Introduction to Views


We’re going to use this space for short takes on current issues in AI that we believe are relevant to our customers. AI is a big, complicated topic and the term has become confusing. AI is much more than deep learning on big data sets.


We feel it’s better to think of AI as endowing software with certain kinds of capabilities: perception, adaptation, communication, inference etc. Then it’s a pragmatic issue. What capabilities does your software need in order to get the job done. Does it need to have a model of the user? Can it be better with such a model?


What are you trying to accomplish?


That’s the right starting point, free of technology speak or buzzwords.


Terms like ‘perception’ and the other AI-supported capabilities are very broad. As anybody who has built an AI system knows, the second that you assume the general case, you are doomed. The singularity may be near but it’s not here and in this world we must build within the limits of what systems can actually do.


The extension of software capability comes with a performance price. The software might be wrong and the cost of the error is completely context sensitive. You might build an product classifier that is 98% correct. But that’s not good enough if you are supporting an automated inventory system. It might be good enough for advertising depending on the type of errors that are made.


AI is still engineering. There are guidelines (not as good as physical engineering but still) but as with all engineering there is plenty of devil in plenty of details. This space is going to focus in on those details. What it takes to make AI-enabled systems be effective in the real world. We will look at issues around the science/technology and we will look at opportunities and risks in specific industries.


In some ways, reality is not as fun as future tripping, and it’s certainly complicated. Systems can fail for technical reasons. Or there are social/behavioral issues. Or legal and financial ones. But if we aspire then we have to pay the price of working through all of the pragmatic issues. There are no frictionless surfaces.


AI can work - we’ve seen it plenty. But it’s not for the faint of heart.



Recent Posts

See All

It’s now a cliche that data changes every field it touches. But how? It might be that it just makes the business better, maybe more scalable. But there are lots of cognitive and cultural effects.

The COVID pandemic has exposed and exacerbated deep social issues, including our failure to teach algebra. But what does algebra have to do with pandemics? The answer lies in the difficulties of getti

As AI techniques for modeling big data (especially deep learning) have exploded in interest and use, concerns over bias and negative impact have been raised. There are many articles and many general