Random

For the past 6-7 years video science communicators via you tube channels like Veratisum, fern, Physics Girl, Vox, Simone Giertz. Cleo Abram, Jhonny Harris etc., have kindof kept my love for science. This was a whole new world that was introduced to me by my partner and have never left me since then. I never knew it was even called science communication or science journalism, but knew that with my teeny tiny ability of story telling I would definetely imagine myself as a science communicator in the future (I am sure I have told this to a lot of people so far and also wrote it in 100s of motivation letters for various reasons). I also talked with a lot of people that if I get into a PhD this would be something I would like to acheive before my graduation.

Until today, when I saw this video from fern released 2 days back. Although I study AI and love to go deeper into the field in a long run, I am openly also vocal about how it needs to be sustainable, can’t be at the cost of human life, energy crisis and people getting dumber with its uninformed use. fern pointed out the crazy amount of AI slop videos (which I was aware of but just not at this scale) of the exact same content from the popular educational creators. And then there are enormously huge amount of money charged in course to teach common people to do the same. The numbers I am talking about are insanely high. This makes the real creators really demotivated and gets lost in the crowd and I see no way to rescue the human out of this loop as it is almost indistinguishable. Good part is we know that AI is really improving but at the same time at what cost and if this is even sustainable in a long run if left uncontrolled?

I am not a person who is anti tech or anti innovation. I so much want that world sees the era of general intelligence (or AGI as commonly used) in my lifetime, which is ofcourse a super hard problem and I believe will require another AI “revolution” both architecturally and philosophically. But now the uninformed use of tech which is still heavily in research and still in the moulding stage is honestly worsening. Fast(?) pace work for early career researchers is ofcourse a concern but even more concerning is the masses who have little to no knowledge about anything behind the blackbox and taking this huge potpourri of knowledge from ChatGPT like LLMs by their word. My colleague suggested, but mankind has reacted the same way when browsers came and same for many other things like mobile phones which are now a part of our daily life.

But the scale of AI revolution is huge in orders of magnitude, especially the pace at which it is penetrating the common households. Knowingly or unknowingly the penetration is in different hierarchies and is much more different than the other tech revolutions. Whether you have to write a basic email asking for a meeting or write a birthday card or make a basic collage for a blog, or ask a recipe, ChatGPT is the way to go. Undoubtedly, they are of huge assistance to us for repetative tasks, or something that is boring in my perspective but part of my daily chore or something like writing software. Ofcourse our BIG dream or what the big companies claim is to live in a world where AI does the boring stuff and we do the fun part. But where do we draw the line?

Do you think that is even possible if people give up thinking? We become paranoid to even write a basic birthday message? We donot know how to study a whole chapter from a book as AI summaries are way easier and simplier to read? We no longer know how to make music, discover new art forms, and most importantly forgot how to think? Taking assistance and compromising on thinking are two very different things. Thinking is basic to human which is why we exists as the superior being in the world and if any revolution is compromising with that, we seriously need to think deeply about restructuring the system and the society, the regulations, the functioning, and “the new world” where we have to coexist with the “AI robots” (if someday they exists) for the goodwill of the society. There are job cuts, depressions, creative people and creators are either going bizzare with slop content or going to emotional trauma to be lost in this sea of knowledge dumb. And honestly in this unregulated use I see no way out, except for countries coming together to impose regulations and having serious discussion, policy makers of the world joining hands with the scientists and researchers to bring the situation under control and ofcourse forcing/imposing the enterprises, businesses or any organization in general to use this tech responsibly.

Coming back to fern’s video and the content, we viewers should responsibly acknowledge and support the people who are still trying to be creative and putting up well researched content with AI assistance. LLMs are still data driven and are consuming more and more sloppy content from the internet and with this amount of unauthentic knowledge dump, I am afraid that if internet browsing and even LLMs using the web and getting quality and truthful knowledge will even be a thing in the future.

While I firmly believe in this revolution, I believe in this human crisis too, which is slowly but steadily becoming a habit of less and less thinking which is not sustainable in a long run AND we need to talk more about this.

I am very happy to be introduced to slow and honest research at a very early stage of my research career. However, while I am currently applying for PhD/research position it is certainly true that practicing slow science is surely a career option in CS atleast at the moment. Only a tiny fraction of people can really practice this.

Why?

I was fascinated by the field of AI and representation learning in particular very late in my masters which already sets me in a disadvantaged stage. I always from the past few years that I loved research and most certainly chose a domain to continue from a year or so. However, now that I look back I was unknowingly was part of a rat race I didn’t want to be. It was simple for me. I love science. I loved the cool stuff modern AI could do. I wanted to learn why things worked like they do. I wanted to dive deep in representations and make something for social good. It was and is still a simple goal.

I was and still am flabbergasted by the amount of expectations and the fast pace of the field of CS or AI in general. Nothing seems enough. I was having a chat with my mentor the other day (he is a CS professor with quite some academic experience ;)) about this and he had similar concerns about how much slow science is now acceptable and how this system will be in future when the throughput of science is even higher. Will real science experiments be faster? Will learning to program no longer be necessary? Fundamentally is it weird to think that how this would even work for the new generation. I use AI assisted tools all the time and adapt to whichever new framework of agentic tools come up, however I also admit that back in 2023 or before I took time to learn the fundamentals of these things slowly and carefully. I was telling him about the expectations nowadays compared to two years back and how everything is moving so fast and we should have to adopt it fast as well to catch up, but then what about the new generation who are grinding publications and giving talks since high school. Is that sustainable in the long term? This has been a thing in all the scientific revolutions and the transition is always uncomfortable.

However, I am optimistic about this revolution in general and we should catch up and find the right balance of slow science yet fast developments. The scientists and the researchers should and will definitely come up with better ways of evaluating or rewarding science which might be the only way out for the betterment of mankind and to get the best out of this immensely impressive yet fast revolution.