How many junior engineers can be replaced by AI, and how many AI agents can a senior engineer successfully manage? This seems to be a common theme these days, especially in the software development world where there are lots of freely available examples on which AI can be trained. Whether that can expand into embedded software is still somewhat uncertain. There are numerous reports about new graduates with computer science degrees not being able to find jobs. The unemployment rate for them is now higher than the average, something that has always been the reverse in the past. Enrollment will likely drop, especially since AI is increasingly being seen in a negative light amongst the younger population. In the semiconductor space things are moving a little slower, but there are signs that it is accelerating. A common message is that agentic AI is being used in the verification space and having a fair level of success. At the same time, most companies are saying that verification will never go out of fashion — you can never do enough verification, and you can never do it quickly enough. The demise of verification has been portended many times in the past, and yet the demands have continued to grow, and the problem becomes increasingly complex. But what about those junior verification engineers who have been displaced? The first round of AI introductions, the chatbots, were slated as being a benefit to junior engineers. They would not have to learn how to use complex EDA tools because AI would do that for them. They no longer had to master new languages. AI would deal with the necessary syntax. But none of this really takes long for a budding engineer. They can pick most of this up fairly quickly, but okay, AI could help them do that quicker. Still, that isn’t what makes them a senior engineer. That requires experience, judgment, and being able to do the things that cannot be captured by AI yet. Two things are cause for concern. First, if junior engineers have no value, how do they become senior engineers? Eventually, they will be needed to replace the current senior engineers as they retire. This assumes that the role of senior engineers will not be displaced equally when AI super agents become better. That means a development path for all levels of engineering has to be found. Second, creativity. When RTL and logic synthesis first appeared, they were disliked by the existing logic designers – the senior engineers of the time. They could design better logic than the early synthesis engines, and they were not ready to pick up something that could potentially displace them. It was junior engineers, who had nothing to lose, who adopted synthesis and showed that while the senior engineers could still beat them on PPA, they could do it in a fraction of the time. The productivity boost was significantly greater than the PPA they were giving up. There is often an overlooked opportunity. I can remember my first job and a couple of projects I was given. One of them was for an avionics system that had always consumed two boards in the standard sized box. A new component had just been made available, and PALs were starting to emerge. These were the very first programmable logic devices, long before FPGAs became available. They asked me to see if the system, using these new devices, could be shrunk onto a single board. This was not about whether I could create the perfect design, or even one that would be worthy of a production system. It was to see if I could demonstrate a new architecture. Normally, we think that is the role of the senior engineer or architect, but they can become stuck in their ways, and perhaps they are not the best equipped when it comes to researching new devices or new possibilities. Their time is better spent working on getting the production systems designed, working, delivered. As a junior engineer, I could spend the time reading lots of manuals, performing studies about how the devices functioned, learning how to use the new equipment for programming them, what it would take to use them in a production system, understanding implications for getting air certification — all of that sort of stuff. Of course, this was all under the supervision of the department head, but they did not get in my way — just offered helpful advice when I needed it. As it happens, on my first attempt, I didn’t quite reach the goal. I never had the chance for a second attempt because that is when I moved jobs and entered the fledgling industry we now call EDA. It may be enticing to reduce the number of junior engineers a company employs, but that is short-term thinking. If senior engineers are busy managing 10 or more agents, they do not have the time to do other things. They do not have a lot of time to do the things that AI cannot do. So how do you effectively use junior engineers to help them be more creative? How do you use junior engineers to reimagine the task? How do you train junior engineers to be what a senior engineer will be expected to do in five or ten years, which is certainly not what they are doing today? The general thinking today is, ‘Adopt AI or be left behind.’ But at the same time, companies need to reimagine the role of all levels of engineering and work out how to invest in their employees so they remain competitive after AI can do all of the easy stuff. Perhaps companies need to have a people architect, someone who can transform the development process and how humans and AI fit into it. The post Invest In Junior Engineers appeared first on Semiconductor Engineering.
Source: https://semiengineering.com/invest-in-junior-engineers/