We need to understand our own intelligence is competition for our artificial, not-quite intelligences.
A few days ago, Elon Musk likened artificial intelligence (AI) to “summoning the demon.” As I’m sure you know, there are many stories in which someone summons a demon. As Musk said, they rarely turn out well.
There’s no question that Musk is an astute student of technology. But his reaction is misplaced. There are certainly reasons for concern, but they’re not Musk’s.
The problem with AI right now is that its achievements are greatly over-hyped. That’s not to say those achievements aren’t real, but they don’t mean what people think they mean. Researchers in deep learning are happy if they can recognize human faces with 80% accuracy. (I’m skeptical about claims that deep learning systems can reach 97.5% accuracy; I suspect that the problem has been constrained some way that makes it much easier. For example, asking “is there a face in this picture?” or “where is the face in this picture?” is much different from asking “what is in this picture?”) That’s a hard problem, a really hard problem. But humans recognize faces with nearly 100% accuracy. For a deep learning system, that’s an almost inconceivable goal. And 100% accuracy is orders of magnitude harder than 80% accuracy, or even 97.5%. Read more…
Nate Oostendorp on manufacturing and the industrial Internet, and Tim O'Reilly and Rod Smith discuss emerging tech.
The Industrial Revolution had a profound effect on manufacturing — will the industrial Internet’s effect be as significant? In this podcast episode, Nate Oostendorp, co-founder and CTO of Sight Machine, says yes — where mechanization ruled the Industrial Revolution, data-driven automation will rule this next revolution:
“I think that when you think about manufacturing 20 years from now, the computer and the network is going to be much more fundamental. Your factories are going to look a lot more like data centers do, where there’s a much greater degree of automation that’s driven by the fact that you have good data feeds off of it. You have a lot of your administration of the factory that will be done remotely or in a back office. You don’t necessarily need to have engineers on a floor watching a machine in order to know what’s going on. I think fundamentally, the number of players in a factory will be much smaller. You’ll have much more technical expertise but a fewer number of people overall in a factory setting.”
According to Oostendorp, we’re already seeing the early effects today in an increased focus on quality and efficiency. Read more…
It's all about software, but it's a little harder than that.
If you Google “next industrial revolution,” you’ll find plenty of candidates: 3D printers, nanomaterials, robots, and a handful of new economic frameworks of varying exoticism. (The more generalized ones tend to sound a little more plausible than the more specific ones.)
The phrase came up several times at a track I chaired during our Strata + Hadoop World conference on big data. The talks I assembled focused on the industrial Internet — the merging of big machines and big data — and generally concluded that in the next industrial revolution, software will take on the catalytic role previously played by the water wheel, steam engine, and assembly line.
The industrial Internet is part of the new hardware movement, and, like the new hardware movement, it’s more about software than it is about hardware. Hardware has gotten easier to design, manufacture, and distribute, and it’s gotten more powerful and better connected, backed up with a big-data infrastructure that’s been under construction for a decade or so. Read more…
Predixion service could signal a trend for smaller health facilities.
Analytics are expensive and labor intensive; we need them to be routine and ubiquitous. I complained earlier this year that analytics are hard for health care providers to muster because there’s a shortage of analysts and because every data-driven decision takes huge expertise.
Currently, only major health care institutions such as Geisinger, the Mayo Clinic, and Kaiser Permanente incorporate analytics into day-to-day decisions. Research facilities employ analytics teams for clinical research, but perhaps not so much for day-to-day operations. Large health care providers can afford departments of analysts, but most facilities — including those forming accountable care organizations — cannot.
Imagine that you are running a large hospital and are awake nights worrying about the Medicare penalty for readmitting patients within 30 days of their discharge. Now imagine you have access to analytics that can identify about 40 measures that combine to predict a readmission, and a convenient interface is available to tell clinicians in a simple way which patients are most at risk of readmission. Better still, the interface suggests specific interventions to reduce readmissions risk: giving the patient a 30-day supply of medication, arranging transportation to rehab appointments, etc. Read more…