It's as if Neal Stephenson wrote a book about AI without knowing what AI stands for. Still, some interesting bits even though I wasn't interested in most of them. ( )

The late Freeman Dyson's sometime adventurer (and societal dropout?) of a son presents us with a mix of 3 kinds of story: (1) North American history, including Russia's early acquisition of Alaska and the US's later subjugation of the Apaches; (2) autobiographical periods, including childhood in Princeton and boat operation (and tree-house living!) on the continent's northwest coast; (3) technological history, including heliography, vacuum tubes, the Orion scheme for nuclear-powered space travel, and (most importantly) modern computer technology and AI. The second-to-last chapter refers to digital computing as "an alien construct growing in our midst", and the culminating chapter uses Cantor's continuum hypothesis (no levels of infinity between countable and uncountable) as a framework for arguing that analog computing is bound to loom much larger in the future.

We have been dealing with unintended consequences of technology for hundreds of years! Yes neural networks rely on internal models we cannot immediately fathom (although actually these days there is a lot of research into probing them to visualise features they are using in their 'deliberations'- deepdream and even pixma are examples of this), this is not that far removed from dealing with the many other complex algorithms and deterministic paths in all complex software, for which there is a large and still evolving canon of methods.

The real danger we are currently in with 'AI' and the IoT is the hype driven danger of premature deployment in many situations with potentially serious consequences. Also, inevitably, when the technology doesn't live up to the promises, we will have another AI winter.

So is it that the algorithms are doing something other than what the people who designed them designed them to do, or just that they eliminate the need for office workers?

If the former, then that's really a problem for the influencers; if the latter, it's the same problem for those being influenced as it always was. Strange that manipulation of the populace has suddenly become this terrifying problem.

Anyone who works with technology, and has half a brain, knows that technology isn't even close to becoming beyond control.

It's not the technology that's beyond our control, but the companies behind that technology.

Programmable beyond control? This is a rather anthropomorphic view. It implies the "technology" is actively pursuing a goal.

It isn't. We, the users - the people who have ultimate control, since we know where the OFF switch is - are building machines with the ability to "control" aspects of our lives (such as feeding us an endless supply of FAKE NEWS) and then we are actively setting them to do these things.

Nobody has been forced to use Google or social media. Nobody is legally obliged to evade their parental responsibilities for a smartphone. Nobody HAS to let go of the steering wheel of their Telsa. We willingly allow third parties to assume control, just as we have third parties that remove our rubbish, make our takeaways and create of TV programmes (and our TVs). And neither do we feel the need to understand how to do those things, ourselves.

However, we must also understand that humanity has not evolved any new intellectual powers over the past 30 years. Given that most people in the 1980s were utterly bamboozled by their video recorders and microwave ovens, it is unrealistic to suggest those same people - or their children and grandchildren - would be any more capable of setting up their Alexa, smart TV, Facebook sharing options or roomba.

There are no "algorithms" in existence today that can interpret news headlines with an accuracy anywhere near what would be needed to determine market strategy. At best, modern Natural Language Processing techniques can classify the sentiment of a document (such as the title of a news article, or its content) into "positive" or "negative", etc., categories.

Whether this is sufficient to change High Frequency Traders' behaviour is very doubtful. Most probably, the humans in control of the HFTs take action and modify the HFTs' settings, then the HFTs carry out the chosen strategy.

In the end these machines are just designed and programmed by people...so how can they be perfect or make perfect decisions when the people who designed them are far from perfect.

I barely trust the largest companies in the world to design a usable word processor or an App for booking a haircut...the idea that so many people trust the same geeks to program algorithms to run entire Govt Depts or our financial system is just bizarre.

We call it "holding up the sky". Because software is everywhere. Western and Chinese lifestyles are dependent on it to get food to them, to make them electricity, to run almost everything. Much of it is badly written because it must be cheap and good enough software engineers are rare and expensive. More is badly written because no-one can make clear decisions on how it should behave for those engineers to implement. Hardware is just as bad; to be cheap enough, protocols are only half implemented. Much of it is designed by software much of that ad-hoc and difficult to use...But now, somehow, it all HAS to work.

No, technology evangelism based on technology-fuelled capitalism has evolved beyond our control.

It's a pity there are no modern equivalents of Leibniz. The separation of science and philosophy that has become the inevitable modern paradigm (due in part to the sheer extent and mathematical depth of modern knowledge), does not serve humanity well. This separation is best evidenced in the leading (at least most resourced) hypothesis of modern science – Multiverse Theory. What would Leibniz make of what Multiverse or the Measurement Problem might offer as a conception of the universe (or is it 'multiverse')? Might he suggest that the emperor is without clothing? ( )