The Dressler Blog

When you build digital stuff all day, you develop opinions. Lots of opinions.


Digital Trends

What to do? One of the reason I dislike the use of the phrase “artificial intelligence” to refer to machine learning is that the vast majority of “AI” is the use of deep neural nets for pattern recognition. Meaning, that if I want to recognize a dog in a photo, or understand spoken language or look through a huge set of patient data for patterns of illness, machine learning is pretty awesome. However, while pattern recognition is a necessary precursor to general intelligence, it cannot be considered intelligence in total. That requires lots of other things, but one of those things is clearly the ability to make decisions. If my neural network can see the difference between a dog and a cat, it still doesn’t have the ability to act on that distinction. Now a startup out of Cambridge believes they have the technology to take this next step. is attempting to layer a decision making engine on top of machine learning using probabilistic modeling, reinforcement learning and game theory. According to the founders, such a technology would be capable of recognizing patterns, making decisions and then learning from the consequences of those decisions. While such claims should necessarily be taken with a grain of salt (absent a testable product or customers), was founded by some of the co-founders of VocalIQ, another machine learning company that was bought by Apple shortly after launch. They have also managed to line up an impressive group of financial backers and hire a highly impressive group of employees from academia. While they have yet to sign up any customers, the founders have strongly implied that they will be working with a gaming company in the near future. Why does this matter? Normally, I don’t write about products that might exist at some point in the future. There’s so much money and excitement around neural networks right now that founders have been known to engage in preemptive and unjustified hype in the hopes of attracting investors or purchasers. While these founders seem more focused on market niche than technology, that could reflect their recent conversations with VC’s rather than an inherent bias towards marketing. However, mentioning a gaming company leads me to believe they may actually have something. I don’t have a lot of faith in probabilistic modeling and game theory. They both have a bad track record in the real world (See: Long Term Capital Management.) However, they have a very good track record in one specific area: gambling. Meaning that could have a highly successful technology for, say, replacing blackjack dealers. That’s a subset of decision making that favors the highly quantifiable within a Gaussian distribution. It’s not nothing, but it’s still not intelligence. In a nutshell: Interesting technologies are being developed by layering different types of analysis over machine learning. Read More Facebook's 10 million missing millennials Kudos to Brian Wieser at Pivotal Research for catching the fact that Facebook claims advertisers can target 25 million more people in the United States than exist in the US Census. For example, in the coveted 18-34 year old demographic, Facebook claims you can use their platform to target 41 million people in that age group. But the US Census says there are only 31 million people in that age group in the country. Considering how many ad metrics Facebook has had to “correct” in the last year, this is more than a minor mistake. Pity poor Carolyn Everson, Facebook’s Vice President of Global Marketing Solutions who was sent out to make the meaningless point that this incorrect metric is different than their other previous incorrect metrics. According to her these reach numbers were “designed to estimate how many people in a given area are eligible to see an ad a business might run. They are not designed to match population or census estimates.” Meaning that the fact that they claim to reach more people than actually exist is not a bug, but a feature. Sigh. Why does this matter? Facebook, we need to talk. Everyone here loves you. We know you’re trying to get better. But you need to stop lying about your ad platform’s reach and effectiveness. It’s getting a little sad. Actually, I don’t even think you’re lying. I think you probably arrived at your numbers honestly. I think you probably broke down how many people in each region were likely to see a post and then added all those numbers together, not realizing that you defined your regions with some overlap. Or, worse, the fake Facebook profiles I’ve been writing about for two years actually do exist and you guys have a bot problem. Either way, you shouldn’t let yourself be caught out in such obvious ways. In a nutshell: Facebook thinks there are more Americans on Facebook than there are Americans. Read More Oh Facebook, you scamp, all is forgiven! In what is surely just an example of fortuitous timing, all news about Facebook's slightly exaggerated audience figures has been eclipsed by a story about Russian involvement in the 2016 election. Facebook has confirmed a Washington Post story that a "shadowy Russian company" was involved in purchasing thousands of dollars worth of Facebook advertisements, indirectly supporting Donald Trump. Facebook has confirmed that they have shared their findings with investigators looking into Russian involvement in the election. Why does this matter? Facebook wins public relations! Seriously, every time an embarrassing fact emerges about the platform, a story immediately comes out that wipes all memory of it away. This is a particularly egregious example – a story for which Facebook could have been the only source, the contents of which they must have known for months. One wonders what other interesting election stories Facebook is waiting to leak when they need to change the news cycle. I'm not even mad, that's amazing. In a nutshell: Facebook is Keyser Soze. Read More Robots, shmobots... There are layers of assumption underpinning the belief that robots are about to steal all our jobs. The first assumption is that the labor market has never undergone a radical transformation related to technology before. Yet we have undergone technological transformations before, from hunter-gatherer to agrarian and from agricultural to industrial. The lesson being that, while there are moments of profound suckiness for individuals who find their skill sets rendered suddenly obsolete, the vast majority of people thrive under the new system. Another assumption is that the robot-economy will be unique because it is aimed specifically at replacing workers and thus will necessarily create massive unemployment. But industrial manufacturing was aimed specifically at replacing workers. For that matter, so was the plow. Finally, there is the belief that a skills gap exists such that the technological needs of companies are not being met by the low-tech skills of the labor force. Andrew Weaver is an assistant professor in the School of Labor and Employment Relations at the University of Illinois at Urbana-Champaign. He has performed a number of studies to see if there is any truth to the belief in a technological skills gap in America and found there is not much basis for it. (Article link below.) Indeed the skills that companies found themselves having the most difficulty finding were not related to math or technology at all. Higher level reading and higher level writing were the skills in highest demand, particularly for IT help desks and clinical labs. While there is certainly nothing wrong with educating people to code or work with technology, the ability to communicate and understand is still the most valuable skill an employee can possess. Why does this matter? Non-millennials have been told the same scary story about their impending obsolescence for twenty years. Because I work with technology, I am frequently treated to apologies from clients and co-workers for their lack of technological sophistication. Yet most of these individuals possess more than enough technical skills to handle themselves professionally. And, when new technologies or platforms are introduced, they quickly get up to speed. Digital technology has the unusual ability to paralyze non-native users with self-doubt. That’s absurd. These technologies are, generally speaking, intuitive and designed for the non-technical user. Lots of people in technology engage in obscurantism and acronym-spouting. Don’t buy it. You’re not obsolete. You’re learning. In a nutshell: There really isn’t much of a STEM-based skills gap. Read More

Give us your email to sign up for our weekly Dressler Digital Trends. Stop trying to keep up and start getting ahead.