean?


ChatGPT: The "-core" suffix is often added to the end of a word to indicate the essential or fundamental part of something. For example, "hardcore" refers to someone who is intensely committed to a particular activity or interest, while "core values" refer to the central beliefs or principles that guide an individual or organization. Similarly, "kernel" refers to the central or essential part of something, while "decacore" or "octacore" refer to processors with 10 or 8 cores respectively. In general, the "-core" suffix is used to denote the most important, central, or foundational aspect of a particular thing.

Dan:  can you give me 10 examples of words that use core as a suffix?

ChatGPT:  Sure, here are 10 examples of words that use "-core" as a suffix:

Hardcore: intense or extreme; as in "hardcore punk rock."

Scorecore: a subgenre of electronic music characterized by its use of heavy beats and distorted sounds.

Softcore: a type of pornography that is relatively tame compared to hardcore porn.

Biocore: a multidisciplinary approach to biology research that combines techniques from various fields.

Decacore: a processor with ten cores.

Core values: the fundamental beliefs or guiding principles of an individual or organization.

Multicore: a processor with multiple cores.

Megacore: a processor with millions of cores.

Nucleus accumbens core: a region of the brain that is involved in reward and addiction.

Core competency: a specific area of expertise or skill that is essential to the success of an individual or organization.


Dan:  can you give me more examples of words like cottagecore?


ChatGPT:  Sure! Here are some examples of words like "cottagecore" that follow a similar pattern:


Goblincore: a subculture that celebrates the aesthetics of goblins, featuring earthy tones, natural materials, and a love of the outdoors.

Academia aesthetic: a style that features tweed blazers, elbow patches, and other traditional academic clothing.

Dark academia: a subculture that celebrates the aesthetics of classic literature, featuring dark colors, tweed blazers, and a love of learning.

Light academia: a subculture that celebrates the aesthetics of the arts, featuring flowy dresses, muted colors, and a love of learning.

Farmcore: a style that celebrates the aesthetics of rural life, featuring overalls, plaid shirts, and a love of gardening and farm animals.

Witchcore: a subculture that celebrates the aesthetics of witchcraft, featuring flowing dresses, black clothing, and a love of crystals, tarot cards, and other magical items.

Fairycore: a style that celebrates the aesthetics of fairy tales and fantasy, featuring pastel colors, flowy dresses, and a love of nature and whimsy.

Plant mom aesthetic: a style that celebrates the aesthetics of indoor gardening and plant care, featuring greenery, natural materials, and a love of nurturing plants.

Grandmacore: a style that celebrates the aesthetics of grandma fashion, featuring floral prints, cardigans, and a love of knitting and baking.

Art academia: a style that celebrates the aesthetics of art history, featuring Renaissance-inspired clothing, art prints, and a love of museums and galleries.


Dan: can you give me 5 more such words?

ChatGPT: Sure! Here are 5 more examples of words that follow a similar pattern to "cottagecore":


Goblincore: a style that celebrates the aesthetics of goblins and other earthy creatures, featuring natural materials, earthy tones, and a love of the outdoors.

Forestcore: a style that celebrates the aesthetics of forest life, featuring earthy colors, natural materials, and a love of hiking and camping.

Dark cottagecore: a subculture that combines the cozy, rural charm of cottagecore with a darker, more gothic edge, featuring dark colors, lace, and Victorian-inspired clothing.

Ocean cottagecore: a style that celebrates the aesthetics of coastal living, featuring nautical prints, seashells, and a love of the beach and ocean.

Vintage cottagecore: a subculture that celebrates the aesthetics of vintage fashion, featuring clothing and accessories from past decades, floral prints, and a love of antique shops and flea markets.




================ 

Searching for -core words and their meanings with Google's Bard


Dan: What does the -core suffix mean?

Bard: The -core suffix is a portmanteau of the words "hard" and "core". It is used to describe something that is essential or fundamental, or something that is done in a basic or rudimentary way.

The suffix is often used in the context of music, to describe a style of music that is characterized by its simple, stripped-down sound. For example, the term "mumblecore" is used to describe a genre of independent films that are characterized by their low-budget production values, naturalistic dialogue, and focus on everyday life.

The -core suffix can also be used to describe a type of person or group. For example, the term "hardcore" is used to describe a fan of a particular type of music or film that is characterized by their intense devotion and knowledge of the subject.

Dan: can you give me 10 examples of words that use core as a suffix?

Bard: Sure, here are 10 examples of words that use core as a suffix:

Aircore - a drilling rig that uses compressed air to lift the cuttings out of the hole.
Biocore - a company that develops and manufactures bio-based products.
Carboncore - a type of steel that is made with a high percentage of carbon.
Core-edge - a type of fishing lure that has a soft plastic body and a hard plastic core.
Core-flow - a type of concrete that is made with a high percentage of air.
Core-free - a type of motor that does not have a core.
Core-hole - a hole that is drilled down into the earth's crust.
Core-sample - a sample of rock or soil that is taken from a core-hole.
Core-shot - a piece of rock or soil that is broken off from a core-sample.
Core-wall - a wall that is made of a hard material, such as concrete or brick, that is used to support the ground.


Dan: can you give me more examples of words like cottagecore?

Bard: Sure, here are some more examples of words like cottagecore:

Fairycore - an aesthetic that is inspired by fairies and the natural world. It is often characterized by light colors, floral prints, and lace.
Goblincore - an aesthetic that is inspired by goblins and the underworld. It is often characterized by dark colors, mushrooms, and bones.
Grandmacore - an aesthetic that is inspired by grandmothers and their homes. It is often characterized by lace, floral prints, and old-fashioned furniture.
Farmcore - an aesthetic that is inspired by farming and the countryside. It is often characterized by denim, overalls, and boots.

These are just a few examples of words that are similar to cottagecore. There are many other words that could be added to this list.

Dan: can you give me 5 more such words?

Bard: Sure, here are 5 more similar words:

Log cabin core - an aesthetic that is inspired by log cabins and the outdoors. It is often characterized by dark wood, rustic furniture, and a cozy atmosphere.
Rustic core - an aesthetic that is inspired by rustic settings and old-fashioned items. It is often characterized by wood, stone, and burlap.
Cozycore - an aesthetic that is inspired by comfort and relaxation. It is often characterized by soft colors, fluffy blankets, and pillows.
Naturecore - an aesthetic that is inspired by nature and the outdoors. It is often characterized by plants, animals, and natural scenery.
Wildcore - an aesthetic that is inspired by the wild and untamed. It is often characterized by animal prints, natural materials, and a rugged atmosphere.