So today I created my Kaggle account at long last. It is not that I was delayed for some business. I was well entrenched in my Economics & Finance, Marketing, Sales and Strategy career and not interested in the world of ML and AI till very recently. As I did not belive AI could be sentient. I still do not believe in that. But then my friend convinced me that there was a lot do be done in the next decades without the promise of the sentient nature of the phenomenon. In-fact, that is what drew us to the works of Hubert Dreyfus in the first place about 4 years back. I had argued that I did not believe AI could in a very very long time, begin to understand the seat of human consciousness for our decision-making goes much beyond than mere logic based assertions. I turned from an AI skeptic to someone who now likes to postulate about the re-electrification of the world through the force of AI, or neuricity as I like to call it personally.
But there are skeptics and cynics alright. But Copernicus knew them. And so did Socrates. Or a Tesla too. I want to take you to that June day in 1752 Philadelphia wheen Ben Franklin performed his famous stormy cloud kite experiment. In pursuit of more pragmatic uses for electricity, remarking in spring 1749 that he felt "chagrin'd a little" that his experiments had heretofore resulted in "Nothing in this Way of Use to Mankind," he planned a practical demonstration. While that demo was a little too grotesque for my taste, the kite experiment was just fine. He must have been a little scared of electrocution and wore insulation as another Russian scientist names George Wilhelm had died Years before conducting the same experiments. So you see Ben Franklin believed in harnessing electricity someday. So did Michael Faraday who generated little "sparks to harness" as early as the year 1832 on the other side of the pond. But the progress had been very slow. Till Edison lighted the bulb they say in 1878 which shone for an hour or whatever.
So Imagine how it would have been in the year 1880. For backdrop, consider the Colorado Springs made famous by the movei where Tesla was doing his thing. Westinghouse was ready to invest the money for the projects. The promise of electricity had been made. But I am sure, people would be skeptical sometimes. In 1880, a lighted bulb would not satisfy their expectation. Some pretentious scientists may have given up. Some would become nay-sayers. But not Tesla and Edison. Or Westinghouse. They would march on. They would march on. Right until that fair in Chicago in the year 1893 when Westinghouse finally lit the festival using Tesla's AC, outcompeting Edison's DC for it was more economical. But the win was for all three of them in my opinion against the forces of cynicism, skepticism, nihilism, lethargy, guerillaism and all the other "isms" which stall the progress of humanity. Those people meet their fate.
So in my opinion, we are at a similar cusp of change in the timeline for AI. Like it is year 1880. The light bulb has been lit in a controlled environment like Edison did. What are the self driving cars vrooming through the streets of SF if not the equivalent of the light bulb shining in a room for Edison. The questions really is, can we light the fair in a few years with grid. A grid like newron.ai. I do not pretend to be any kind of a leading participant at all in this field. If at all, I am humble student of it as I come from Economics, Marketing and Finance background. But I made my Kaggle account at long last. And I would invite you to make yours and explore this adventure together with the actual leaders in the space. Like the team at newron here. Or a myriad other product startups the worldover who want to make this re-electrification happen. As I told my friend, afterall it is fun to imagine and deliver the future. Or try. Rids you of the cynics in the least.
The Asphalt ahead - unlaid mostly
The unlaid road ahead I much longer than the laid behind
I have been trying to understand the AI field now as I intend to continue my work in this field for a long time as I am now convinced that this exciting future is upon us sooner than anticipated. But one question that had been nagging at me for a few days earlier was regarding the depth to which I should go onto as far as models and algorithms were concerned. I started reading material on the different kinds of algos and stuff like how Neural Networks were more associated with deep learning. Just beginning to understand the ideas of a layered set of neural connections so to say. Now the numbers 8 16 64 120 etc used as postfixes had started to make sense to me in the literature. I guess. The top level understanding like k means algos used for clustering results and SVM for more prediction based stuff had settled in. I guess. I was reading through concepts like Backpropagation and Actovation Functions. Without understanding too much. But I atleast started to know what the meaning of assigning weights and biases to models was. What was the hyperparameter that was supposed to be tuned. Then I was wondering as to how much I needed to understand gradient descent or even the mathematics behind some of these algorithms which were always doing the rounds in the literature. Honestly, I had completely forgotten simple regression methods I had learnt more than 15 years back when I was building a career in finance and prepared for interviews. Black Scholes was a gades memory. And I was getting overwhelmed. I had introduced myself to some basic sysyems like the Kaggle Notebooks and some frameworks like Pytorch and made signups on someone these sustems. I was not sure I needed to learn all of the deeper stuuf in the first place as we were talking about building a model tracking and data-model veesioning software. But I was getting sucked into the models I assessed. Even when job was to build the plumbing lines.
Then I stumbled across some research and market studies by authentic organizations and bodies dedicated to the cause of AI. And that demonstrated as to how the actual models were infact going to be just a part of the bigger piece more and more in the future. Ofcourse that is the central theme of the Data Centric AI paradigm for the future that guys like Andrew NG speak of. But the research I quite spoke of things like Governance and Security. And once I understood what that stood for, I realized how crucial the whole dataset aspect of the model-data combine was going to be. In addition to the plumbing lines. The challenges which this massive re-electrification would have would be so much more than just the models.
And most of that roadmap lies ahead of us. I cannot quote a number for I am still learning and forming an opinion. But I reckon nobody truly can and that is because the roadmap shall be much longer compared to what we have trudged so far. Even with all the accomplishments and achievements made in the field. And that realization made all of this so much more exciting. We are just beginning the journey. Something like they were on a day in 1880.