Last month, I wrote a blog entry about the TV show Battlestar Galactica. To update, I rapidly fell out of love with the show for many reasons and am struggling to just finish the first season.
The last thought-provoking episode I watched was entitled “33,” in which the hodgepodge human space fleet finds itself mercilessly pursued by the machine Cylon fleet, with no more than 33 minutes between attempted Cylon attacks. At the start of the episode, it is made clear that the pattern of furious engagement has been going on for five day straight, and the human crewmembers are fraying from lack of sleep, whereas the Cylons are completely unfazed.
Not needing to sleep is a real, crucial advantage that intelligent machines will have over humans in the future. For example, autonomous warplanes could fly missions around the clock, only stopping to reload their weapons, refuel, or get repaired. As a result, a fleet of autonomous warplanes could, at any given moment, put a greater fraction of its planes in the air than an identically sized fleet of human-piloted warplanes, since at any given moment, some fraction of the human pilots need to be sleeping. The readiness gap would become a chasm over the course of a sustained air campaign like the Battle of Britain as the effects of poor sleep and overwork added up on the human side.
Recently, two American warships–the USS Fitzgerald and the USS McCain–collided with merchant ships during patrols with fatal results, and while the investigations of both incidents are ongoing, it’s likely that overwork and lack of sleep were causal factors. Automated warships wouldn’t have had such problems.
Outside of the military context, not having to sleep will enormously advantage intelligent machines since it will let them get more work done each day, and hence to earn more money. Consider a future scenario where intelligent robots exist, but are no smarter than average humans. Even with that important aspect of the playing field leveled, the robots would be vastly superior workers and would take over most or all of the economy. Without the need to sleep, eat or do recreational stuff, they could just stay at work 24/7. That’s equivalent to 168 hours of work per week, which means one robot of average intelligence could replace four humans who work 40 hours per week.
And actually, it’s even worse than that once you factor in all the time a typical human worker spends at lunch, taking smoke breaks, resting, or doing nonsense during a 40 hour “work” week. A conservative estimate is that 50% of time is wasted in a typical American workplace. So one robot that stayed on-task could actually replace EIGHT humans.
And if the robot just stays at work all the time, that means it doesn’t need an apartment, house, or car, which means a massive cut in its personal expenses and way more disposable income. Aging, illness, and disability wouldn’t be things it would have to worry about, so it wouldn’t spend any money on health or dental insurance, nor would it lock up any of its pay in long-term retirement plans. The robot’s disposable income would be massive compared to a human making the same salary.
As a result, jobs and money would become concentrated in the hands of machines and business owners, and human workers would get shafted. And remember, I’m not even assuming that the robots will be smarter than humans–I’m pointing out that they could quickly muscle humans out of the job market by banking on their other inherent advantages, notably their ability to function without sleep. Even robots that are only as smart as dumb humans will be a major threat.
This little gem comes from the 1979 Biblical Doomsday “documentary” The Late, Great Planet Earth:
‘I am one of those scientists who finds it hard to see how the human race is to bring itself and bring the human enterprise much past the year 2000.’
That dire prediction was made by famed scientist Dr. George Wald, who was by all accounts a brilliant man who won a Nobel Prize for his work.
The phrase “much past” makes Wald’s dooms-date ambiguous, though I consider it a failed prediction at this point, since we’re 17 years into new century without civilization collapsing, and without any evidence it’s about to. To the contrary, since Wald’s quoted statement, we’ve managed to add three billion more humans to the planet while also sharply reducing global rates of malnourishment and absolute poverty. Across a wide variety of metrics, the human race has grown larger, healthier, richer, and less violent, and there are no signs the trends will abate anytime soon.
Making accurate future predictions is always fraught with uncertainty, but it becomes especially conjectural when people start making predictions about things outside of their areas of expertise. Wald’s mastery of biochemistry left him with no better a grasp of the human race’s trajectory than an average person, and his inclusion in this religious doomsday documentary is an example of the “Appeal to Authority” logical fallacy, in which a person’s credentials are erroneously substituted for reasoned and fact-based argumentation.
In my recent blog entry about Richard Branson, I pointed out that predictions should not be trusted if the person making them stands to tangibly benefit if other people believe them, and to that I’ll add that predictions should not be trusted if the person making them doesn’t have relevant expertise. Moreover, name-dropping and credential-dropping should never substitute for independently verifiable facts and transparent methodologies.
UPDATE: (8/28/2017) Coincidentally, I just came across the article, “A Nobel Doesn’t Make You an Expert: Lessons in Science and Spin.” The author (a former New York Times science editor) uses the example of James Watson, who won a Nobel Prize for co-discovering the structure of DNA, to show that the opinions and predictions of “experts” are often of little value when they pertain to subjects outside their areas of expertise. In 1998, Dr. Watson erroneously predicted that cancer would be cured within two years. The author also sets forth a few tips for evaluating predictions from “experts,” which partly overlap with my own and which I’ll summarize here:
Ensure that the person’s education and professional credentials are relevant. A useful measure of a scientist’s level of expertise is the quantity and quality of the peer-reviewed papers they have produced.
Be suspicious when experts have conflicts of interest that may bias their opinions and predictions.
Remember that experts whose theories fall far outside the scientific mainstream are usually (but not always) wrong.
Be very suspicious of scientists and other experts who feel aggrieved or persecuted by the mainstream of their professions. If an expert with an outlier theory also believes there is a conspiracy against him or her, it should raise a red flag in your mind.
I got my hands on several years’ worth of Consumer Reports Buying Guides, and thought it would be useful to compare the 2006 and 2016 editions to broadly examine how consumer technologies have changed and stayed the same. Incidentally, the 2006 Guide is interesting in its own right since it provides a snapshot of a moment in time when a number of transitional (and now largely forgotten) consumer technologies were in use.
Comparing the Guides at a gross level, I first note that the total number of product types listed in the respective Tables of Contents declined from 46 to 34 from 2006 to 2016. Some of these deletions were clearly just the results of editorial decisions (ex – mattresses), but some deletions owed to entire technologies going obsolete (ex – PDAs). Here’s a roundup of those in the second category:
DVD players (Totally obsolete format, and the audiovisual quality difference among the Blu-Ray player models that succeeded them is so negligible that maybe Consumer Reports realized it wasn’t worth measuring the differences anymore)
MP3 players (Arguably, there’s still a niche role for small, cheap, clip-on MP3 players for people to wear while exercising, but that’s it. Smartphones have replaced MP3 players in all other roles. The classic iPod was discontinued in 2013, and the iPod Nano and Shuffle were discontinued last month.)
Cell phones (AKA “dumb phones.” The price difference between a cheap smartphone and a 2006-era clamshell phone is so small and the capabilities difference so great that it makes no sense at all to buy the latter.)
Consumer Reports recommended the Samsung MM-A700 in 2006
Cordless phones
PDAs (Made obsolete by smartphones and tablets)
Scanners (Standalone image and analog film scanners. These were made obsolete by printer-scanner-copier combo machines and by the death of 35mm film cameras.)
Here’s a list of new product types added to the Table of Contents between 2006 and 2016, thanks to advances in technology and not editorial choices:
Smartphones
Sound bars
Steaming Media Players (ex – Roku box)
Tablets
As an aside, here are my predictions for new product types that will appear in the 2026 Consumer Reports Buying Guide:
4K Ultra HD players (Note: 8K players will also be commercially available, but they might not be popular enough to warrant a Consumer Reports review)
Virtual/Augmented Reality Glasses
All-in-one Personal Assistant AI systems (close to the technology shown in the movie Her)
Streaming Game Consoles (resurrection of the OnLive concept)–it’s also possible this capability could be standard on future Streaming Media Players
Single device (perhaps resembling a mirrorless camera) that merges camcorders, D-SLRs, and larger standalone digital cameras. This would fill the gap between smartphone cameras and professional-level cameras.
It’s also interesting to look at how technology has (not) changed within Consumer Reports product types conserved from 2006 to 2016:
Camcorders. Most of the 2006 models still used analog tapes or mini-DVDs, and transferring the recordings to computers or the internet was complicated and required intermediary steps and separate devices. The 2016 models all use flash memory sticks and can seamlessly transfer their footage to computers or internet platforms. The 2016 models appear to be significantly smaller as well.
For a brief time, there were camcorders that recorded footage onto internal DVD discs.
Digital cameras. Standalone digital cameras have gotten vastly better, but also less common thanks to the rise of smartphones with built-in cameras. The only reason to buy a standalone digital camera today is to take high-quality artistic photos, which few people have a real need to do. Coincidentally, I bought my first digital camera in 2006–a mid-priced Canon slightly smaller than a can of soda. Its photos, which I still have on my PC hard drive, still look completely sharp and are no worse than photos from today’s best smartphone cameras. Digital cameras are a type of technology that hit the point of being “good enough for all practical purposes” long ago, and picture quality has experienced very little meaningful improvement since. Improvements have happened to peripheral qualities of cameras, such as weight, size, and photo capacity. At some point, meaningful improvements in those dimensions of performance will top out as well.
TV sets. Reading about the profusion of different picture formats and TV designs in 2006 hits home what a transitional time it was for the technology: 420p format, plasma, digital tuners, CRTs, DLPs. Ahhh…brings back memories. Consumers have spoken in the intervening years, however, and 1080p LCD TVs are the standard. Not mentioned in Consumer Reports is the not-entirely-predictable rejection of 3D TVs over the last decade, and a revealed consumer preference for the largest possible TV screen at the lowest possible cost. It turns out people like to keep things simple. I also recall even the best 2006-era digital TVs having problems with motion judder, narrow frontal viewing angles, and problems displaying pure white and black colors (picking a TV model back then meant doing a ton of research and considering several complicated tradeoffs).
DLP TVs were not as thick or as heavy as older CRT TVs, but that wasn’t saying much. They briefly competed with flatscreen TVs based on LCD and plasma technology before being vanquished.
Dishwashers. The Ratings seem to indicate that dishwashers got slightly more energy efficient from 2006-16 (which isn’t surprising considering the DOE raised the energy standards during that period), but that’s it, and the monetized energy savings might be cancelled out by an increase in mean dishwasher prices. The machines haven’t gotten better at cleaning dirty dishes, their cleaning cycles haven’t gotten shorter, and they’re not quieter on average while operating.
Clothes washers. Same deal as dishwashers: Slight improvement in energy efficiency, but that’s about it.
Clothes dryers. Something strange has happened here. “Drying performance” and “Noise” don’t appear to have improved at all in ten years, but average prices have increased by 30 – 50%. I suspect this cost inflation is driven by induced demand for non-value-added features like digital controls, complex permutations of drying cycles that no one ever uses, and the bizarre consumer fetish for stainless steel appliances. Considering dishwashers, clothes washers, and dryers together, we’re reminded of how slowly technology typically improves when it isn’t subject to Moore’s Law.
Autos. This section comprises almost half of the page counts in both books, so I don’t have enough time to even attempt a comparison. That being said, I note that “Electric Cars/Plug-In Hybrids” are listed as a subcategory of vehicles in the 2016 Buying Guide, but not in the 2006 Buying Guide.
Billionaire Richard Branson in front of a scale model of his experimental space ship
From December 2013:
Sometime in 2014, entrepreneur Richard Branson and his two children aim to be on the first commercial flight of SpaceShip Two, Virgin Galatic’s rocket for propelling eight people 100 kilometers above the Earth. (SOURCE)
Sadly, SpaceShip Two broke up during a test flight in October 2014, killing one of its pilots. A replacement was constructed, and as of August 2017, it is undergoing sub-orbital test flights, but Branson and his children haven’t used it or any other craft to go into space (in fact, Virgin Galactic has only had three manned spaceflights in its history, all taking place in 2004). Yet hope springs eternal, and there’s a new deadline:
One area Branson has been less keen on speaking out on recently has been his project to take people into space. Virgin Galactic, as the fledgling business is known, has been beset by technical and other difficulties, not least the fatal crash of its SpaceShipTwo in California’s Mojave Desert in October 2014.
Despite the idea proving popular with future travellers – some 500 potential customers have spent $250,000 on reserving their spot on one of its trips– it is perhaps the one business he has found the hardest to get off the ground.
After the crash, Branson said his dream of space travel may have ended. But Galactic, under boss and former NASA chief of staff George Whitesides, has regrouped, redoubled its focus on safety, and appears to be making progress.
…“The test programme is going really well, and as long as we’ve got our brave test pilots pushing it to the limit we think that after whatever it is, 12 years of hard work, we’re nearly there.”
When exactly will he be nearly there? After all, Branson himself – and some of his family – have committed to being on the first flight.
“Well we stopped giving dates,” he confesses. “But I think I’d be very disappointed if we’re not into space with a test flight by the end of the year [2017] and I’m not into space myself next year [2018] and the progamme isn’t well underway by the end of next year.” (SOURCE)
This underscores the need to always be skeptical of future predictions, even if they come from people who have been enjoying a lot of recent success and who appear to know what they’re talking about. Skepticism is doubly warranted when the predictions are self-serving and possibly designed to boost interest and investment in the person’s business ventures (i.e. – inflate the stock price of the predictor’s company, of which he is the majority shareholder). On that note, I’m a fan of Elon Musk, but I fear he might be dangerously over-reliant on self-generated hype to keep his portfolio of businesses going. At some point, his investors will lose faith in him without bona fide profits.
I had to swing by the local pharmacy last weekend to get a prescription. There was no line, so the trip was mercifully short and efficient. But as usual I couldn’t help shake my head at the primitive, labor-intensive nature of the operation: Human beings work behind the counter, tediously putting pills into little orange bottles by hand. The pharmacist gets paid $121,500/yr to “supervise” this pill-pouring and to make sure patients aren’t given combinations of pills that can dangerously interact inside their bodies, even though computer programs that automatically detect such contraindications have existed for many years.
We have self-driving cars and stealth fighters, computing devices improve exponentially in many ways each year, and a billion people have been lifted out of poverty in the last 20 years. Pharmacies, on the other hand, don’t seem to have progressed since the 1980s.
For the life of me, I can’t understand the stagnation. Pharmacies seem ideally suited for automation, and I don’t see why they can’t be replaced with large gumball machines and other off-the-shelf technologies. Just envision the back wall of the pharmacy being covered in a grid of gumball machines, each containing a unique type of pill. Whenever the pharmacy received an order for a prescription, the gumball machine containing the right pills would automatically dispense them down a chute and into an empty prescription bottle. The number and type of pills in the bottle would be confirmed using a camera (the FDA requires all pills to have unique shapes, colors, and imprinted markings), a small scale, and some simple AI visual pattern recognition software to crunch the data. This whole process would be automated. Empty pill bottles would be stored in a detachable rotary clip or something (take inspiration from whatever machines Bayer uses to fill thousands of bottles of aspirin per day). Sticky paper labels would be printed as needed and mechanically attached to the pill bottles.
Every morning, a minimum-wage pharmacy technician would receive boxes of fresh pills from the UPS delivery man and then pour the right pills into the matching gumball machines. Everything would be clearly labeled, but to lower the odds of mistakes even further, the gumball machine globes would have internal cameras and weight scales to scan the pills that were inside of them and to verify the human tech hadn’t mixed things up. (And since the gumball machines would continuously monitor their contents, they’d be able to preemptively order new pills before the old ones ran out.) The pharmacy tech would spend the rest of the day handing pill bottles to customers, verifying customer identities by looking at their photo IDs (especially important for sales of narcotics), and swapping out rotary clips of empty pill bottles. If a customer were unwittingly buying a combination of medications that could harmfully interact inside their body, then the pharmacy computer system would flag the purchase and tell the pharmacy technician to deny them one or other type of pill.
I’m kind of content to stop there, as automating just those tasks would be a huge improvement over the current way of business (one human could do the work currently done by three), but here some more ideas:
Confirming a customer’s identity before giving them their pills could also be automated by installing a machine at the front counter that would have a front-facing camera and a slot for inserting a photo ID card. The machine would work like the U.S. Customs’ Automated Passport Control machines, and it would use facial recognition algorithms to compare the customer’s face with the face shot on their photo ID. I’ve used the APC machines during overseas trips and never had a problem.
The act of physically handing prescription bottles to customers could also be automated with glorified vending machine technology, or a conveyor belt, or a robot grabber arm.
Eighty percent of pharmacy customers are repeat buyers who are already in the computer system and are just picking up a fresh bottle of pills because the old bottle was exhausted. There’s no need for small talk, questions, or verbal information from the pharmacist about this prescription they’ve been taking for months or years. That being true, the level of automation I’ve described would leave pharmacists with a lot of time to twiddle their thumbs during the intervals between the other 20% of customers who need special help (e.g. – first-time customer and not in the patient database, have questions about medications or side effects). Having a pharmacist inside every pharmacy would no longer be financially justified, and instead each pharmacy could install telepresence kiosks (i.e. – a station with a TV, sound speakers, a front-facing camera, and a microphone) through which customers could talk to pharmacists at remote locations. With this technology, one pharmacist could manage multiple pharmacies and keep themselves busy.
An Automated Passport Control machine in use
As far as I can tell, the only recent advances in the pharmacy/pill selling business model have been 1) the sale of prescriptions through the mail and 2) the ability to order refills via phone or Internet. If you choose to physically go into a pharmacy, the experience is the same as it was when I was a kid.
Is there a good reason it has to be the way it is now? I suspect the current business model persists thanks to:
Political lobbying from pharmacists who want to protect their own jobs and salaries from automation (see “The Logic of Collective Action”).
Unfounded fears among laypeople and politicians that automated pharmacies would make mistakes and kill granny by giving her the wrong pills. The best counterargument is to point out that pharmacies staffed by humans also routinely make those same errors. Pharmacists will also probably chime in here to make some vague claim that it’s more safe for them to interact with customers than to just have a pharmacy tech or robot arm hand them the pills at the counter.
Fears that automated pharmacies will provide worse customer service. Again, 80% of the time, there’s no need for human interaction since the customer is just refilling a prescription they’ve been using for a long time, so “customer service” doesn’t enter into the equation. It’s entirely plausible that a pharmacist could satisfy the remaining 20% of customer needs through telepresence just as well as he or she would on-site.
High up-front costs of pharmacy machines. OK, I have no experience building pharmacy robots, but my own observations about the state of technology (including simple tech like gumball machines) convince me that there’s no reason these machines should be more expensive than paying for human labor. Even if we assume that each gumball machine costs an exorbitant $1,000, you could still buy 121 of them for the same amount a typical pharmacist would make in a year, and each gumball machine would last for years before breaking. It’s possible that pharmacy machines are unaffordable right now thanks to patents, which is a problem time will soon solve.
Well, at the end of my pharmacy visit, I decided to just ask the pharmacist about this. She said she knew of pharmacy robots, and thought they were put to good use in hospital pharmacies and military bases, but they weren’t suited to small retail pharmacies like hers because they were too expensive and took up too much physical space. I would have talked to her longer, but there was a long line of impatient people behind me waiting to be handed their pill bottles.
