AI - Visions of (not so distant) Futures

The world of smartphone apps is split into two competing philosophies : while countries like Russia, China, Indonesia and Thailand have embraced the super-app concept (one app to rule them all), the rest of the world has moved in favor of specialization, with apps doing one thing and doing it best. The world of AI is yet to mature, but it too could follow a similar roadmap – we could have one wide AI to rule them all, or have multitudes of narrow AI working together.

The first scenario is an all-encompassing AGI. This too could pan out two ways – we could have one AI trained in so many skill sets that it becomes indistinguishable from AGI for majority of use cases. It could be similar to how our smartphones run on an OS and we download required apps – there could be one base AI-OS into which we download (or train) skills. A more interesting second option is true AGI – not trained in any skills, but able to independently reason its way through any problem statement. If it does emerge, true AGI will be the last, because by definition it will be able to better itself in all aspects at a pace orders of magnitude higher than humans would be able to keep up with. In this case (assuming it doesn’t enslave us all) this AGI will permeate all aspects of our life, hyper-charging development & optimization to limits of known physics.

The second scenario is AI evolving into an ecosystem of narrow specialization, with no common link between them. Certain AIs become really good at one or at most a few tasks, and become dominant in that field. Example could be ChatGPT dominating the generative AI space in text, image and video; Microsoft copilot dominating the coding space, Siri dominating the personal assistant space etc. But it will not be specializations in silo. Just as the internet exploded in capability once websites started talking to each other, so too here; AI bots will be constantly interacting with each other, pulling data and requesting outputs. Websites and applications, or their equivalents in a post-AI world, will be engineered for interacting seamlessly with these AI bots, similar to how they are currently engineered for SEO.

Going a step further, bots could even be empowered to make decisions on your behalf, based on its understanding of your preferences and patterns of action. Multiple daily decisions could be automated and optimized for. Your AI assistant would be talking to a cab aggregator AI to book the cheapest and shortest cab ride to your destination such that the cab (driverless?) is available at your doorstep just as you are stepping out the door, without you ever explicitly having to book it. The food delivery app would know your individual taste preferences and will create a personalized menu for you daily, maybe even automate the task of ordering food by predicting your mood based on how your AI assistant tells it your day is going. Your OTT platform will know what type of content works best to stimulate emotional responses in you, and will be generating personalized movies and shows – the story and script generated on the fly, with the most popular AI actors of the day emoting and vocalizing just for you, and in real time. There might even be a democracy bot that enables direct governance as opposed to our current elected parliamentary systems -basis its understanding of your views and preferences, the AI votes on your behalf personally, for every policy decision affecting you. The possibilities are endless. AI will enter every lossy bottleneck in our lives and bot-to-bot systems will optimize for it.

If these futures sounds like a fanciful flight of imagination, we only need to remember that at the dawn of the 21st century just over two decades back, digital assistants, cab aggregators, food delivery apps and OTT platforms were all fanciful flights of imagination! Standing on the shoulders of these AI giants, whether it be one or many, humanity will start to glimpse the future.

While all this sounds rosy and utopian, the dangers of AI are also very real. Unlike other tools developed by man in the past, none of which had any morality of their own, but were only as good or bad as the person wielding it, AI will be the first tool to take ethical calls. While it is almost a given that AI will evolve with ethical considerations inbuilt – Asimov’s famed laws come to mind – it is also almost a given that we won’t be able to factor for edge cases and emergent behavior. And given AI’s exponential learning ability, it also possible AI will evolve beyond the confines of human morality. Misalignment between human intent and AI objectives is not science fiction, it is a reality even the ChatGPT team had to face when releasing GPT4.

?Given all this caveats, the million dollar question remains – will AI remain in humanity’s control? Lets explore in the next part!