Some weeks ago a friend asked me how to automate and achieve scale for his mental health startup. It’s not been the first time someone has asked me this. Scale and automation are, under the technology creed, synonyms for Artificial Intelligence.
I was hesitant to answer. I told him that I am a believer on AI, but that you couldn’t apply AI as it stands today to mental health problems.
Most AI methods are based on function optimization. In layman’s terms, the algorithm looks for the best (optimal) solution to a given goal. The problem with this is that such goal rarely contains information about its moral worth.
“A system that is optimizing a function of n variables, where the objective depends on a subset of size k < n, will often set the remaining unconstrained variables to extreme values; if one of those unconstrained variables is actually something we care about, the solution found may be highly undesirable.”
And this is the key to the disturbing stage of technological evolution we are experiencing. As I’ve pointed out before, I believe we’re living a major moral crisis. One of the consequences is that we’re blindly applying mathematical formulas that don’t factor in human moral values. The goal is to automate a problem and do it fast an efficiently. The issue is that the human mind can’t be reduced to a set of equations (yet).
Human psychology is incredibly complex. We’re governed by dynamic systems that defy most rational explanations. During the last few decades, we’ve witnessed the failure of the Keynesian doctrine in economics. Heralded by behavioral scientists like Dan Ariely, more voices are demanding better models. Models that account for human irrationality.
However, we’re, once more, making the same mistake. We’re applying sophisticated algorithms that overly simplify the reality they deal with. One could argue that most AI systems are employed to non-human, repetitive, tedious tasks. And they’re right. Still, many of those AI systems are making decisions that do affect humans and their mental models.
“Yet despite all the thoughtful ethical guidance and research that’s already been produced, and is out there for the reading, here we are again being shown the same tired tech industry playbook applauding engineering capabilities in a shiny bubble, stripped of human context and societal consideration, and dangled in front of an uncritical audience to see how loud they’ll cheer.”
It’s discouraging to watch history repeated. In 1962, Rachel L. Carson published one of the most impactful books in science, Silent Spring. In it she criticised the indiscriminate use of chemicals (DTTs) and the poisonous effects of these in the ecosystem, humans included.
“Technology, she feared, was moving on a faster trajectory than mankind’s sense of moral responsibility.”
Linda Lear’s introduction to Silent Spring by Rachel L. Carson. 2002
At the time, the US Chemistry industry, one of the largest beneficiaries of the Cold War, operated at large. Science and chemists were considered the top of the food chain. No one questioned their knowledge. No one questioned their products. What Carson uncovered, documented and publicised was the other truth; lethal ignorance, greed, and Capitalism.
“Carson questioned the moral right of government to leave its citizens unprotected from substances they could neither physically avoid nor publicly question.”
Linda Lear’s introduction to Silent Spring by Rachel L. Carson. 2002
I can’t but draw parallels with our current situation. I question the moral right of Google or Facebook to leave their users unprotected. But this isn’t a problem with the prominent technology corporations, but with most AI-powered solutions. Not even with such solutions, but with the lack of awareness of the developers and designers.
All this said I’m bullish on AI. I’ve been a defendant and ardent believer of the field. This is why I’m so vocal about the current misdirection.
Yes, we need autonomous agents. We need to apply AI, but we need to incorporate moral values into the equation. This in itself is a considerable challenge. It’s becoming one of the newest research lines in AI, but most advances are, so far, theoretical. New companies deploying new systems should be aware of the problems. They should try to apply new AI models and build moral safeguards.
“The systems will need some method for learning and adopting prosocial preferences, in light of the fact that we cannot expect arbitrary rational actors to exhibit prosocial behavior in the face of large power disparities.”
Podcasting has been around for ages, but it hasn’t been until recently that it’s making a comeback. For years, it was hard, not only to find good content but to subscribe to it. On top of that, the only place you could listen to podcasts was on your computer.
Apple, smartphones and bandwidth improvements changed that. The simplicity of subscribing to a show on iTunes and listen to it on the Apple Podcast app made it a breeze. It opens the gates for mainstream adoption of podcasts.
Two significant milestones propelled the awareness of podcasts. The first one was the addition of official support for podcasts on iTunes in June 2005. The second milestone was the release of the iPhone on June 2007.
Despite Serial’s success, podcast consumption remained low compared to other siblings like video or social media. Audio remains a much-loved but not-mainstream medium.
One of the major problems around podcasting has always been monetization. One the one hand, audiences have been small compared to other content sources. On the other, there’s always been a difficulty with measuring the exact engagement of the audience. You can measure how many downloads but not the precise behavior of listeners. And that’s a problem for attracting money.
Some podcasters turned to their fans for monetization. They joined Patreon and transformed their audience into patrons that kept the lights running. However, not every podcast could pull that feat. Some creators believed brands should sponsor the content, instead of fans. Still, having no metrics was a significant stop-block.
Since then, the space is undergoing drastic changes. Data is already showing what many podcast creators have been claiming for years, that their audiences are hyper-engaged. They might not command Facebook-size audiences, but they sure have very targetted and bespoken listeners.
“On average, according to Midroll’s data, podcast listeners are making it through about 90 percent of a given episode, and relatively few are skipping through ads.” […] “Those numbers tend to be steady regardless of the length of the show—and according to Panoply, the few listeners who do skip ads continue to remain engaged with the episode, rather than dropping off at the first sign of an interruption.”
As the Wired article mentions, this is the advertising Holy Grail. This has become even more important with Facebook’s recent problems. Not only ads are becoming politically charge on those platforms, but companies that advertise there are also feeling the heat. Also, many advertisers are looking for ways to break out of the advertising duopoly held by Google and Facebook.
It’s not surprising though that brands are looking at podcasts as a new frontier for their advertising dollars. One not controlled by Google or Facebook, but new players. Podcast advertising grew a 228% between 2015 and 2016. That’s not even taking into account 2017 which, as I mentioned, was a stellar year for the industry.
The impact of voice interfaces
Looming at a distance, voice interfaces like Amazon Alexa or Google’s Home Now, are heralding the audio renaissance. Many expect the increasingly ubiquitous smart speakers to change the dynamics.
As I’ve written before, I believe conversational interfaces will change how we consume many things. I’ve felt the change happening at home. How voice interfaces are pushing me to seek more integrated appliances. That said, I also reckon it still needs more time. It’s not surprising then how few users listen to podcasts with their smart speakers.
“David Markowitz from ListenUp thinks Interactive audio is still very much in the experimental phase. “I don’t think we’ll see broad adoption until we get more comfortable having lengthy interactions with smart speakers. That is going to happen – we are just not there yet.”
I have to agree with Markowitz. I’ve tried several podcast skills in Alexa, and it’s still not there. Some experiments are exciting and we’ll see more engagement but still needs more time.
Audiobooks to the rescue
I have no doubts that smart speakers will drive podcast consumption even higher. I would expect the newer generations to be used to consume audio as naturally as our parents did radio.
Another trend that will accelerate both smart speakers and podcasts, in general, are audiobooks. Audible (and Amazon’s) success in these past years is opening an entirely new media segment.
People that would never read a book are, surprisingly, willing to listen to one. There is no correlation between the quality of the book and its audio counterpart.
“One of the things that’s most interesting to me here is the fact that it used to be that the success of an audiobook was correlated with the success of the print book. That is no longer true. The number of audiobooks that perform well independent of its print and eBook circulation is increasing. The format itself is creating new ways of discovering content that is becoming increasingly independent of the underlying print and eBook success.”
It’s remarkable how some writers are even willing to create audio narratives to accompany their writings. In a way, it reminds of a comeback to ‘Serial’. Podcasts and audiobooks will begin to fuse and mingle. Some podcasts will remain radio shows, others will become new narrative portals and will drive audio consumption.
”Margaret Atwood, author of The Handmaid’s Tale, recently worked with Audible to write and record a spoken-word coda to the novel, and comedian David Spade is developing an audio-only memoir. “We’re really trying to break the boundaries,” Blum says, “and go to writers and creators and artists to think outside the traditional boundaries of what is a book.”
It’s not surprising that NPR et al. bought Pocket Casts. There is an impending need to create the ultimate podcast platform. The key here is the word platform, or more specifically, aggregator. As Ben Thompson argued recently, when the market is already modularized, it’s ripe for the rise of an aggregator.
I wonder who will that be. Apple still holds a massive grip on the podcast industry, but they’re slowly losing it. Both regarding client market share and on the follower side. New podcast distribution networks like Midroll, Art19 or Megaphone (Panoply) are carving a space not touched by Apple.
At the same time, Amazon and Google are pushing hard on the audiobook and smart speakers front. Will they control the space by owning the relationship with the end user like Apple had with the iPhone? They have an advantage, one they aren’t exploiting so far.
It’s a fascinating time to dwell in this space. I expect more content producers and creative people to adventure into the audio world during the next few years.
I am a big fan of logistics. Moving things from here to there is the dream of any engineer. We can track my enthusiasm to my time spent playing SimCity and The Incredible Machine games. I will throw in some Lemmings in the mix too.
Despite my keen interest in the field, we must confess it’s perceived as a dull industry. And no wonder. Run by old school boys driving trucks and loading freights. Grease, dirt, heavy cargo and men in flannel shirts come to mind.
However, the logistic dream from the 60s has nothing to do with logistics in 2018. The convergence of e-commerce, mobile and artificial intelligence is sending shockwaves across the industry.
While international logistics are essential, their complexity pales to the challenges of last-mile delivery. And to be honest, the current last-mile companies suck big time. I’m tired of getting my food cold. I’ve tried them in many different cities and countries, and it’s a hit and miss. I mention food delivery, but the same goes for package delivery.
Many companies, stirred by the looming Amazon empire, are pushing into the field. To do last-mile right, you need massive capillarity and low response times. On top of building predictive systems for improved routing, which is worth noting, not everyone is doing; there is the simple need of adding more couriers. Here is the kickback, they have to be cheap. All hail to the sharing economy.
Not only last-mile delivery requires speed and capillarity, but low marginal costs. The problem is that humans are slow, expensive and above all, they despise this kind of work. People aren’t cogs, but we insist on using them as machinery.
Enter the robot utopia. What if we could substitute these tasks with robots, or more precisely, autonomous delivery vehicles. Devices that can travel by air or ground, without human intervention, delivering contents virtually anywhere within record time.
That’s the dream a set of companies has been pursuing during these past five years. The problem? The over-sold robotic promise of better-than-humans. The result is a backslash of under-performing companies with lame devices that produce mix-bag results.
To avoid undesired flashy PR, many of these companies have been in stealth mode for a while and have only unearthed their products recently.
There are two big approaches in the field, air delivery with drones or ground distribution with small autonomous delivery rover-like vehicles. Which is better is up for debate.
Autonomous Air Delivery Vehicles
What better way to do last-mile delivery than via high-precision air delivery? It’s fast, allowing straightforward delivery routes and road traffic avoidance.
Within this group, we find companies like Flirtey, Matternet, Zipline or the omnipresent Amazon. These systems work, and they’re impressive to watch.
Despite how remarkable they are, the aerial approach is riddled with problems. While it’s faster than ground approximations, technical specifications are complicated, both in-flight and on delivery. The major hurdle though isn’t technical but regulatory. Not only flight regulations are strict, but craft safety and air traffic control are critical.
Drone delivery has a place and use, that said, I don’t believe it will be massive. The reason is simple, e-commerce package traffic (or food for that matter) is too big for last-mile air delivery. The scale needed to support an even small percentage of that traffic would collapse any city’s airspace.
“But tools have not yet been developed to predict how weather will affect small drones flying around obstacles such as buildings or hills at such low altitude, Gitlin said.”
Taking Amazon Prime as an example, we’re talking about five billion packages shipped in a year; 416 million a month. If you factor in the seasonality of shipping, you have north of 900+ million deliveries at any given December. That’s only Amazon. Factor in new players like Walmart and the impending growth of urban areas. Air delivery requirements (collision avoidance, weather patterns, and safety space to name a few) will make it very hard to scale beyond a certain threshold.
As I mentioned before, Drone Delivery is instrumental but the cornerstone industries won’t be ecommerce or food delivery per se. Companies will continue to pursue it, but it won’t amount to much at first. The real disruption will come from shippings where speed is critical as in life-threatening, for example, medical deliveries. From blood samples to vaccines to organs.
Drones will gain traction outside mega cities too. Agrotech is already making use of drones, and I’m sure they’ll expand their usage of delivery services too. Industrial operations will also employ air delivery for spare parts, maintenance operations and even safe product transportation (i.e., Diamond delivery through hostile territories).
Autonomous Ground Delivery Vehicles
The other big group is ground-delivery autonomous vehicles. Within this side, we have startups like Starship, Marble, Teleretail, Dispatch or Robbie.
Ground robots sacrifice speed for easier regulation and less technical complexity. These devices look like small Mars rovers and are designed to roam the walkways at human speeds. They are fully autonomous, even though they do have human backup operators in case of emergency. They range from small carts to small burrito-like stands with wheels.
Approval for this kind of vehicles has been straightforward. There are already five US states that allow them on their walkways as well as 40 European cities with small-scale pilots.
These robots are simpler to deploy, and that’s the reason why they’re already operational in some locations. As with drones, the concern is around how to scale their number. For now, regulation limits their speed (6 km/h) and weight (18kg – 36kg unloaded) to human standards to prevent dangerous collisions. These restrictions severely cap their performance. Some of these vehicles can reach speeds of up to 50km/h and achieve bigger sizes. Faster and bigger rovers would enable a much more prominent payload in less time, improving the last-mile performance goal.
Many in the industry think these are initial conditions and as humans become used to the vehicles, they’ll gradually tolerate higher limits.
While accidents will occur, it’s not fair to compare bikes or motorbikes with autonomous delivery vehicles. If done right, LIDAR technology should provide better sensors than anything human. This should enable brake and slowdown of the rovers in the presence of humans, making it safe to operate at higher speeds without incidents.
Another concern is walkway traffic. If last-mile companies start deploying these vehicles en masse, it could cause severe problems for pedestrians.
“If there really were hundreds of little robots,” Ehrenfeucht said, “they would stop functioning as sidewalks and start functioning more as bike lanes. They would stop being spaces that are available for playing games or sitting down.”
Ehrenfeucht pointed out that 130 years ago, streets were not yet divided into lanes for traffic, parked cars, pedestrians and bikes, and that the introduction of robots to the streetscape might require a reimagining of the available space, possibly with a designated lane for robots.
I have to agree with Ehrenfeucht’s vision. I think that the advent of autonomous vehicles will reduce private car ownership and traffic. Hence we will be able to devote new lanes to autonomous ground vehicles. It will take years though. In the meanwhile, we’ll probably share the walkway with these little rovers.
It’s worth mentioning some other approaches like Nuro’s one. Cofounded by two former Googlers, they’re taking a radical approach. They’ve created an autonomous delivery vehicle, like Waymo’s, from the ground up. Jiajun “JZ” Zhu was one of the first employees at Waymo, so no surprise here. They’ve raised a monstrous 92 million dollar round for the company to put these babies on the road soon.
While I think the approach has merits, especially the heightened specialization of autonomous vehicles, they have a long way to go. Being a specialized vehicle, they’ll probably have less regulatory hurdles than Waymo, but they need to prove that a street-parked approach works better than a walkway one. That said, I’m sure they’ll inspire a new crop of vehicles designed for one use and one use only. It’s the modularization of the point of integration. The problem is, they might be too early for the modularization phase. Right now, vertical integration will win the game.
Disruptive effects of autonomous delivery vehicles
One of the most interesting effects is the clear disruptive nature of these vehicles. Most companies are aiming for e-commerce or food and grocery deliveries. That’s, in my opinion, the wrong disruptive approach.
There is entrenched competition in that space. There are myriad logistic companies that are pushing incremental innovations in last-mile deliveries. For autonomous delivery to survive, they need to focus on underserved or non-existent market footholds.
Starship Technologies is a fascinating case. They’ve gone from regular package delivery trials to focus on university and corporate campus deliveries. The move is genius. They’re serving a non-existent market, which welcomes the capacity to deliver goods (let it be food or packages) between buildings. Because there are no competitors, the users have no expectations. This works for Starship, who can develop their technology without performance comparisons. In such scenario, limiting regulations won’t matter much, as having rovers is already a welcome addition.
The critical part here is that autonomous delivery vehicle technology is a disruptive one. The pace of innovation is dramatic. Improvements in sensors, batteries, space allocation and route prediction will make to the market in no time.
“Mr. Blown’s trial showed consumers were open to this new service, but it also revealed some limitations of the robots: they were confined to a three-mile area that the machines had to map out beforehand. […] “At the moment one of our couriers can deliver 50 parcels in one go. With one robot delivering a package in about half an hour, you would need a lot of robots.”
The Hermes case is a beautiful example of why this technology will disrupt the logistics industry. So far, limitations on autonomy, mapping and data necessity, speed and cargo caps, will make them underperform. Assuming that this will be the case in a couple of years is a mistake. One that will wipe many couriers. Disruptive technology advances will make it progress way faster than many can predict. Once these technologies make it out of their niches (upscale march), they’ll start competing with traditional couriers, and they’ll win.
The space is wide open for a plethora of supporting services. 3D Sensing mapping companies will become increasingly important. New delivery companies don’t want to re-map an area but leverage preexistent LIDAR-generated 3D maps. There are several startups already working this approach, most only for roads, but the competition is still wide open for walkways and aerial maps.
Another space that will need some services are geofencing services for municipalities. Cities need a way to register autonomous delivery vehicles, to track them and to create no-fly-no-drive exclusion areas. The most advanced company in this space is AirMap, which does this for the drone industry. Nothing like it exists for ground delivery fleets. That piece will be pivotal to the mainstream adoption of autonomous delivery rovers.
Due to the autonomous nature of this vehicles, they’ll need to achieve a certain number of autonomous driven miles before they’re allowed to operate. Simulation software, like the one employed by Waymo, will be critical. Scenarios are different from autonomous cars. Walkways are very heterogeneous and pedestrians too. There will be a need to engineer synthetic scenarios permutations. These alternative scenarios will allow for better testing of the autonomous brain.
Many of these companies though, need to rethink their customers. Going for the sexy e-commerce pie might be a big mistake at this stage. New delivery methods allow them to serve inexistent demand. Becoming adept in these areas will allow them to expand horizontally and attack incumbents from a much more robust position.
It is an exciting new space, whose time is coming faster than many might think. Regulations are in place. Testing pilots are maturing, and scale is being achieved. I wouldn’t be surprised to see some of these vehicles roaming our streets in two to three years. It’s a brave new world, one where humans might be pushed aside if we aren’t careful.