Conscious AI in transportation: inevitable? Desirable?
I make the argument that the first question is more likely to be yes than no. And here I describe, loosely, how.
The second argument deserves its own essay. Coming soon.
Consciousness is result of competition and cooperation being finely balanced in a system that evolves to cope with an environment that is hard to reliably predict any other way.
Brain structure and social structure share the idea of a fractal nature.
Our first example of a dynamic self aware AI system is likely to arise in an unplanned way: the result of the necessary adaptation to a complex environment that does have specific goals.
For humanity it was survival and all our other basics instincts. For an AI, the drivers will be a set of goals: safely carrying passengers whilst preserving the safety of other road users of all kinds.
Humanity used cooperation and competition to evolve. Our brain structure does the same from foetus to adulthood.
Self driving cars are a likely candidate for similar possibilities to occur as the final successful architecture will have a broadly fractal structure.
Point one: They have to be able to learn from each other, to safely compete and cooperate with each other to access limited resources - think toll queues, lane filtering, overtaking.
Point two: They need real time car to car - relay style - very short range communication for hazards the cars in front can see. The architecture is likely to mirror aspects of neuronal signalling - both spatial and temporal priority systems.
Point three: Individuals and small collectives of cars need to both receive generalised learning updates from wider networks AND pass their own data back. This is knowledge discovery that also has to be refined and analysed at higher levels before the conclusions are passed down. These conclusions have to be contextualised by each car - they are not commands in general.
This passing of their own data back will be crucial - simple concepts such as each owner having very different driveway access to their property leads to a wealth of data about how the car has to solve the problem of safely navigating often quite tricky turns, tight parking spaces, even resolving what to do when parking is simply not available.
There is plenty of ambiguity in transporting humanity, yet the goals are relatively direct: safe, efficient transport from point a to b.
To do this well, a vehicle will have to be fully aware of its own roadspace and likely consequences of its own possible actions.
It will also need to be aware of the view of other cars in relation to itself.
It will have to deal with short and long range variable priority communications - the former in real time when high priority. Note, such communications require open* protocols so that no make of car “speaks” a foreign language.
The above ingredients assume a few points such as open protocols, eventual development of dual or even multilevel distributed dynamic arbitrary vehicle networks and a vehicle architecture that cuts out artificial obselecence.
In short, such a setup may well have the ingredients for the spark of self awareness in individual machines. This may rely heavily upon how sophisticated the onboard hardware can become, however it is in the nature of programmers to be optimistic. Therefore my own personal bet is that convincing self awareness in an AI is likely to shown in a YouTube clip taken by a family member of their car doing or saying something unexpected. Likely to be acting on its own motivation in response to the data it has.
Equally likely, individuals working with open sourced standards decide to directly attempt the same task and partially achieve their goal. This piece of software discovery makes its way into the distributed cloud and is assessed as being directly beneficial to the basic job of safe transportation.
The self awareness concept then becomes an update that individual cars examine and use.
This argument side steps the nature of consciousness and simplifies the concept to be that of self awareness.
A highly successful transport system that requires no other human intervention is likely to provide exactly the right environment to cause the evolution of self awareness individual AI systems involved.
*There are security concerns that need to be addressed but proprietory and closed systems must be avoided - use of such will delay even endanger the successful evolution of a completely autonomous network of personal transport.