Ai has been associated to massive number/data crunching and big data, the bigger the better. The (near) future may see this sort of “top-down” AI flanked by a different sort of AI emerging in a bottom up way, different both in terms of quality and in terms of processes (i.e. how it is achieved).
As a matter of fact if we look at Nature we can see both types of Intelligence at work:
- If we take human intelligence this intelligence makes use of intense data crunching, provided by the trillions of synapses in our brain. If you look for the computational power of the human brain on the web you find a broad range of values (the truth is that we don’t know, and most importantly we do not know how to measure it) but the general agreement is that it may compare to the processing power of a supercomputer (that does not mean at all that our brain can multiply numbers as fast as a supercomputer, it cannot even add as fast as a pocket calculator in your smartphone!). Anyhow, we can say that in terms of intelligence our brain processing shows similar potential of the one of a supercomputer and, as a supercomputer it needs to crunch a lot before becoming intelligent. The crunching takes place over years by continuous solicitations that starts as we are born and keep going on. Our is a learned intelligence.
- If we take a hive intelligence (hives are pretty smart) its intelligence is emerging out of swarms of bees, single bees, individually closer to mechanical automata than to intelligent beings, when playing together result in an intelligent behaviour. The processing power of a single bee is pretty low, a bee-brain has about a million neurones, much less than an IoT chip, but the number of possible interactions among a swarm of bees is staggering (bees communicate by vibration, by sight and even by temperature).
What researchers are now doing is to leverage on the growing number of IoTs, each one embedding some form of processing power, to create system wide awareness and intelligence. This is knows as swarm intelligence (yes, swarm like in the bees case) as well as Massively Federated Intelligence -MFI-, a topic that is the focus of the 2021 activities of the Digital Reality Initiative.
Edge Impulse, as an example, is providing tools to create “intelligence at the edges” by making possible for IoTs to learn. They are using TinyML, a hot area of Deep Learning targeting tiny machines with limited processing power, on IoTs. Nordic Semiconductor just announced a partnership with Edge Impulse to have TinyML running on their wireless IoT enabling learning capability for each of them and creating a whole ambient awareness.
The age of Massive Federated Intelligence may be closer that we thought possible.