Video Transcript:
Recent advances in Artificial Intelligence open numerous possibilities, at the same time underlining the limitations in synthesizing human-like intelligence. For instance, the autonomy devoid of freewill imparted to self-driving vehicles, capabilities of the language models stripped of comprehension and common sense, subsistence sans sentience of the virtual entities in meta realities mark these inherent confines of human programming endeavor. Essentially, the algorithmic thrust for such human-like faculties remains superficial – with automation reigning over Autonomy, correctness over Consistency considerations and hence the emulated Originality over genuine simulation of human intelligence.
The objective of A3ilabs lies in the intersection of all attempts at synthetic intelligence. The fundamental foundations laid out for a general variant of AI within the parameters of Autonomy, Originality, and Consistency provide an alternate though coherent worldview to intelligence & related notions. Resultant method revolutionizes every aspect of state-of-the-art AI, giving reason not just for newer technology stacks but to entirely unprecedented technological paradigms. These paradigms transform very motivation behind the acts of programming, modeling, & testing – that impact the training & consumption patterns of AI. A more rigorous evaluation criteria as an alternate to ‘Turing Test’ becomes possible, for differentiating human and machine intelligences. Within these altered contexts, both the method and application of AI explode to make way for measurable, evolutionary, and perpetual AI systems like the world we live in. The complexities involved are immense though!
Video 1 Summary
Simply speaking, Video #1(/10) provides an overview of the research at A3ilabs and talks about the following –
- Background: New AI technology is impressive, but it is nowhere near the level of true thinking exhibited by humans. Examples of current limits include:
- Self-driving cars can act on their own but are not truly autonomous.
- Language models (like chatbots) can write more efficiently than humans, but don’t really understand what they write.
- Virtual worlds are becoming increasingly lifelike, yet the virtual entities lack both sentience and awareness.
- Context: In essence, current-day AI is all about automation, imitation, and ‘seeming correctness’. On the contrary, research at A3ilabs is defined by the parameters of –
- Autonomy: Based on true comprehension, like humans.
- Originality: Not mere ingenuity of the output, but the capacity to work outside preset programmed or trained patterns, in building such patterns as needed.
- Consistency: Not mere logical reliability of the output, but that which is synonymous with ‘being human’.
- Motivation: The research is driven by the goal of creating a method for developing genuinely human-like AI. However, nothing we know or can do today comes handy – not the finest of our sciences, not the most exacting of our mathematics, not even the profoundest of our philosophies.
- Approach: The problem lies in the strict alignment of our knowledge utilities with what was called the human natural attitude. This attitude nurtures human intellect but meets insurmountable hurdles in creating an alternative or a synthetic source of intellect. So for any method to attain a truly human-like AI, this attitude needs to be negated.
- What’s in it: The self-referential nature of such negation gives rise to operational complexities of an unknown kind. Reconciling these complexities with human abilities paves the way for entirely novel technology, science, and mathematics etc. The envisaged result is a self-sustaining perpetual intelligence system, much like the human world we live in.
