Earnest Projections and Advice, page-101

Richard, you have hit the nail on the head, but my concerns go further - how does a mining company have the management experience to evaluate and implement the commercialisation of a computer hardware product.

I have had an ongoing interest in artificial intelligence since the early 1980's when I was dabbling in LISP and Prolog, the then wunderkids of the field providing specialised software tools to replace general programming languages and thus representing a leap into the future.

The one thing I have observed from my travels is that AI (coined in the 1950's) is always 10 years, or one amazing technological advance, away from changing the world. But it seems the dream has been fully realised.

I accept that implementing neural networks in hardware rather than software does offer significant advances in speed and parallelism, but that is not the end of the story.

The chip does not learn by itself, it has to be "programmed" or loaded with data in the field with which it is to apply. Think of a doctor, or any expert in their field, they have to be trained for years by means of a rigorous program to bring them to a significant level of expertise. So to must the brainchip be loaded with expertise, and that is what takes the time.

The area where the human brain analogy falls down is that a neural network must be trained externally, it must be provided with the required data, so it reliably and consistently gives the same answer for the same input. The chip lacks the human capacity to develop internal hypothesis and choose its own direction of learning - and thank goodness for that or we risk SKYNET.

There is much to understand in the field, and very easy to be caught in the excitement and hype. Prudent investors will recognize this, and ask pertinent questions like - how the heck do a bunch of hole diggers suddenly acquire the expertise to build a bunch of computers?