Note/AI Job Market — »A Moving Target«

👀 — »Hardware and chip design is in scarce supply and well paid. The Terafab project, the Stargate buildout, and every AI accelerator company need electrical and computer engineers who can do real silicon work. Power systems engineers. Controls engineers. Embedded engineers. RF engineers. Some of the best-paid and most stable specialties on the board. Few new graduates pick them.«

This observation enunciated in Srivastas insightful Substack article “AI Jobs: A moving target” is indeed correct. The chip design »niche« is in demand if the buildout is really going to go forward as envisioned.

It (i.e., chip design) is also a »niche« that AI had a hard time penetrating (despite official corporate rhetoric) due to the creative aspect of the design process that requires one to discover often highly unlikely and unusual ideas, approaches, and solutions at various levels of abstraction. To meet the complexity of translating an abstract computational representation (e.g., a design expressed in Verilog) into advanced technology nodes requires a gargantuan planning horizon with many feedback-directed steps, an area that seems to pose particularly recalcitrant problems in the realm of computational complexity. Finally, at the lowest levels, challenging constraints imposed by the hard sciences—i.e., the physics of today—pose limits that await creative solutions and understanding in problem domains that require—yet again—new insights and explanatory power to overcome some of the mysteries setting seemingly impenetrable contemporary limits.

One ought to not underestimate factors such as the veil of secrecy and confidentiality of IP design and silicon manufacturing that severely restrict access to the processes and problems in that domain to a broader audience (except for those that maintain exclusive ownership over the intellectual property components that are involved), posing real limits on the broad applicability of AI in this segment of the industry.