Open Letter to Ipsos: Get Your Act Together

I recently got a letter from Ipsos, the polling people.  They must be spending like nobody's business, because they included a dollar bill in their envelope.  Sure, that's not a whole lot for me individually, but that can add up in a hurry.  See Riemann Sum.  They're taking a gamble, knowing that direct marketing like this has an abysmal return rate, like in the single digits of percent.  Perhaps it helps a little sort-of guilting people into taking action.  After all, you got paid a dollar, shouldn't you in turn provide some value for this back to Ipsos?

To quote Herb Tarlek: OKfine.  I did really want to participate, measly $1 payday notwithstanding.   So I sat down at my computer and typed in the URI provided for the online "version."  First, I will warn you, Ipsos, you need better IT people...MUCH better IT people.  It is extremely easy, at least at the topmost level, to program defensively around things like content blockers (see NoScript).  If you are going to have dependencies like JavaScript, you darned well better make sure those dependencies are there and working.  And that applies to any subsequent levels too, such as that JavaScript file you just had the browser fetch specifying more code to be fetched, or you using JavaScript to insert more into the DOM to trigger other JavaScript to be loaded.  It's just not all that hard to pick a function or variable from these inclusions, then inside a try/catch block, referencing these objects.  Even if you think it's not "worth it" to expend resources to program defensively like this, consider that even "normies" are getting ever more privacy conscious, and will seek browser settings and extensions which try to block as much as feasible.

After entering in the code number they sent in the mail, I was greeted with a page that didn't seem to be working right/as intended.  So sure enough, I went to my NoScript icon, and found that I probably needed to let through (at least) the first level of JS loads.  That in turn came up with a page, paraphrasing here, that it looks like either a.) we already have enough people who have answered the survey, or b.) there was an error.

Huh.  OK.  I've encountered this before, the design of certain sites just will not work in my browser of choice, even if I turn off or disable the page-altering things, like uBlock Origin, NoScript, Tampermonkey, and Stylus (maybe others).  So, I keep Brave around for just such circumstances.  As many know, Brave is Chrome/Chromium/Blink/V8 based, but also tries to be the browser of choice for many wanting to preserve privacy.  The difference for the profile I normally use is, it doesn't have any of these aforementioned tools installed which make the modern Web a lot more tolerable.  It's pretty much, allow most if not all JavaScript, and only block those things which the Brave people find objectionable.  Mind, the Brave people want as few sites as possible to break outright because of their privacy preserving measures.  So this usually does the trick.  Usually.

But no, same result, I put in my code, the page dutifully informs me that the survey is done, looks like you, Ipsos, don't need my opinion.  OKfine.  At this point, you have gotten me to stop caring.  Your message mostly clearly tells me this.  You've sapped my willingness to retry.  Until....

...Until some days later, I received a postcard-sized piece of mail from you urging me to take the survey.  Whiskey tango foxtrot.  Uhhhh...which is it?  Are you done with that survey, or do you want my input?  Okey dokey, let's run Brave again.  I put in my code.  I get told once again that either the survey is over or there was an error.  Once again, you have sapped my willingness to retry.

There is something FUNDAMENTALLY wrong with this.  I do not know why you have chosen NOT to make "survey over/not needed" and errors distinct.  It reminds me of the old DR/MS/PC-DOS error "bad command or filename."  Surely you, command interpreter/shell, DO know the difference, and could tell the user which is the error, but the programmers who programmed you are too lazy to make the distinction.  If you, Ipsos, REALLY do want people's opinions, why are you not making the distinction, and for true errors, reporting so and urging the participants to try again later?

So, what choices do I have?  You have an email address where I can ask questions.  Yeah, no thanks.  I get enough spam as it is, I don't need to give you yet another address to which either you or some other entity you're working with to send additional emails.  You also list a toll-free number and a +1(202) number.  You don't answer either of those with a human, you instead want me to leave a name/other identifying information and my number.  Again, no thank you for the same reason, I get enough calls from entities I don't want to talk to, and I really have no assurances that my email address or phone number won't be given out to anyone else.  You only make assertions about the contents of the survey ("And don't worry, everything you say in the survey will be kept confidential").

This needs serious refactoring.  If nothing else, if it would indeed be part of your standards and practices, you could also tell people in these letters and postcards that any emails, email addresses, conversations, and phone numbers are also confidential.  Don't take the lazy way out, make errors distinct and report them as such.  It shouldn't be a bunch more effort to code more defensively.  And for gawd's sake, do not just simply tell the user "something' went wrong, tell them EXACTLY what went wrong, or at THE VERY, VERY LEAST, provide something on the page, like a link which will change "display: none;" into display: block;", that explains PRECISELY what went wrong so that users will have more incentive to give you the data you want by making adjustments (for example, allowing more to pass through NoScript, or turning off extensions temporarily).

You want the data (as in, survey takers), you have to put in more effort to get the details right.

---

English is a difficult enough language to interpret correctly when its rules are followed, let alone when the speaker or writer chooses not to follow those rules.

"Jeopardy!" replies and randomcaps really suck!

Artificial Intelligence Is (Still) Half-Baked

If your car failed to start, let's say, 8% of the time, you'd be looking at your state's lemon laws and demanding your manufacturer fix it.   (I suppose it might be province's, prefecture's, oblast's, etc., but for Yanks like me, it'd be "state's".)  Why major information providers and software authors have chosen to hop on the AI bandwagon, sometimes I don't know.

Don't get me wrong, these AI services are actually quite useful in certain contexts.  For example, after doing a search engine search for something, and getting far too many irrelevant results, even after tweaking the search query, it's a fair bet that Gemini, Copilot, ChatGPT, Claude, Rufus, Leo, etc. can cut through the noise and produce useful results.  And it's also quite useful that they're conversational, in that they remember previous prompts, and all one needs to do is reference them, if only indirectly, to refine the answers.  But sometimes, you have to wonder why companies have bothered at all, and they're quite useless.  I have two instances I'd like to relate.

The first is pitting Siri against Google.  I was driving around in unfamiliar territory, in the city next to my hometown.  I was employed at the time by a company whose standards were 1.) that all their IT staff carry a cell phone (so that they can be, minimum, consulted any time, day or night...i.e., be on call), and 2.) that it be Apple (due to their track record of lack of security bugs/compromises).  No worries, they bought the handset and paid for the service directly (instead of previous companies, which would have you do that and submit expense reports).  I tried, hey, Siri...tell me about such-and-such (it was a while ago, I can't remember exactly what the query was about).  It gave me irrelevant after irrelevant answer, even after trying to refine my question.  Finally, I set up the Android tablet I had with me to be tethered to the iPhone, and asked, hey, Google.  It had far more relevant results.  Yay, Alphabet.  I wondered why people liked Siri so much considering its popularity.  I find it to be more frustration than anything else.

The second happened yesterday.  I was in a certain multinational discount department store whose theme is blue, their logo is six yellow rays coming out of a central spot, seeming like the Sun, on a blue background ;-), and their workers wear blue vests.  In fact, they have an (Android) app, which I wanted to find clothes hangers.  I have used the app before; often if you ask for something, it will at least give you an idea of where to look by pointing out some product's aisle designation (ultimately turned out to be H30 in this case) given that the store you were in was selected.

But such is their zeal to sell you SOMETHING that they'll include mostly irrelevant search results.  For starters, they quite understandably have, like many retailers with apps and Web sites, ship to store, pickup, and order for delivery.  I am pretty sure I had selected the correct store location and pickup.  The results showed me hangers that I guess they sell somehow, but (of course, why not?) would show not available for pickup.  I'm going to guess the ones listed first have the highest GPM, to encourage you to buy them and therefore earn the most money from the sale. Okay, scroll, scroll, scroll, tap on a few of the results, still no aisle designations on any of them.  I just need to get in the vicinity of any of these products, I'm sure they're not going to scatter hangers all over the store.  Indeed, a lot of the product pages will show you they're in aisle such-and-such.  Finding that designator is another story altogether, like, where are the G aisles, where are the H aisles, and so forth.

What's this?  They have an AI assistant?  Great.  Let's give that a whirl, maybe it'll be like getting too many search engine results, it'll be helpful.  Where would I find these?  Well, you'll find them in closets, and in hotel rooms, and...(on and on about general, non-helpful things).  No, no...where are these within the store (paraphrasing)?  Typically they'll be found in the housewares.  Well, duh, I know you don't know this, but I am standing somewhere in housewares.  What is the aisle designation where I can find this product?  Typically, they'd be in with housewares.  Grumble.  I had to go find an employee and ask them.  Even they were at first less than helpful by pointing and saying down that way.  Do you know the aisle designation?  Nope.  We ended up walking there together, where I said to her, ah, yes, thank you, H30, noting the sign for the aisle.

Hey, discount department store IT guys and gals, or their subcontractors:  I don't need "typically," I need specific information.  And why the F do you not have easily found aisle designations on each and every product...like right near the top of the description?  Or, if it happens to be a not-in-person, only-online product, how about something like "Aisle: (online only)"?  Why, oh why, was it considerably easier to find and to ask an employee instead of using your mostly worthless app?

I am also looking at you, Rochester, NY-based popular supermarket chain whose logo is a stylized W and also has an Android app.  You tell me a certain cheese product is "in dairy".  Have you looked at the size of your dairy section?  Don't you think I might have been through your dairy section if I'm looking for cheese?  It's kind of like telling people how to get to your store by telling them it's in Depew.  Okay, that's not about AI failure, but still, it's a similar sort of app and organization failure.

If I recall correctly, it was Andrew S. Tanenbaum who wrote in his seminal OS design book with regard to input buffers for typing, those who did not program them into their OS ought to be forced to use exclusively their own system.  Both the supermarket and discount department store IT guys and gals ought to be, on a regular and frequent basis, forced to find something in their respective stores with only their company's app.

---

English is a difficult enough language to interpret correctly when its rules are followed, let alone when the speaker or writer chooses not to follow those rules.

"Jeopardy!" replies and randomcaps really suck!

Bye, Bye, Chrome, Hello Firefox (or, Out of the Frying Pan and Into the Fire)

Well...tonight "it" finally happened.  I did an apt-get update followed by going into aptitude to apply the latest updates.  Among them was google-chrome-stable.  Next time I ran it, it promptly told me 3 of my extensions were disabled because...well, don't know the exact wording, something like they're unsupported.  uBlock Origin, EditThisCookie, and SQRL all fell victim.  It's the first one of those which is my dealbreaker.

Google have apparently finally pulled the Manifest V3 trigger.  The author of uBlock tried for a while to support a V3 compliant "lite" version of the extension, but ultimately concluded it's not worth his time, effort, and the headaches. So, uBlock for Chrome is, as best as I can tell, dead.  Other Blink browsers have announced they'll try to support V2 for a while longer, but are not to my knowledge making hard and fast promises.

I have been planning this transition for months.  I have already been running Firefox as my preference at home (on Linux), but at work (Win 11), for the longest time, Chrome was my go-to.  I figured, the sooner I start the transition, the less I'll be fumbling around when my must-haves won't work anymore.  So I set up Mozilla syncing accounts, one per purpose (home, work), and several profiles (home, work, work as admin, home with alternate email address, etc.).  I got used to the less convenient ways of switching profiles (buttons on the builtin about:profiles page) and the less convenient Firefox View for finding open windows, tabs, and pages.

I'm...kinda...fine with where I am, except for the recent Mozilla change in terms of use.  So...anything at all I use FF for can potentially be used by Mozilla in just about any way they please.  Anything I type, any UI elements I click, any features on the pages I visit with which I interact, heck, maybe even the URIs themselves are potential fodder for...who knows what.  My inkling is it's for LLM or other AI training.  Still, their new policy is a bit broad for my tastes, but at this point, it's the least of all evils.

With so much of the world swirled into the Blink/V8 universe, with only Mozilla's Gecko and Apple's WebKit to act as competition, I can hardly wait to see what Ladybird is like.  Seems like that's gonna be "a bit" before it's a usable program though, they've only just begun.

---

English is a difficult enough language to interpret correctly when its rules are followed, let alone when the speaker or writer chooses not to follow those rules.

"Jeopardy!" replies and randomcaps really suck!