The accelerated evolution of technology in modern times is a bizarre and fascinating phenomenon. We love to debate how it happened and where it will end but need to remember that while the tech side is interesting, it is not always that relevant:
There have been many innovators who did everything right on the tech side of an interesting idea but were unlucky on timing and did not see results. This can sometimes lead to selling off the opportunity only to see it succeed for someone else in the near future as the market has evolved.
Predicting these trends is a mysterious art for business startups but also has an impact on tech evolution. We are very aware of the maturity of facial recognition algorithms as there is a high demand for the service. Something like photogrammetry is more niched by default so will never have the same market demand.
However, as the applications of AI become more mature, more accessible and ultimately more affordable, we can definitely expect some serious evolution.
- Music was an analogue process - it evolved into an electronic one.
- Photography was a chemical process - it also evolved into an electronic one.
This concept is from the highly recommended "Prediction Machines: The Simple Economics of Artificial Intelligence" by Ajay Agrawal, Avi Goldfarb, and Joshua Gans.
The video below provides an overview - not to be missed.
A recent innovation with neural networks - “Instant NeRF” - is analysing the light in a small number of photos and effectively reverse engineering the geometry from this. This process of interpreting perspective is similar to how we process an image naturally so is an interesting example of AI.
As well as being fascinating, it is also amzingly quick with spectacular results. AI and neural radiance fields (NeRF) is not ready for mainstream consumption yet but the entire concept of the Economics of AI is that it's evolution will follow the same trend as the transistor and branch out even further so we could expect to see this in the near future.
This puts even more emphasis on one of the most important rules:
Do not discard the photos used to form the models.
As the systems evolve, we can re-process them, especially the projects that didn't work so well initally. It is a frustrating experience to search for these photos in an archive and realise that they were erased when they could now be useful.
