The Robots Are Here and They Can Spell

[Image courtesy of World of Weird Things.]

I warned you, fellow puzzlers. You can’t say I didn’t warn you.

The robots are coming, and they want our puzzles and games.

Let’s look at the hit list:

  • Deep Blue defeated Russian chess grandmaster Garry Kasparov under standard chess tournament time constraints
  • IBM’s supercomputer Watson bested previous Jeopardy! champions Brad Rutter and Ken Jennings to nab a million-dollar prize
  • An AI program called DeepMind taught itself to play several Atari games with superhuman proficiency
  • There are several robots constructed out of LEGOs that solve Rubik’s Cubes in seconds flat
  • Dr. Fill, the crossword-solving computer program, competes at the American Crossword Puzzle Tournament, and in a matter of five years, it has jumped from 141st place in the 2012 tournament to 11th place in the 2017 tournament
  • Just last year, an AI developed by Google, AlphaGo (a product of DeepMind), twice defeated Ke Jie, the 19-year-old Go tournament champion ranked number one in the world

And Scrabble fans, you’re the next ones in the crosshairs of the machines.

During last week’s Consumer Electronics Show (CES), the Industrial Technology Research Institute out of Taiwan debuted the IVS Robot — aka The Intelligent Vision System for Companion Robots — a machine capable of defeating human competitors at Scrabble.

[Image courtesy of ABC News.]

Instead of tiles and a standard Scrabble board, the IVS reads letter cubes (similar to a child’s alphabet blocks) played on a slightly larger gameboard. But time limits for play and standard rules still apply.

From an article on Engadget:

It’s hard not to be impressed by all the moving parts here. For one, the robot has to learn and understand the rules of the game and the best strategies for winning. It also needs to be able to see and recognize the game pieces and the spots on the board. That means it can read the letters on the cubes and identify the double-letter and triple-word score spots.

And, last but not least, it needs the dexterity to place the pieces on the board and not disturb the existing letters — which is especially difficult when you’re laying down two words next to each other to rack up those two-letter combos.

A quick Google search confirms that the robot bested practically every reporter, tech-savvy or otherwise, that crossed its path.

In the video below, North American Scrabble champion Will Anderson teams up with reporter Lexy Savvides to battle the robot, but a technical error prevents the game from getting very far:

Still, you can see the potential here. I’m sure it won’t be long before the IVS Robot is making appearances at Scrabble tournaments, attempting to establish machine dominance over another puzzly activity.

Stay strong, fellow puzzlers.

Thanks for visiting PuzzleNation Blog today! Be sure to sign up for our newsletter to stay up-to-date on everything PuzzleNation!

You can also share your pictures with us on Instagram, friend us on Facebook, check us out on TwitterPinterest, and Tumblr, and explore the always-expanding library of PuzzleNation apps and games on our website!

How to Make a Crossword: Cluing Advice

After constructing the grid, cluing is the most daunting task facing a constructor. Constructor Robin Stears readily confessed, “Writing the clues is the most time-consuming process of puzzle construction, especially for common words that appear in many puzzles. Crossword Compiler allows me to keep a database of words and clues, but I try not to use the same clue twice.”

Constructor and puzzle archivist David Steinberg: “When writing clues, it’s important to strike a balance between original clues and clues that exist in databases. For entries that appear frequently and/or have a limited number of cluing possibilities, such as ALAI (traditionally clued as [Jai ___]), I feel it’s best to go with a database clue. In the case of ALAI, almost all clever cluing possibilities, such as [Half-court game?], have been exhausted.

“Original clues for such an entry often end up feeling strained or wordy and/or rely on a less common usage of a word, which solvers generally don’t appreciate as much.”

Constructor Ian Livengood also stressed finding a balance between creativity and accessibility: “Keep you clues relatively short, especially if you’re creating puzzles for outlets with strict line counts. But don’t just use one-line clues for everything, since that will bore solvers. Try to toss in some fun trivia, wordplay, etc. that seems interesting to solvers.

“And, like filling the grid, make sure you clues are appropriate for the intended day of the week. [High line?] for ELEVATED TRAIN works well for a tough puzzle, but would only fluster new solvers in a Monday puzzle. [Above-the-street transportation] is easier and more welcome for beginners.”

Constructor Robin Stears reminds you to utilize the many resources available: “Personally, I use a number of websites to help me write clues: Wikipedia, Wolfram Alpha, The Free Dictionary, Crossword Tracker, Internet Movie Database, Reddit, and Google. In the old days, I used to spend hours in the reference section of the library, but nowadays, it’s much easier to search the Internet. There’s a wikia for almost everything.”

When it comes to determining the difficulty of a clue, constructor Doug Peterson suggests doing a bit of research: “My best advice is to solve lots and lots of puzzles of varying levels of difficulty. It won’t take long for you to get a feel for what types of clues are found in what types of puzzles. And it’s OK to have a few hard clues on a Monday or a Tuesday. Just make sure their answers don’t cross.”

Los Angeles Times Crossword Editor Rich Norris and assistant Patti Varol touched on how cluing can set a puzzle’s difficulty: “With the right grid, a talented, creative editor can transform an expert-level puzzle into one that any newbie can solve (and that an expert would still enjoy). Every editor has his or her own ‘familiarity test,’ which is the educated guess we make to determine if a clue or an entry will resonate with or be recognized by our solvers.

“It’s subjective, sure, but there are also pretty straightforward guidelines. If an entry is not in most major dictionaries and only gets 10K or so Google hits, well, that’s pretty obscure and probably shouldn’t be used at all. The editor needs to know the solving audience and needs to know how to balance current and older pop culture references -– much of what the Millennials find easy will completely baffle the Boomer solvers, and vice versa.”

They were also kind enough to offer an example of how cluing a given word can affect clue difficulty:

“Take the entry SMITH, for example. Will SMITH and Ozzie SMITH are arguably more famous than Patti SMITH, who is arguably more famous than Matt SMITH. Plain clues — [Actor Will], [Baseballer Ozzie], [Singer Patti], [Actor Matt] – tend to be hard. Ozzie is the exception in this example, because Ozzie is such an unusual name in baseball that even nonfans are likely to word-associate their way to the correct answer.

“What information is added to those semi-naked clues is key. Thus, [“Men In Black“ actor Will] is an easy clue for SMITH because “Men In Black” is a popular movie franchise and having the first name narrows down the potential answers. Compare [“Just Kids” memoirist Patti]. Even with that first name in the clue, it’s a tough clue — more solvers know Patti Smith as a singer than as a writer, and the title of her memoir doesn’t even hint at her singing career. For a nonfan, [Hall of famer shortstop Ozzie] is about the same difficulty as [Baseballer Ozzie], but, alas, it’s likely only a Doctor Who fan will recognize any clue for Matt SMITH.

“There’s Kate SMITH, and Bessie SMITH, and Agent SMITH of the Matrix movies. Each one resonates with a different solver. The easiest kind of clue for SMITH is, of course, the generic kind: [Common alias] or [Popular surname] or [Suffix with lock or gun]. And then there’s the fun, inferable kind: [Surname that comes from an occupational suffix].”

Teasing your brain (and your job applicants)

Did you ever have a job interview where someone posed a mental test or brain teaser?

These were all the rage a few years ago and I’ve heard plenty of stories from friends and acquaintances who applied for jobs only to find themselves wandering down tangential rabbit holes instead of presenting their credentials in the best light.

How many golf balls can fit in a school bus? How much should you charge to wash all the windows in Seattle? Why are manhole covers round? How many piano tuners are there in the entire world?

Google (among other companies) became notorious for this sort of on-the-spot cognitive analysis, but in an interview with The New York Times, senior vice president of people operations at Google Laszlo Bock admitted that these kinds of questions proved completely worthless as predictors of employee creativity or performance.

As a puzzle guy, I can appreciate the spirit behind asking these questions. When you present a seemingly unsolvable puzzle, you’re not really looking for the solution, you’re looking for the resourcefulness of the solver. When you present a brain teaser that demands great results with only two tries, you’re examining the interview’s insightfulness and efficiency.

The problem is… abstract problem-solving isn’t the same as actual problem-solving. I daresay the interview is the most stressful part of many jobs, so the pressure you endure sitting in the hot seat and trying to earn a job overshadows the pressure you’ll endure actually doing that job. After all, there’s not a yes-or-no implied after each question when you’ve got the job, but that uncertainty permeates the interview process.

But, Bock also explains how properly-framed questions about problem-solving can be more useful indicators:

Behavioral interviewing also works — where you’re not giving someone a hypothetical, but you’re starting with a question like, “Give me an example of a time when you solved an analytically difficult problem.” The interesting thing about the behavioral interview is that when you ask somebody to speak to their own experience, and you drill into that, you get two kinds of information. One is you get to see how they actually interacted in a real-world situation, and the valuable “meta” information you get about the candidate is a sense of what they consider to be difficult.

I’ve never had to answer a brain teaser like the ones listed above, not even when I interviewed to be a puzzle guy. Of course, if you ask me how many golf balls can fit in a school bus, my first answer would probably be “more than I could ever need.” Plus I don’t do windows.

[Check out this io9 article for greater detail, including source links and list of former Google interview questions.]

Thanks for visiting the PuzzleNation blog today! You can like us on Facebook, follow us on Twitter, check out our games at, or contact us here at the blog!