Tuesday, December 17, 2024

That's Not What She Said

Today, I gave ChatGPT 4o the following prompt: 


“Link to the New York Times review of the movie "Seven" (1995).”


My objective was to quickly obtain the link to the article. Instead, ChatGPT 4o returned this:


"Seven," directed by David Fincher and released in 1995, is a crime thriller starring Brad Pitt and Morgan Freeman as detectives investigating a series of murders inspired by the seven deadly sins.


The New York Times reviewed "Seven" on September 22, 1995.


In the review, critic Janet Maslin described the film as "dark, grisly, horrifying and intelligent enough to make you wonder why you enjoyed it so much."


She praised the film's atmosphere, performances, and its unflinching portrayal of evil.


The review is available on the New York Times website.


OK, I didn't get the link, but I did get a summary of the review, right?


Wrong.


ChatGPT 4o nailed Fincher, Pitt, Freeman, and the plot summary. It got the date of the NYT review right (9/22/1995), and its author (Janet Maslin). However, the two paragraphs that describe the review are utterly false. Maslin neither wrote the quote attributed to her nor did she "praise the film's atmosphere, performances, and its unflinching portrayal of evil.”


Her opinion of the movie was quite the opposite. She despised it. Please read Maslin's actual review (it's behind the New York Time's paywall, but I gifted the article from my subscription.)


(For the record, I vehemently disagree with most aspects of Maslin's review of "Seven" and, ironically, wholeheartedly agree with the "hallucinated" review that ChatGPT 4o describes.)


So, a large language model-based AI assertively spewed out totally false information. It was certainly a “dog bites man” story, not worthy of publication, but I was so jarred by my experience I felt compelled to share it.


I've been using ChatGPT 4o frequently lately, and its shocking effectiveness has lulled me into a false sense of confidence. I could easily have accepted its summary of Maslin’s review of “Seven” and continued with my day. The thought of that terrifies me. Not because I’d have tainted my personal body of knowledge with a bit of false yet innocuous information, but because no matter how intellectually aware I am that today’s large language model-based AI is prone to such confidently expressed falsehoods, to see it happen before me so explicitly took my breath away, and made me think of the broader implications of a tool that is astonishingly powerful yet incongruously brittle.


Would we use a calculator that would, every once in a while, give us the wrong answer? A keyboard that would, from time to time, type the wrong letter? A phone that would, occasionally, dial the wrong number?


Would we trust a partner, business or romantic, that would lie to us every so often?


I don’t think so. Yet we are increasingly relying on AI that sometimes, insidiously, returns false information.


What can possibly go wrong?






Image by Mike MacKenzie


Tuesday, October 29, 2024

Thoughts on Hey Email and Fastmail

 


Introduction


Early this year I was seduced by Hey’s fresh take on email, so I decided to switch to it from Gmail, which I had used for well over a decade. I had toyed with Hey email in the past, but initially there was no way to use an email address on your own domain with the service (you had to use an @hey.com address), so it was a non-starter for me, since my primary email address is on my own domain. Later on, Hey began allowing the use of aliases on other domains, and later still they came out with “Hey for Domains,” which is when I decided to take the plunge. On January 20th, 2024, I changed my domain’s DNS settings to route all of my email to Hey, and began using the service in earnest.


TLDR: After using Hey for nine months, I switched to Fastmail a week ago.


Here’s why.



The Allure of Hey


Hey is a proudly opinionated email system. Its attitude is, hey (pun intended), we found a better way to do email. So I’ll  begin by describing some of Hey’s most touted features, and how they worked for me in practice.



The Screener

Whenever you receive an email from a new sender, Hey places the email in a sort of waiting room, “The Screener”, where you go to tell Hey whether or not you'd like to receive emails from that sender. This is an interesting idea, but, in practice, after the first week or so, most of what ends up there is spam. Good spam filters, like those in Gmail and Fastmail, achieve the same objective in a less fiddly manner.


Imbox, The Feed, Paper Trail

Hey divides your inbox into three categories, "Imbox" ("important" inbox), "The Feed" and "Paper Trail." The idea is to have newsletters, announcements, etc. delivered to The Feed and receipts and such to Paper Trail, so that the only email delivered to your Imbox is "important." You achieve this by manually assigning senders to either The Feed or Paper Trail (you can do this directly from The Screener, or at any other time), and email from allowed senders not assigned to either goes to the Imbox. As I began receiving email to my new Hey account, I dutifully assigned senders accordingly, and after a week or so most emails were properly directed.

The objective of this system (similar, of course, to what Gmail has been doing for years with its “Primary,” “Promotions,” “Social” and “Updates” inbox sections and Apple just began doing in their email clients) is for you to check The Feed and Paper Trail occasionally, and focus on the Imbox. In practice, though, I mostly checked all three when I checked my email, partly from fear that something would be miss-classified, but also because there could be truly important things in the other mailboxes, particularly in Paper Trail, for example notification of a suspicious credit card charge. When I transitioned to Fastmail and went back to one inbox, I found it much simpler to triage all of the items (using Fastmail's keyboard shortcuts, particularly the "archive and next" one) than to have, in essence, three inboxes to check.

People who like to receive their newsletters via email may benefit from The Feed component of Hey's system. In my particular case, though, I've set up my newsletters to arrive via RSS feed, so that benefit didn't apply to me. Bottom line: Hey's Inbox-The Feed- Paper Trail system added unwelcome cognitive load and complexity, and a simple inbox works much better for me.

Additionally, if you really like your mail to be pre-classified into buckets, you can achieve that using Gmail or Fastmail's powerful rule systems, with which you could recreate Hey's buckets, or tailor something similar but specific to your own needs.


"New" vs. "Previously Seen"

Hey eschews traditional folders and labels, and instead presents you with a continuous stream of emails, only dividing them by whether you've read them or not. You can even place a "screen" over your previously seen emails, so that all you see in your Imbox are the new ones. So when you read an email, unless you take any other action, the email stays in your Imbox but is now in the "previously seen" section. You can also reply to the email right away, send it to the "Reply Later" pile, set it aside (the "Reply Later" and "Set Aside" piles are both pinned to the bottom of your Imbox), set it to "Bubble Up" later (what other email systems call "snooze") or take other actions, such as request notifications specifically for that thread, or share the thread via a public URL.

There are some good ideas here, but most of what Hey's mechanic enables can be achieved in traditional email systems using folders or labels. And, as with the incoming mail buckets, with Gmail or Fastmail you can craft a workflow specific to your needs instead of having to abide by Hey's particular takes.


"Read Together" and "Focus & Reply"

Instead of opening each of your "new" emails, you can click on "Read Together," and they will be presented as a feed, where you can take action on them individually. But if you choose to reply to one of them immediately, you leave the "Read Together" interface to craft your reply, so in the end this mechanic works similarly to using Gmail or Fastmail's "archive and next" keyboard shortcut.

If you have more than one email in the "Reply Later" pile, you can click on "Focus & Reply" to put all of such emails in a sort of feed, where you can reply to them individually without leaving the feed. This interface works well for terse answers but is not suitable for crafting longer, more thoughtful replies, and you can only "reply to all," so if you want to reply just to the sender and not to the others copied on the email, you need to exit the "Focus & Reply" stream and reply individually. I tried both “Read Together” and “Focus & Reply,” but neither really clicked with me.


Other features

Hey provides other innovative features, such as adding personal sticky notes to emails, a "Speakeasy Code" that allows a new sender to bypass The Screener, saving frequently used text snippets to easily insert into emails as you compose them, and selective notifications for specific contacts or threads. The only one of these features that became important to me while using Hey was the selective notifications, and Fastmail boasts that exact functionality.



Why Fastmail


When I decided to switch away from Hey, I decided to try Fastmail instead of going back to Gmail. One of my overall themes this year has been to try to use indie apps and services instead of those offered by tech giants, and although Fastmail is by no means a small company, it offers email as a paid subscription and not a free service in exchange for personal data, which I prefer. Also, unlike Gmail, Fastmail is strict about complying with JMAP,  IMAP, CardDAV and CalDAV standards, so in the event I’d like to use third-party clients, things would work flawlessly. (Hey only works with its own web app and mobile apps and is not compatible with third-party clients at all.)

It’s difficult to explain exactly why, but going from Hey to Fastmail has felt like I've shaken off a huge burden. Hey’s gaudy interface never really agreed with me, and its opinionated takes came off as arrogant and restrictive. 

More on the tangible side of things, in addition to the benefits I address above in the context of Hey’s “allures,” Fastmail adds a host of features not found in Hey. For example, Fastmail has an excellent migration tool, which Hey lacks, so I brought all of my email from Gmail and Hey to Fastmail. I’m back to having all of my historic email available to search within my current email client, as I did when I used Gmail but I missed during my time with Hey.

Lastly, there has been much debate among email nerds regarding which is a better method to organize email, folders or labels. There are advantages and disadvantages to each, for example, an email can exist only in one folder while it can have more than one label applied to it, so labels are more flexible organization tools, but, on the other hand, third-party IMAP clients have difficulty dealing with labels. Guess what: with Fastmail, you can switch between folders and labels on the fly, at any time, as many times as you want! Just awesome.


Bottom Line


I’m glad I spent the better part of a year with Hey. It is indeed, as they tout, a fresh take on email. I’m sure there are many people who enjoy its quirkiness, and others that greatly benefit from its features. For me personally, though, Hey’s features were either easy to recreate with traditional email systems or not useful to me at all. Once I decided to switch away from Hey, Fastmail (which I had previously used as a backup email provider and a historic email repository) seemed an obvious choice. So far, my experience with Fastmail has been fantastic, and I wouldn’t hesitate to recommend it to any technically-inclined user.


Friday, January 20, 2023

Bootstraps

 

Preface

From Martin Luther King Jr.’s “The Other America” speech, Stanford University, April 14, 1967:

And [it’s one thing] … to say to people that you are to lift yourself by your own bootstraps, but it is [a cruel jest] to say to a bootless man that he ought to lift himself by his own bootstraps. And the fact is that millions of Negros, as a result of centuries of denial and neglect, have been left bootless. 


_____________________________________________________


The USA Today crossword puzzle is a treasured part of my daily routine. Primarily edited by the brilliant Erik Agard, it unfailingly provides a smooth, amusing solving experience, and, more often than not, I learn a few things along the way. Agard himself was the constructor of last Monday’s puzzle, as well as its editor, but that was not unusual per se. What was unusual was the grid itself. It was almost totally divided diagonally into halves:



Even more unusual (in fact, unprecedented, at least in my experience): the chasm between the solving difficulties of the halves. I had no problem with the northwest half of the puzzle, but as soon as I hit the southeast half, my solving suddenly came to a screeching halt. I was flummoxed. 

Merely attempting to solve the SE felt frustrating in the extreme. That part of the puzzle seemed beyond me, as if I was ill-equipped even to have a shot at it. It felt infuriating to be stumped by clue after clue, to not gain any toeholds anywhere, to simply spin my wheels in gut-wrenching futility. 

As it turns out, that was exactly what Agard intended.


_____________________________________________________


Published on Monday, January 16th, Martin Luther King day, Agard’s puzzle captured the essence of “The Other America,” but not by merely describing it. Instead, in a clever piece of symbolism, Agard gave the puzzle’s solvers just a taste of the impotence that many Black Americans feel when attempting to solve problems that, through no fault of their own, and not for lack of effort, intelligence or character, they simply cannot solve. The crossword’s halves represent Dr. King’s “two Americas:” the NW “overflowing with the miracle of prosperity and the honey of opportunity,” and the SE “an arena of blasted hopes and shattered dreams.” 

The southeast half of Agard’s puzzle is virtually unsolvable by design

Failing to solve any part of a crossword puzzle is usually frustrating, but failing to solve the southeast half of Erik Agard’s January 16, 2023 USA Today puzzle was enlightening. For, by doing so, I momentarily became Dr. King’s “bootless man,” fruitlessly attempting to lift myself by my nonexistent bootstraps. Agard brought “The Other America” to life, and thus singularly honored Dr. King’s legacy. 

The cruel jest was on us. Well done, Mr. Agard.


_____________________________________________________



Attribution note:

It was only after reading Sally Hoelscher’s comment on “Diary of a Crossword Fiend,” and her commentary that I fully understood the meaning of Agard’s gem and was moved to write this essay. Thank you Sally!


Monday, July 25, 2022

Boswords 2022 Summer Tournament


I can’t say that participating in the Boswords 2022 Summer Tournament was a humbling experience, because I came into it already sufficiently humbled. However, the Tournament did bring the blurred edges of my crossword puzzle-solving humility into razor-sharp focus.


Having heard about the contest from the inimitable Rex Parker, I forked over my $30 entry fee (an incredible value, by the way) and signed up as an online contestant. I assumed that, although my opponents would possess varied skill levels, they would generally coalesce around my own solving prowess neighborhood (I’ve been solving for approximately three years, with 2,200+ puzzles under my belt).


When I finished the first Tournament puzzle, which had a 20-minute time limit, in 11:16, I was quite pleased with myself. I finished the second, another 20-minute puzzle, in 12:49, so I figured I was on a roll. But then I took a look at the standings


My goodness, the names. It turns out that the Boswords tournaments attract some of the most brilliant luminaries in the crossword world as contestants. People like Will Nediger, Rebecca Goldstein, Hoang-Kim Vu, Jeff Davidson and Max Kurzman, just to name a handful. Hilariously, the morning of the Tournament, I had completed the Washington Post Sunday crossword, constructed by the amazing Evan Birnholz… and Birnholz himself was a contestant in the Tournament!


The standings, at that point, served to emphatically buttress my solving humility. I mentioned I was pleased with my 11:16 solving time for Puzzle #1? Well, Will Nediger solved the identical puzzle in... 1:54. (I’m pretty sure that even if I were given all the answers to the puzzle beforehand, I wouldn’t be able to simply type them into the grid in 1:54!) Hoang-Kim Vu solved Puzzle #1 in 3:14, Rebecca Goldstein did it in 3:16, and our friend from that morning’s Washington Post, Evan Birnholz, solved it in 3:18. Literally dozens of other solvers in the Tournament solved Puzzle #1 in less than five minutes. 


Humility appropriately buttressed, I gamely continued solving. Puzzle #3 proved too much for the man. I was unable to finish it in the 35 minutes allotted. Of course, the dastardly Nediger solved it in 3:04. Yes, 3:04, for a 117-word puzzle that, given a couple of hours, I only may have been able to solve.


I sort of redeemed myself with Puzzles 4 and 5, cleanly solving both within the 30 minutes allowed (29:14 and 28:09, respectively). Mr. Nediger? 2:27 and 3:40.


For the online solver championship, Nediger faced off against the aforementioned Davidson and Kurzman in a three-way, live, on- air solve of a wicked puzzle dreamed up by the brilliant Paolo Pasco. Pasco had come up with two sets of clues for his Championship grid; one labeled “mild,” the other “sizzling.” Using the “mild” set of clues, I was unable to finish Pasco’s puzzle in the allotted time (I believe it was 20 minutes). The three finalists, of course, were required to use the much more difficult “sizzling” clues for their on-air solve. They all solved the puzzle cleanly, all well under 5 minutes. Will Nediger won the Championship match, and the online tournament. (For the record, I ended up with a rank of 262 out of 314 in the overall standings. I write that sentence with a strange combination of pride and humility. )  


When I signed up for the Boswords 2022 Summer Tournament, I expected a certain experience. What I ended up with was completely different. I had the unexpected privilege of solving the same puzzles at the same time as a multitude of people whom I deeply admire. I was reminded that there are individuals in the world whose intellect surpasses mine to such an extent that it seems impossible that we are of the same species. And I got to spend an afternoon deeply immersed in the world of crossword puzzles, which, as the past three years have shown me, is a magical place, chock-full of intelligent, genuine, welcoming, generous, and inclusive people. It’s a place with nothing less than the best that humanity has to offer.


I’d like to extend a big thank you to John Lieb, Andrew Kingsley, Lisa Bunker, Ruth Margolin, Paolo Pasco, Erica Hsiung-Wojcik, Parker Higgins, Rob Gonsalves and Brad Wilber for making the Boswords 2022 Summer Tournament possible, and available to all.


I look forward to the next one, hopefully in person!


Sunday, May 16, 2021

Break

Since I rebooted Azout Askance in March of 2020, I’ve penned nineteen essays and produced fifty-one videos. A total of seventy posts in roughly 60 weeks, slightly more than one per week. Beginning on August 2nd, 2020, I settled into a weekly-on-Sunday cadence of videos. Last Sunday’s “The Default Human Condition,” was the forty-second consecutive weekly video.

As any baseball fan or (Douglas Adams fan, for that matter) knows, forty-two is quite a special number. So, now seems as good a time as any to take an indefinite break to recharge, refresh, and evaluate new ideas.  


As always, I deeply appreciate that you’ve taken the time to read and/or view my material, and look forward to connecting with you again in this space soon.


Sunday, May 9, 2021

Sunday, May 2, 2021

Saturday, April 24, 2021

Vulcan

Saturday, April 17, 2021

Sunday, April 11, 2021