AI out; 5G and EVs in

Could you ever imagine a situation where recruiters would tell students that they cannot be shortlisted because they only have AI/ML on their resumes? They don't have other "old school" skills such as theoretical CS, hardware design, OS, compilers, networking, etc. This is exactly what I as a program coordinator am seeing and students are narrating this year.?Two new stars have replaced the reigning king -- AI/ML.

A quick flashback is due before we finally talk about our new technical lords at the end.

Let us appreciate that mass hysteria has a lot of power -- a large collection of people who think the same even though each person may have a different reason to do so and may claim to think independently. Such "thought waves" define the trajectories of modern institutions starting from democracies to stock markets to cryptocurrencies. They have fueled communism, fascism and other "isms" in the past. A small announcement is made and stock markets start crashing all over the world creating many new millionaires and destroying the fortunes of many others in the process. There is some logic but a lot of sentiment as well. People tend to think as "one" regardless of their differences, be it a banker in Wall Street or a tribesman in Namibia.

AI was one such wave, where tons made tons of money by convincing people that this is really what they needed. We had car companies say that they are actually "data companies", semiconductor chip vendors introduce themselves as "AI companies", and e-commerce sites rename themselves as "data-driven smart vendors". The fallout of this was that engineers who drove the real business such as the ones designing better engines?in cars, or creating more efficient digital circuits and distributed systems got discouraged. They felt slighted and sidelined and some moved on to companies who were not ashamed to say that they only designed chips or wrote operating systems and did nothing else. Some focused on designing cars with better emission, mileage, comfort and safety figures, whereas a few focused on software to automatically change lanes. No credits for guessing who prevailed.

On the other hand, the hype about AI/ML was so high and the stakes were so large that sooner or later companies had to show big money flowing in. They had after all taken the risk of alienating a large part of their own engineers and unrealistically raising the expectations of their consumers and investors. The Ukraine war proved to be the straw (or the log) that broke the camel's back. In the last few months, the news of mass layoffs in this sector has risen from a trickle to a flash flood. It has become quite clear that the money is simply not there and a lot of overhiring or misdirected hiring was done in the hope that a thorough analysis of let's say consumer behavior will yield disproportionate benefits. Little did folks realize that this is a constant-sum game and what matters much more is the quality of the movie rather than the recommender system that predicts the movie that one may want to watch. The "big" money never came. There are information theoretic limits to what can be learnt and how much of that is useful. The economy also places limits on the total amount of ad or e-commerce revenue; it can only pass from Peter to Paul but does not increase as a whole.

Hence, given the pressure that was building up, a market correction was due. Now at the end of 2022, the change is quite visible in MS/Ph.D applications and the job market. Car companies are now "car companies", and chip companies are now "chip companies". Those who spent their time on Keras and TensorFlow are now having second thoughts, especially after getting rejected in job interviews because they don't know how a pipeline or virtual function works. Needless to say nobody has heard of a Turing machine, the Pumping Lemma, Shannon's theorem or the Banker's algorithm. Sadly, we are back to the basics.

In the tech. world such boom-bust cycles are the norm. It was clear from day one that AI/ML will not reign forever. However, this era has left behind its indelible footprint on academia and industry, in a good way. There is a big gap between what is taught and what is needed in the area of computer algorithms, which is why AI/ML arose in the first place. Beyond a basic book on algorithms, we really don't have much in terms of deterministic and explainable computing. For harder problems, either fast solutions don't exist, or there are approximation algorithms that are seldom useful in practice. Other than AI/ML, there was no other technique to turn to if difficult problems like face recognition had to be solved. Hence, we can perhaps look at AI/ML as Advanced Algorithms -- a semi-explainable black box. In my view, an ML course should be a core at all levels (Bachelors to Ph.D). It is hard to thrive without it, even though the current state of affairs also prove that you cannot thrive only with it. Second, data has its value, it is not the new oil, but it is clearly something that cannot be ignored, and the techniques to process data need not always be explainable -- it can perhaps be friendly snake oil. We should learn whatever we can. In today's world, it is hard to think of an organization that does not mine its data and use the derived insights to optimize its operations.

Now, the new stars appear to be 5G and related technologies (security, wireless comm, malware) and electric vehicles (EVs). Their time will also come and they will also one day be replaced by something else.

A short Hindi couplet by Sahir Ludhianvi.

Aaj main hoon jahaan kal koi aur thaa. Ye bhi ek daur hai woh bhi ek daur thaa.

Today I am there where yesterday someone else was. This is an era and that was also an era.