A friend of mine, an experienced circuit design engineer, told me the other day that one of his managees had recently submitted a paper to a well-respected conference on signal processing. The paper was apparently on a novel modification to a commonly used signal processing algorithm and the resulting simplifications and savings in its hardware design. Two of the reviewers had hailed the contributions of the paper and recommended acceptance. A third reviewer, possibly a confused graduate student trying to find their way in the current AI frenzy, had rejected the paper with a comment to the following effect (I am paraphrasing as I have not seen the comment myself): “The area of circuit design for signal processing is very old and not of interest to the community. I do not see how this work contributes to the general framework of LLMs.”

I grew up in the pre-Internet age. The large corporations or rich individuals that we would hear about or occasionally see portrayed in movies were mostly energy companies and oil tycoons. There were of course various large manufacturing companies as well, from cars to home appliances, and many other products. I may be wrong or misremembering things, but the image that I have in my mind about those companies is that they were heavy weight. They had a lot of inertia and would not react instantly to every new and half-baked development in the industry. They wouldn’t make major financial or workforce decisions based on short-term market fluctuations. I would imagine academic and scientific research organizations had similar approaches to their decision makings on investments in research projects.

Fast forward to our current time (and past a couple of financial bubbles), modern large corporations seem to be making investment and hiring/firing decisions based on market reactions to their quarterly earnings announcements, or worse, based on transient and emotional public sentiment towards new and shiny technological developments. With the advent of social media and the flood of new opportunities for digital marketing, many tech companies were in a hiring frenzy for software engineers, some hiring by the hundreds a month, not because they had projects planned and lined up for them, but only to make sure they were not hired by their competitors. And of course, we saw them layoff employees at the same pace in reaction to new market developments. Advertisers themselves do not seem to be spending money in digital advertisement so that they get the word out about their products, but rather to gain a sizable share of the flood of the ads that are constantly shown to the social media users, adding to the same flood in the way. Even the current frantic competition of companies of all kinds to purchase as many GPUs as possible, in many cases seems to be only to ensure that they won’t fall behind their competitors, possibly in the development of a new chatbot for which they don’t really have a plan or need. And as we all can see around us and is especially evident from the anecdote at the top of this post, academic research does not seem to be immune to this frenzy either.

We seem to be living in an era when scientific research and technological developments have turned into a continuous fashion show where with every new shiny topic or buzzword, the entire community flocks towards the same thing, investing billions of dollars and thousands of researches into often pointless projects with marginal contributions at best. Most of the resources of our major institutions seem to be spent in fighting/competing amongst each other, and not in the direction of solving real and serious problems that humanity is facing. We take pride in our advanced technologies that have broken many geographical boundaries and enabled the people around the world to have instant communications with each other, but that opportunity is mostly used to share AI-generated content that contributes to the constant flood of clutter on the Internet. And companies are investing billions in new data centers to house increasingly larger models trained mostly based on the content that earlier models have generated. In my own field of telecommunication systems engineering, you can see research and standardization organizations desperately trying to somehow insert AI/ML into any layer of the protocol stack, whether it makes sense or not. Just today, a friend of mine overseas was telling me that when he was offered collaboration with a publication in abstract mathematics, the editor told him he would welcome any submission in the area of machine learning. I know I am stating the obvious, and that there are many who are constantly thinking or talking about these issues, but I feel the subject needs a lot more engagement from everyone in the community, to address questions like: Is this really the direction we should be heading towards? Shouldn’t our science and technology leaders try to slow down a bit, take a step back, and assess the true value of what we are investing so much into its development?