Miscellany

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 many disparate 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.

Great Design

Coffee Mixology

I’m kind of a lukewarm follower of technological advances, and one of the trends I finally tried out last year was a single-serve (pod) coffee maker.

The product I bought was a basic dual-use coffee maker by Ninja, and I would recommend it to anyone.

Just based on the convenience alone, I am never going back.

However, there is another useful feature: you can easily mix different coffee flavors together, or, for example, coffee and hot chocolate.  That could be a fairly big hassle to do with standard kitchen equipment even once, much less every time I happen to feel like it.

All in all, the coffee costs more (and unfortunately it’s still worse for the environment), but it’s definitely worth it.

Great Design

If You Had To Learn Spanish From Scratch…

I will diverge from standard academic thought, and say that if one is learning Spanish from scratch, or essentially from scratch, then they might as well learn “standard” conjugations for the vos person (not to be confused with vosotros) – which are only used regionally (most notably in parts of South America).

I would say the cutoff point is when one finally commits to memorizing conjugations for regular Spanish verbs.  At this point, one can easily include conjugations for vos, along with everything else.  These vos conjugations often replace conjugations in regional Spanish.

The reasons academia has not done this include the following:

•  the demand for vos is less

•  it increases learning complexity

•  vos is used very differently across the regions that do use it

I decided that thinking is wrong based on the following observations:

 a significant amount of print and film media is produced in countries that use “standard” vos – in Rioplatense-Spanish-speaking countries in particular (notably Argentina, Uruguay, Paraguay, and Bolivia)

•  the increase in learning complexity is actually a function of the learning method; a bad method (rote memorization) means incremental additions to the teaching material are (unsurprisingly) difficult, but a good learning method means incremental additions have a negligible cost

  the problem of regional differences is sidestepped if you simply teach the “standard” version of vos; furthermore, this standard version is a reasonable foundation should you want (or be forced) to try your hand at the other regional vos dialects (e.g. Chilean Spanish)

Great Design

Tonic Mnemonic: NOISE Soon

In Spanish, the tonic syllable is the syllable that sounds loudest when pronounced.

• When an accent mark (tilde) is present, it is very clear.

sofá:   so – FA

• When then there is no accent mark, a set of rules must be used instead.  When the word ends in a consonant other than n or s, the last syllable is the tonic syllable.

feliz:   fe – LIZ

• And when the word ends in the consonant n or s, the second-to-last syllable is the tonic syllable.

oasis:   o – A – sis

So how do we remember the n-or-s rule for penultimate syllable stressing?  As usual, from a hodge-podge of mostly complementary mnemonic aids – it’s not necessary to remember all of them for the mnemonic to work.

First, noise and soon both have the operative consonants n and s.  These are the only consonants the mnemonic is keyed to.

Second, when “noise is soon” to occur, you obviously cannot be on the last syllable yet.  So we’ll say you must be on the penultimate syllable.

Third, there are two words in the mnemonic itself (noise and soon), which corresponds to the number 2, and the number 2 corroborates the penultimate nature of the mnemonic.

Fourth, the mnemonic words themselves, conveniently enough, have only two consonants (n and s).  The number 2 once again corroborates the penultimate nature of the mnemonic.

Finally, in the mnemonic phrase, the (capitalized) word NOISE is the only word out of the two words that you would expect to have tonic emphasis.  By its very definition, noise would have audible precedence over a temporal concept like soon.  So from the reading of the mnemonic itself, once again the penultimate syllable is stressed.

Great Design

Language-Learning Efficiency

I recently realized that the most efficient time to start reading a book in a foreign language is after you can reliably pronounce words in that language (at least to a reasonable degree).  That capability allows you to get near-instant translations from translator apps on your phone via voice commands.

Great Design

El Dorado Is In The Mind

Technically, El Dorado – if it exists – is massively distributed underneath South American soil and in similar South American sites, at locations that only Inca record keepers had memorized.  According to a documentary I just saw, this is a close and in fact probable approximation for the El Dorado that many (including myself) had considered to be mythical.

When Pizarro held the final Inca Emperor Atahualpa hostage, and demanded gold, an untold amount was brought to him, from parts unknown, and without any understanding of how it could have possibly been stored somewhere.

The best theory is a large, distributed set of Inca record keepers memorized countless hidden locations where portions of the mammoth Inca gold reserves were individually secured.  After Atahualpa  was killed, the details of any such system may have died with him.

The suspicion is much more Inca gold remained hidden throughout that part of South America, and in fact treasure seekers still haunt that region and occasionally find more Inca gold that unfortunately ends up on the black market instead of museums.

Great Design

Use The Bottom Half Of A Spray Cleaner Bottle To Soak Paintbrushes

I needed to soak a paintbrush in paint thinner overnight to clean it.  While it’s said that used paint thinner is still usable, my goal was to use as little new paint thinner as possible, and no typical container really met that goal.

I find it hard to believe no one else has ever thought of it, but none of the most popular internet articles and instructional videos suggest using the bottom half of a spray cleaner bottle.  In fact one of the more popular videos – by someone who obviously had a lot more experience than me – illustrated how one can balance a cylindrical can on its edge precariously so that the gasoline he put inside the can would soak more of the paintbrush.

We had dozens of different spray cleaners – 409, Lysol, generic brands, and a near-empty Frosch-brand cleaner.  The Frosch bottle had a thinner profile, fit the paintbrush like a glove, and worked perfectly.

Great Design

Diphthong Mnemonic: The Spanish Language Is Greater Than You & I

In Spanish, two vowels side-by-side may be pronounced as one syllable or two syllables, depending on the vowels in question.  (Note that the order of the vowels doesn’t matter.)

“Strong Vowels”A  E  O

“Weak Vowels”:  U  I

• 2 strong vowels together = 2 separate syllables

aeropuerto:   a – e – ro – puer – to

• 2 weak vowels together = 1 syllable (diphthong)

ciudad:   ciu – dad

1 strong vowel + 1 weak vowel together = 1 syllable (diphthong)

pueblo:   pue – blo

An accent mark on a weak vowel will undo what would otherwise be one syllable.

día:   di – a

The true pronunciation of aeropuerto still kind of messes with my mind.  I also used to think of ciudad as 3 syllables.  That’s the whole idea behind having the mnemonic.

As for the mnemonic itself, it’s simple and pretty self-explanatory – just convert “you & I” to “U & I”, and those are the two things noted as being smaller or weaker than something else.  It helps you remember which two vowels are the “weak vowels”.  As for the rest of the rules, one could make up something like the following: Two “strong vowels” side-by-side will each demand their own limelight; they each demand to be pronounced separately.  Conversely, two “weak vowels” will band together to be pronounced.

Of minor interest is the fact that if you perhaps disagree with the jingle, in its surface-level reading, then that mental dissonance would, in general, probably create a more effective mnemonic for you.

Great Design

Committing To Advance The User-Experience State-Of-The-Art

Like millions and perhaps even billions of other people, it happens to me at least several times a day: I’m prompted by a computing device to make a decision I don’t care to make.

In most cases, I don’t want to make decision.  In fact, I don’t want to even read about what the decision is about, much less make the decision.  But if I had to make a decision, I would usually just want to keep the current state of whatever it is the prompt it talking about.

But of course, the prompt almost never tells you what the current state is, nor how your potential decision might affect the current state.

A lot of times, there is an ‘X’ somewhere that you can click to make the prompt go away, but even that is not ideal.  I usually have to think for a split second and translate that ‘X’ in my mind to “probably non-committal so I’ll click that”.

The industry standard for user experience (“UX”) ought to be updated such that this entire idea is finally recognized as a first-class concept.

Specifically, there ought to be a separate button, that is at least as prominent as the other buttons, and which says something to the effect of “Non-Committal”.

The collective amount of time that would probably save humanity can not be under-estimated.

Poor Design

Cheating Windows Out Of Cheating You (The Fingerprint Reader)

For all I know, this is a well-known problem with an easy solution, but after searching online, it seems like many people have the same problem, and nobody has a solution.

The problem, for some reason, is this: If you have a fingerprint reader (as a faster log-in method so you don’t have to type your password), and the sensor/Windows doesn’t verify your fingerprint after about 3 tries, Windows will require you to type your password.

It seems like Windows is being insane.  I can’t think of a reasonable security rationale for not allowing unlimited fingerprint tries, when of course you have unlimited password attempts.

The workaround is this: Alternate the fingers you use in between unsuccessful fingerprint verifications.

If finger #1 doesn’t work, try finger #2 next.  If that doesn’t work, go back to finger #1.  Windows allows you to register at least two fingerprints, so everyone who has this problem ought to be able to use this workaround.

When you use this workaround, the number of tries that Windows gives you goes way up.  It seems like more than enough to eventually get a successful fingerprint verification, at which point you’re logged in.

Of course, this makes the whole thing seem even more insane.  If there were no security rationale to begin with, then I really, really have no idea how to describe the security rationale for this extended workaround.