Posted by We may never have self-driving vehicles. Or we may; time will tell. It is of interest that although accident rates are already below that of the driving public, a death by machine is far more objectionable than one caused by an individual’s own drunkenness, for example. That’s a curious phenomenon but one with which the automobile industry must grapple.
So an entire idea and potentially time-saving technology is written off because of the aberrant idea that machines would then take over our lives and cause our death.
Now, we will have electric vehicles; and already do. I just bought a new Toyota Prius Prime. Every day it goes about 40 kilometers – 25 miles – on a daily charge. And though I travel over 150 km (94 miles) a day, having 25% of that be gas-free miles is very appealing to me, especially as I approach retirement and decreased commute distances – yes, a spring chicken I am not.
But listen – that same Prius Prime has features birthed by the self-driving experiments. The leaving-the-lane warning and the radar-assisted cruise control are just as important to me as the distance I can go on a given amount of petrol fuel. The self-driving experimental work has spawned very useful and workable benefit even if its final goal is yet to be realized.
I treasure innovation and experimentation. But when a technology has proven to have bad bones – shaky foundational precepts – it needs to have sober humans (and I speak of the inebriation of hubris) look at it and salvage what they can. Note – there are always salvageable parts built on good ideas with good implementation. And there are always parts that should be discarded. I have long lists of both kinds.
I was involved with Lotus Discovery Server in the early 2000s. It was a venture into the area of Knowledge Management – the sound concept that the unstructured data comprising the corpus of a company’s intellectual expertise could be better ranked, organized and thus, found and used by everyone at that company.
It had a taxonomizer – a product of IBM research that would find groups of medium cardinality words (called “bags of words”) and create the “real categories” among documents. That was a miserable failure. It had bad bones and only the glue of theory holding it together.
Now, DS also had metrics –
- how often a document was read and responded to and linked to
- what people did the most writing that got read and responded to and linked to
- what topics/categories had high counts like those
And that part of the product was very cool. And it mostly worked. Why did it work and the taxonomizer fail? Because it used human interaction as an indicator of value. Was it infallible? Hardly. People read first documents in lists a lot, messing up the metric. However, incoming links require work that people perform to vote “yes” to a document’s value.
This is not a rip-roaring new idea – Google’s seminal page ranking uses metrics the same way. And they have been forever perfecting it. The new wave of big-data cognitive computing rides and advances these same concepts.
But small data – inside the smaller confines of a company, work group, discussion database or team room – has hidden treasure that human beings have buried there, sometimes years ago. Projects and campaigns and white papers and just discussion banter are everywhere. As we proved with Discovery Server (though it was never a commercial success), elegantly exposing it forges a new view of that data that promises to expedite and enrich access and raise in value the exploration of all who venture into that data.
It’s no secret that the big forays into AI (ahem, cognitive processing) are fumbling to really take off. Like the Discovery Server, the systems need help to learn before they can teach their users. And even then they must be fed high quality data (that is, correct) or else they will learn from that which is obsolete or just wrong.
But there is definitely stuff built and found along the way to the elusive market smash that is inherently useful, provides clear value and is relatively easy to implement. It’s known and it’s there for the carving out and harvesting. New Domino templates, anybody?
J D Curtis 