To be or not to be (Intelligent)

 As we stand at the dawn of a critical decade in AI and machine learning, I find myself reflecting on the incredible journey these fields have taken. AI has long been part of our lexicon, yet it's currently at a defining moment that will shape its trajectory for the times to come.

Today, the media either paints AI as the ultimate problem solver to any challenge, or the big evil replacing mankind  and making us obsolete.

As often, it’s difficult to stand in- between without being rated too conservative or not aware enough.

Yet as in many domains it seems we are again both overestimating its short-term impact and underestimating its long-term potential.

Indeed, we have now mastered the art of creating huge machine learning models which can identify the patterns present in specific datasets which such accuracy that they can produce endless creative variants from these initial datasets. a field that's surpassed human ability, and a problem about to be solved. In popular media, AI is chiefly depicted via seemingly sentient language models and AI-generated art that captivate the imagination.

But the reality is that AI and mainly GenAI has excelled in specific, narrow areas. Our AI models are all incapable of true intelligence. Their sophisticated pattern replication is a mere mimicry of human creativity as opposed to true creation. Their achievements, while technologically impressive, don't equate to the friendly and context-aware robots of science fiction. As a result, AI's current application remains often focused and cautious, frequently confined to low-risk environments.

Is it a proof of failure? Not at all – if you think the current technology is mature and future improvements will only be incremental, be ready for major surprises. Looking at the progresses made in the last 12 months, the next 5 to 10 years will bring massive changes in our business and private lives.

But a key question therefore emerges:  How should we develop our models going forward if we want them to be more “intelligent”?

• According to one school of thought, scale alone has been able to produce ‘’intelligent-looking’’ results. Therefore true AI intelligence will emerge if we keep making our models bigger and bigger…
• Another philosophy is that true intelligence only comes from systems that explicitly reason about real-world concepts, developing a true general intellect rather than just copying patterns of behavior from other intelligent beings.

I believe that elements of both will be part of an ultimate, hybrid solution. The massive deep learning models and their remarkable ability to manipulate huge volumes of complex data will work in tandem with smaller dedicated models and advanced symbolic reasoning to create new breakthroughs in the capabilities of AI.

But the push for AI and especially for larger models as they are built today confronts a true environmental reality – the substantial carbon footprint and mainly the energy requirements of training these systems on zillions of domains.

One thing is clear however, the pace at which AI is advancing is nothing short of exponential.