An AI Case Study: Gemini vs Copilot ⨯ sabe vs sepa
About a year ago, I experimented with some of the first public incarnations of customer-facing, search-oriented AI. At the time, I felt like it had a ways to go.
Today I tried out Google’s Gemini and Microsoft’s Copilot, and I am impressed with the results that I was given. I also felt that Gemini outperformed Copilot when I gave it the language-translation prompt I was interested in:
When the English phrase “I think you know it is” is translated into Spanish as “Creo que sabe que si”, why is the subjective conjugation for saber, sepa, not used?
When pasting the question/prompt into this blog post, I realized that I mistyped ‘subjunctive’ as ‘subjective’, and given the context, the AI could reasonably get confused, but neither of the AI products did. They both seemed to realize I meant ‘subjunctive’.
I think Microsoft’s Copilot provided a correct answer, but it wasn’t especially clear:
Google’s Gemini gave a thorough answer which definitively answered the question for me. I feel like its answer knocked it out of the park in terms of removing confusion:
You’ll notice that Gemini used the word ‘sabe’ in a few places where it should have used ‘saber’. However, that error seemed to essentially be a typo, and given my level of knowledge and ability to spot it, the error didn’t detract from the answer.
(Upon closer examination, there is another typo where ‘present tense’ should be ‘indicative’, but that didn’t affect my comprehension either. Additionally, the #3 reason appears to be an AI “hallucination”. I wasn’t reading too deeply after I ascertained the answer, and in hindsight maybe Gemini’s response deserves additional demerits. However, if the #3 reason is tossed out, which I would be able to do given my level of knowledge, my overall assessment would remain the same. I suppose its clear that this information medium is imperfect, and we are all going to have to come to terms with how we think about these AI errors.)
After doing this comparison, I realized I overlooked a use case for search-oriented AI. In addition to sheer convenience and/or necessity on technology platforms that emphasize voice input and voice output, these AI products sort of act as a second-order search. They perform the search on myriad data sources and synthesize the results.
A year ago, I guess I understood that that was exactly what it did, but I didn’t realize there might be cases where I would rather have a machine do those searches.
In this case, I am not an expert in Spanish, and a good answer to that question may need to rely on manydisparate data sources, and so it kind of made sense for a machine to do the searching for me.
So perhaps that is one use case I was overlooking: Questions pertaining to subjects in which I have limited knowledge, that may require searching through many disparate data sources.
Day in and day out, I still use regular Google Search.
I previously shared my thoughts on the new, experimental AI-enhanced search engines.
Here I’ll share some chat excerpts that showcase some colorful and interesting discoveries. At this early stage, you really can’t read too much into them. However, one has to suspect the following: The heart of these issues will be difficult to tamp down.
Searcher Beware: ChatGTP, Microsoft, Google, and Bard
There was recently a lot of buzz surrounding Microsoft’s heralding of ChatGPT, which is an AI technology that Microsoft is incorporating into their search engine Bing. Google responded by announcing the release of Bard, which will add analogous AI enhancements to their own search engine.
I dabbled with some of the new technology, and my top 3 takeaways were this:
1.) If this had all occurred circa 1998, the value would be much higher. But after 20+ years of people using search engines, we have already built up a native understanding of how search engines work. We have a native understanding of how to parse through search results, get what we want, follow-up on what we want, and so on. At this point, it’s not really a burden to us, and it tends to facilitate us getting what we want, as quickly as possible.
If someone wants to try to synthesize those search results for me in a natural-sounding language, that’s great, but I imagine I’ll often want to dissect it back into its original pieces (i.e. the search results we’re all familiar with) and just process this at a lower level (i.e. the search results we’re all familiar with), to maintain a higher degree of search fidelity.
As an analogy, over the course of many years a software engineer will eventually build up a native understanding of a particular programming language. Although there are certain scenarios where it would be useful to have a technology where a programmer can speak in a natural-sounding language to a computer, and have it spit out code, or conversely, have someone explain existing code in a natural-sounding language, often times that’s just not optimal and in fact it’s painful.
The best use cases for this type of technology were, to the best of my knowledge, actually being played out in cars and even homes years ago. When you’re in a car and need a hands-free search, this is very useful. When someone is not tech-savvy, this can be very useful.
You just have to take those synthesized search results with an extra grain of salt, and you accept that potential loss of precision and accuracy, because you’re driving and at that moment that’s the best you’ll be able to do.
The bottom line is classic search results, and the need for them, will never go away. Many times we will prefer them. In a lot of cases, we need them. To articulate what should be a simple, canonical example of this truth, if one wanted a 2nd or 3rd medical opinion, they would not want a clever intermediary to synthesize the net cacophony of all those doctors’ voices. To the contrary, they would want to hear the entirety of each opinion, and they would want each opinion siloed.
2.) Beware of insidious errors!
For reasons unknown to me, one of these new, state-of-the-art technologies refused to “talk like Sub-Zero ”.
OK, so then I asked a more modest follow-up question, at which point it rattled off some synthesized information, and buried in the middle of that synthesized information was the worst type of error possible.
Relative to its context, it was a critical error, but it was not an obvious error.
Nothing drew any attention to this error. It was very inconspicuous.
It would not be immediately obvious to anyone else what your misconception is. Instead, they would assume your sense of humor was just kind of off, you were acting kind of strange, etc. They would not feel compelled to stop and correct you.
There is very little in the ambient environment that would cue you to the fact that you are mistaken!
Indeed, it is the type of error that you can have in your head for decades until you finally realize that you have it wrong!
It had claimed that Sub-Zero was the “main character” in the Mortal Kombat franchise.
And, if someone were using a search engine as they normally would, I don’t believe they would ever come to that conclusion.
3.)The top search companies will compete as they normally do. Anything is possible, but there is little reason to expect this will fundamentally change the search engine landscape.
People noted that Bard got a question wrong. For reference, when I used ChatGPT, most of the answers were wrong, or had one or more significant inaccuracies within the answer. Not surprisingly, this all reflects what we get in our normal search results, and have a native understanding of how to deal with. So in closing, I would also add that I don’t expect these AI enhancements to fundamentally change the accuracy of our searches.
Before I designed this system, I had spotty recall on maybe just a few tenses. I’m now able to write out the entire 18-tense conjugation chart purely from memory, and I also have a better understanding of what the tenses actually are. (If you’re counting, that chart has upwards of 200 or more facts, and those facts are not conveniently packaged.)
Flawless Recall: Universal Memorization Method For Conjugating Regular Spanish Verbs is better than ever in its 8th edition. I highly recommend it – there’s never been a better book for reliably learning this aspect of the Spanish language!
There is not a shot out of place, not a word uttered or note of music that is not just right in this film. I admit I had to watch it twice to get all the nuances. You must listen for when they are speaking Shanghainese or Japanese and you must understand a bit about what was happening in china and shanghai in 1937. There are family ties, triad ties, Japanese secret society ties and love ties. There is betrayal of all these ties. Jumping back and forth in time makes the plot unravel slowly and like a game of mah-jong the final betrayal is revealed when all the tiles are exposed at the end. This is a great movie. It does not have any kung-fu or wire-work or epic slow motion gunfights and white doves – but it is a truly original masterpiece. I recommend it to anyone who wants to see a thoughtful movie and be challenged by what a great piece of cinema is.
So what is the relationship between Shanghai Triad and The Wasted Times? Well, Hollywood has started to develop a vocabulary for this, and it includes terms like reboot, remake, and re-imagining.
But The Wasted Times doesn’t seem to really match any of those terms. I’d contend The Wasted Times is a spiritual successor to Shanghai Triad, where a spiritual successor would be one or two levels beyond a re-imagining. It’s adopting the environment, the general context, and the era, but not any of the specific characters, relationships, or plot lines.
If you enjoyed Shanghai Triad, do yourself a favor and see The Wasted Times. I have to admit, I was touched by the final scenes – complete with an alien form of chivalry. It was extremely well done. In fact, everything was very well done, and it tends to suggest the setting in these two films (1930s Shanghai) really lends itself to a cinematic rendering. Something about 1930s Shanghai tends to be very crisp, and very compelling, with plenty of surface area for an artist to work with.
Although I probably wouldn’t agree with all of the director’s viewpoints, it’s very hard to argue with the quality of the film. The Wasted Times is a very worthy and a very formidable spiritual successor to Shanghai Triad.
And now for the dubious levels of classification.
This film incorporates a creative device that I have never seen any vocabulary for: At the end of the film, the main characters briefly meet in a completely different setting for the final resolution.
I think my mental shorthand for this is a Point-Break-style ending, although there are obviously other films that have the same device.
In The Wasted Times, as in the original Point Break, this style of ending works because it further pronounces and embellishes key traits for some of the characters.
One could characterize it as the “language of resistance” inside China, and abroad by older generations of the Chinese diaspora.
If the glass is half empty, it makes me wonder if Cantonese might soon be one of those “extremely valuable” languages, in terms of government recruitment efforts.
Weighing in its favor, it’s more of a greenfield language right now. If you’re applying for a relevant government job, you’re less likely to be competing against an entire room full of Cantonese experts for that job (as compared to Arabic).
Weighing against it is the fact that many Cantonese speakers tend to also know English (technically, the same can be said of Mandarin). Thus, the necessity isn’t quite as high.
But also weighing in its favor is the fact that the severity of the potential conflicts between the USA and China is quite high – probably a lot higher than conflicts where Arabic was relevant. This fact alone sort of promotes the value of the language up several notches.
I know next to nothing about Cantonese, but several things in that article fascinated me. One of them is that Cantonese can be expressed very similarly to Mandarin, while retaining a completely different meaning. This almost suggests a degree of steganographical potential. Furthermore, it suggests that automated censors will not be able to handle this.
I’ve had to literally run for my life twice in 6 years from killer bees. They’re just a fact of life in this part of the country, and no one anywhere is immune from the threat.
I got away without even minor illness, so I guess my luck with killer bees hasn’t run out yet, but I will say this – the previous time was easier to digest mentally, because they were at least living in a common type of bee shelter that I had disturbed without thinking.
But the race I had recently is harder to forget. I was walking along a clear path, disturbing nothing, and a lawnmower was mowing grass maybe 40 yards away. Without any warning, and out of nowhere, I realized several extremely angry bees were up at my head trying to sting me, and probably stinging me already. I started my sprint out of there without actually having seen a single bee. (In fact, I ended up never seeing any of the bees – only hearing and feeling them, and finding some stingers that were left behind.)
I got a really good jump on the rest of the hive, wherever they were; I was sprinting away in probably less than a second. What’s actually scary though is later I found small bee stings (with no or very little injected venom) on my torso – meaning additional bees were still hounding me even as I had created some very good separation in such a short period of time!
I supposed your mileage might vary, but I ended up taking the longer path to my house, because the shorter path had some obstructions, and I didn’t want to risk tripping and falling. The longer path also afforded me the luxury of thinking less.
In the end, there’s really only one good explanation for what happened. The bees had probably become angry due to the lawnmower noise, and then they happened to find me by accident. I’m going to have to find the hive later (using a bee suit), but the truth is I have no idea where the bees came from (and if they’re even living on my property), and considering I was doing nothing where I’d expect a bee attack, this was pretty memorable.
In a surprise killer bee attack, humans basically have no advantage, but I’m a believer that getting a good jump is important, and a difference of a few seconds can mean quite a lot. I’ve figured that if you create enough separation fast enough, it can buy you a tangible amount of extra time to get inside (or in water), and reduce the number of stings, and can also magnify the consequences of any errors that the bees are making.
One thing I would recommend to everyone is that if you have a bunch of keys on your keychain, you should probably mark the key or keys that get you inside your house. I have a strip of green duct tape around the base of the key that opens my home’s outer door. It didn’t save my life this time, but in a scenario where I’m not as lucky, it could certainly make a big difference.
In a panic situation, all your fine motor skills will go away, and so you’re probably going to fumble around with your keys trying to find the right one.
The other thing that helped me in this case was the clothing I was wearing. There are no guarantees, but when bees sting through your clothing, they usually don’t inject as much venom. So that’s another thing to keep in mind.
And afterwards, I would run all the clothing through the washing machine immediately, take a shower, and not go back outside for a good period of time. You want to get the bees’ alarm pheromone off of you and off of your clothes, and also let the alarm pheromone that was released into the air dissipate.
1. The Safdie Brothers are quietly rejuvenating, re-imagining, and I would even say completely re-inventing the thriller genre, which is no small feat. At the current pace, that could easily be how their recent films are described in the future.
2. Mosul (2019; the one directed by Matthew Carnahan) is definitely underrated at its current 7.1 rating on IMDb. It looks at one part of the Iraq War, and it’s one that doesn’t directly involve Americans. For that alone, it’s already a different type of film.
3. I recently found out a prequel to Predator (1987) will be released this year, and it’s titled Prey. The plot involves the Predator hunting a Comanche tribe roughly 300 years ago.
I’m a fan of the original film, but I don’t think I’ve been excited for another release in this franchise in quite a while! I think the producers picked a great idea to run with, and I hope they pull it off.
I’m very interested to see where they go with this.
Just my two cents on what I would do with it, or what I’d like to see them put together: I would have the Comanche tribe slowly lose a war of attrition to the Predator, and when the tribe has completely lost hope, the elder chief (who’s perhaps on his death bed) pulls out on old relic that had been passed down to him (and that he had until now not understood), and the old relic is actually the severed hand of an even older Predator with the tactical nuclear device still attached. This leads to an abrupt but memorable ending where, somehow, the female warrior divines what it’s for and rides off into a suicide mission against the Predator – perhaps into an open field somewhere (with the tactical nuclear weapon counting down in the alien script). With a final display of bravado, she takes out the Predator, saving the remainder of the tribe.
Then, another prequel is released where we find out how that Predator hand came to be severed. The sequel would be much further back in time – perhaps as far back as the cavemen. I really like this idea because we tend to figure the cavemen were idiots, but it would be interesting to see how a group of cavemen might be especially wily and somehow overcome a single Predator.
That’s how I would run with the franchise.
But no matter what, the female protagonist in Prey absolutely needs to bring back the line, “If it bleeds, we can kill it.” I’m going to be very disappointed if they didn’t think to do that.
I actually liked the last iteration of the franchise – Predators (2010). I’ll give my take on it also, which involves a few spoilers.
I would have shaved maybe 20 seconds off the end of the film. The entire film was pretty bleak, and I don’t see why they can’t just be stranded on the alien planet at that point – without any hope of further human contact.
But what I would have really liked to do is completely change the surprise mole character.
I would have liked to see a twist on the “don’t judge a book by its cover” proverb. I would have liked the group to identify the Japanese warrior as a Yakuza member by the volume and nature of his tattoos, but I think it would have been a great twist if they didn’t look closely enough. In my opinion, the Yakuza member should have had one additional tattoo unrelated to organized crime – one for Aum Shinrikyo, the infamous Japanese death/doomsday cult mostly remembered for their sarin gas terrorist attacks in Japan.
So the idea is that his organized crime involvement, while perhaps congruent with his radical views, is maybe just like a day job for him, and what makes him the most notable is his involvement in that infamous cult.
The idea would be that the Predator clan handpicked him to be the mole in the human group due to his obsession with doomsday and human death, and so, ultimately, he’s essentially working with the Predators against the humans – most likely eagerly so, as he possibly perceives this whole thing as a sort of grand religious consummation.
Perhaps after this is finally revealed, we also see something like a telepathic link between the Predators and the death cult member, which would be a memorable first in the franchise.
Most of all, maybe the Yakuza member was handpicked by the Predator clan as a particularly twisted or poetic element in their competition – an element that would also be a rare demonstration of the Predators’ higher sentience and intelligence. The idea would be that this was a distinct challenge for the other humans – would any of these clever criminals, intelligence agents, and elite soldiers be astute enough to see past the “noise” of the Yakuza tattoos?
1.)In the technology world, the strength of security is often quoted quantitatively in terms of combinatorial complexity (e.g. an 8-character password with all normal keys available has a combinatorial complexity of 94^8), and qualitatively as if there is only one attack vector and the context for that attack vector is full access over the technology in question, with no significant time constraints, by a malicious actor.
That attack vector is more or less equivalent to scientists in a laboratory trying to attack some piece of technology you have.
That’s pretty much the worst case scenario, and so it makes sense to judge security by this standard prominently. But the world isn’t perfect – not even for malicious actors. There are many other attack scenarios that don’t resemble this full-control-with-no-time-constraints setup.
What seems to be missing in security analysis are metrics that are relevant to many other imperfect attack vectors that are liable to exist.
One metric, for example, might be coined time-pressure-complexity, and it would measure how long it would take an intruder to attack some piece of technology if the intruder also wanted to not remove the technology from its current setting.
And imagining a similar setup, another metric might be how much dexterity-complexity an intruder would face when attacking a piece of technology. This metric is not specific to passwords, but just as a simple example, let’s say one created a very long password that was easy for the owner to remember and type, but that would just naturally take longer for an intruder to enter. We can take that further and imagine the password not utilizing common typing patterns (e.g. languages) the intruder is used to. Even if the intruder knew the password letter for letter, he might fail to enter it correctly when under time pressure, which could lead to panic and a further degradation in that entire attack vector.
In a panic situation you tend to lose your fine motor skills, and a high dexterity-complexity would attempt to attack that weakness.
Although these are secondary metrics to consider, they’re still extremely relevant to the real world, and so it would make sense to have a proper accounting of them in many types of security analysis.
2.) Security analysis for multi-factor authentication often describes three separate notions: something you know, something you have, and something you are.
But I believe there’s actually slightly more to this security space than what the current model is describing.
I believe there’s another distinct notion: something you inherently own
Perhaps it would be a subset of something you have.
The difference is that if a malicious actor takes your metal key (something you have), it is now something that the malicious actorhas.
But, for example, if a malicious actor steals a phone with a SIM card tied to your phone number (perhaps utilized for 2FA), you will, generally speaking, have the ability to get that SIM card invalidated by the carrier and have your phone number re-associated with a new SIM card.
To a significant degree, you inherently own that phone number – it’s not just something you happen to have.
Although that was an overly simple example, it demonstrates an additional security notion and a meaningful difference between it and its closest relative.
3.) Major mainstream technology providers (e.g. email providers) offer a sizable amount of security options, ranging from rustic passwords to biometric technology. But I believe the not-so-state-of-the-art communication space therein is serious cause for concern.
Currently, the consumer/user is not provided clear guidance on what’s going on with their security options.
There are generally three different times an email provider will bring those extra security options (e.g. phone text message, biometric) into play:
account recovery (can’t log in to your account)
trying to change security settings for the account
The technology providers allow the user to define their additional security options, but they do not provide any guidance at all regarding what will happen in each of those three different scenarios, and if I’m not mistaken, there are differences.
What will specifically happen in each scenario should be communicated clearly to the user, but currently the system is unclear, ambiguous, confusing, and opaque; you set up your security options and then find out later exactly how they’re employed in each individual scenario.
In my opinion, three visual diagrams should be shown to the user – side-by-side if possible – depicting what will occur.
Knowing how those options are actually employed in those three different scenarios can definitely affect the user’s decision making, and it’s just generally a good practice. But unfortunately the providers are not living up to their affirmed security and corporate responsibility principles when it comes to this.
As a final quiz, imagine a user who has set up a mainstream email account (of your choosing) with a password and a garden-variety authenticator app. The user forgets the password, and also, his phone is completely destroyed. (In this example, it sure is especially clear the authenticator app is something he had, not something he inherently owned.) How do those three security scenarios play out? Don’t know, do you? But shouldn’t you? Additionally, if a phone number was also required to set up the account, perhaps to save the user in this situation, would that or would that not contradict the theory that authenticator apps are superior to text messages for account security? Whatever your answer is, it should serve to further articulate the distinction between something you have and something you inherently own.
My Contribution To The Mnemonics State Of The Art?
In my previous post, I talked about some products designed to improve mental speed and memory. As an absolute beginner, I don’t anticipate too many novel insights on my part, but I did discover a few things that didn’t seem to be covered, and might be worth sharing.
(If you aren’t familiar with the subjects in that previous post, the following won’t be very useful to you.)
1.Kevin Trudeau’s Mega Memory teaches the listener how to organize memories so they are easily accessible later. With the methods that are shared, one tends to end up with a lot of different ordered lists. (Ordered lists are a strong motif in the course, but it’s not the only memory application taught.)
For example, one might memorize all the U.S. presidents in order, the periodic table of elements, the human bones, and so on. For a task like that, at least two memory systems were taught. Using the primary system, each element of each list goes onto a “peg”.
You develop various “peg lists”. The peg lists are ordered. Some peg lists might be 10 pegs long. Others might go up to 100, or even more. The pegs are where you place the individual items to remember.
You pick the peg list you feel is best for the ordered list of data to memorize, and then, using specific criteria, go element by element and associate the ordered elements with the ordered pegs, pairwise.
But with even just a few ordered lists, it seemed to me like sometimes one of the challenges is remembering which peg list you used. And furthermore, it seems to me that when you do remember the peg list that you used, and then remember the first peg/item memory that you created, the rest of the list, at that point, gets recalled relatively easily.
Therefore, if the first peg/item association has greater significance, greater effort should go into creating it.
Taking the example of U.S. presidents in order, George Washington will be the first item. As far as the peg list to use, the listener can pick from various peg lists that were taught in Kevin Trudeau’s MegaMemory: the “tree list”, the “body list”, the “house list”, and also a custom list based on the major system.
The best list to use virtually gives itself away in this case. The “tree list” begins: tree, light switch, stool, … (Although that list only has 20 pegs to begin with, the listener is challenged to add more, and in this case will need to create 26 more.)
The “body list” begins: toes, knees, muscle…
(My) “house list” begins: chest, dresser, map…
The major-system list begins: tie, Noah, mow…
The list to use really is a dead giveaway. The “tree list” is absolutely what should be used. George Washington and the cherry tree he felled are part of the national consciousness. There’s even a little bit of scandal and drama, and that makes it even better. His act of chopping it down will be the memory, that, in this case, doesn’t even need to be dreamed up; all the hard work has already been done for us.
Several years later, when you want to randomly recall the list of U.S. Presidents, and you have also memorized hundreds of other lists, I believe it will be very important to remember the peg list that was used, and the first peg/item association. And since a natural and intuitive first peg/item association has been created (more like adopted in this case), that will not be a problem.
Diving a little deeper into cognitive psychology, the instructor talks about how once you mentally get a handle on something related to what you want to remember, it’s like grabbing ahold of a chain that you can start pulling on. So the idea here would be like we are trying to make that first chain link big and easy to find.
If I were daring enough to create terminology for this concept, I would label this first peg/item association the “door association” and/or the “window association”.
In my opinion, it’s tangibly more important than the rest of the associations to come, and so additional care should be given to it.
Finally, perhaps there is also a generalization of this idea, or at least an application for any level of granularity. Going down the presidents list, I realized I could recall the other presidents easily, but for some reason 27 was difficult.
My 27th peg in the tree list is morion, and there was no trouble remembering that. In the presidents list, William H. Taft needed to be associated with it. My visualization was an NFL player in the NFL draft (“Taft”) being selected at the podium in a celebration, and then he kicks a morion into a TV and it goes into the TV programming, which is showing an old-style H-shaped field goal (“H.”), and it bounces around the goal posts and then through the goal posts, and then the morion finally hits another visual representation I have for “William”. The reason 27 was taking longer to recall is because the morion was not immediately active.
Having to wait for the NFL player to hold up a jersey at the podium (to signify this was the NFL draft) was delaying the introduction of the morion, which was the mental peg. When I got to 27, my mind was searching for action involving a morion, but my visualization started with action not involving a morion, and also made it so that the morion’s action was not obviously the most significant action – the draft selection could be viewed as equally significant or even more significant which just creates more “noise” that distracts the mind. (And to make matters even worse, there was almost no memorable connection between the draft selection and the player then kicking a morion into a TV.)
Just a split-second delay, and just an ounce of ambiguity can stall the memory recall, or prevent it altogether. Therefore, applying the “door/window association” principle at this level of granularity would mean etching the mental peg first and foremost within the visualization, and also striving to elevate its significance as much as possible.
2. In general, and certainly in specific situations, spend extra time creating multiple mental linkages.
The custom major-system list taught in Kevin Trudeau’s Mega Memory goes up to 100. Pegs 60-69, as taught, are below:
I don’t think that list is very good, and I’ll elaborate more on this shortly. Furthermore, 60-69 is especially difficult because the words can begin, phonetically, with three different consonant sounds (ch, sh, j). That alone makes this series of 10 pegs more difficult to remember than the other 90 pegs.
It’s worth the effort to improve the list.
Here’s my customized 60-69:
(For me, “china” as in “fine china” is what works best, but for someone else, it could just as easily be uppercase “China”. )
Here’s why I prefer the customized list that I created:
• Much better verbal flow.
“cheetah” verbally flows/rolls to “china” much more naturally/fluidly than “chain”. “cheetah china…” just rolls off the tongue much more naturally than “cheetah chain”, and most people will quickly realize that when trying to memorize about 100 other pegs.
Good verbal flow from “shim” to “shore”.
And so on.
• Much better semantic flow.
The semantic link from “shore” to “shell” is very good.
The semantic link from “shell” to “judge” is vaguely kind of good once you consider the idea of a fraudulent shell corporation.
And so on.
• Also, just better individual pegs.
For example, the mental image of a “sheikh” was bound to look like an abstract Arab, and this would later present a problem when trying to remember why in the world you have a mental picture of an “Arab” in a memorized list you’re going through. On the other hand, a (horse) jockey is a much cleaner, crisper, louder, unambiguous image, unlikely to be confused with anything else.
But on the other hand, “shore” will definitely be confused with “beach” later on, unless you devise a way to disambiguate it. My method for that was to visualize “shore” as very rocky, without sand, without people, and overcast – basically the opposite of a beach postcard teeming with humanity.
Sub-item #1 (better verbal flow) and #2 (better semantic flow) constitute the additional mental linkages that make the updated list superior, in my opinion.