What about AI evolution ?
Some food for thoughts on incoming AI evolution.
– What do you think are the implications of AI’s recent evolution for technology, business, and society?
First of all, we need to cross that mindset barrier which could be resumed to this quote from 1828 french economist about the rise of motorisation revolution : “Nevertheless no machine will ever be able to perform what even the worst horses can – the service of carrying people and goods through the bustle and throng of a great city.” While experts and strategists know for certain Ai will go far beyond that, some still think it will remain limited to semi accurate chat engines and will never overtake complex functions.
Autonomous driving system is a lead example of this bias, where human still think, despite obvious flaws and failures, that Ai cannot drive better than they do. We are reaching a turning point where mankind will divide on four categories about Ai : Optimists, pessimists, pragmatists and doubters (as categorised in S. Makridakis study about Ai impact on society). From a complete transformation up to genetic control and brain connection to the cloud to a relegation of humanity to a second role where computers and Ai are taking the critical decisions, up to selecting candidates to lead democracies and dictate laws. The future is very likely to be in between, and most conditioned by what humanity will want to achieve with Ai, minus the non-negligible aspect of preserving social peace in a world where many jobs will be replaced by artificial intelligence.
IT environments are the first to rush for Ai. It is a computing landscape where adding a form of intelligence is the continuity of simple automation that was occuring in the past decade. Business environments are now the second one, as it become clear for leaders that Ai can greatly reduce costs in many ways. Societies are the slowest element to change, as it’s mostly human based and much more inertia. It will move at a pace limited by the necessary need to keep social peace, replacing transformed jobs with new roles and activities.
On the model part, the opening of Ai systems (due to its huge data needs) will naturally lead to more open systems. The only limiting element to that free access now is the running cost that has to be charged. Proprietary Ai engines, because they need to be sealed, protected, will not be able to stand versus open ones in the long run. When Ai will reach society level implication, this energy cost could be split at state or region level, becoming almost free to use for individuals.
Main risk I see here, as it was with previous industrial revolutions, is the gap extension between regions or societies with and without Ai strong footprint. Either from a cultural mindset or economical challenges, this could cause massive break between groups of humans accross the world.
Second risk is purely energetic. Ai is a very high energy consumer, and in a world limited by energy availability (i.e. fall of fossil energies and lack of renewable ones), Ai development can be halted to route energy to more vital usages, such as transport, housing, etc. This is a big challenge for the 21st century while we are aiming for a new energy source (like nuclear fusion).
– How do you think current ML techniques could fall short? What could be gained by creating more human-aligned AI systems?
Once much of available data will be ingested by Ai systems, the next frontiers will have to be human-aligned Ai systems. That’s where imagination and creativity production are limitless and can continue to grow Ai ecosystem. Not only it will move into this direction, but this alignment will also be crucial to push further Ai intrication into our world. That will be the moment where the couple Ai/human will be defined, as to where the marker on the decision making process will be drawn between the two.
To explore deeper : https://news.mit.edu/2023/mit-congress-inflection-point-ai-0310
