Now In Tech: AI, Assessments, And The Great Over-Correction

If you are craving some stability and have had enough adventures, then you are definitely not alone. Over the last 3 years, I’ve had enough of rollercoaster rides – metaphorically and literally. Seeing the post-COVID hiring frenzy descend into chaos has not been easy for me or others in the recruiting community.

With tech companies laying off more employees in Q1 ’23 (more than the entire 2022), good news has been hard to come by. For my first update in 2023 – a new endeavor I decided to take up this year – I would have liked the industry to be in a much better place than where it currently stands, but we’ll take the cards we’ve been dealt.

All of us in tech know that this downturn will change in a heartbeat as soon as the markets stabilize. What remains to be seen is if it’ll have a lasting impact on how we hire engineering talent, similar to the move to remote hiring post-pandemic.

There is a bright-eyed-bushy-tailed focus on productivity among the Valley people. Gone is the massive workforce that came to a workplace boasting three-course meals in the cafeteria and needed yoga rooms and sleeping pods to function. In many ways, tech is going back to the basics – show up, use your skills, and be productive.

Companies across the board are reviewing their team structures, looking for removable cushioning, cutting down on middle management, and wanting to do more with less. Public companies have been motivated by their stock performance and the rest are emboldened, seeing the giants lead the charge. Retrenching has been a buzzword around the valley. Tech leaders are relooking at their performance review processes and reformulating their hiring.

While some cutbacks were long overdue (for instance, Google more than doubled its workforce from 2018 to 2022), as an industry we might have been a little overzealous with this wave of austerity. Markets will bounce back, like they always do, and companies will go on a hiring spree to fuel the growth — but who we hire, and how we hire, would have changed.

Despite this recessionary period, the demand for specialized skills like data science, AI, and ML has remained stable. However, the hiring urgency we saw right after the pandemic is replaced by a muted, prolonged hiring process where the focus is not on filling roles faster but on hiring for the right skill sets. Developers reading this should be aware that the Great Over-Correction in tech companies also signals a death knell of the bargaining power they have held since COVID.

Talking of skills…

There are two important ongoing skill-related conversation starters in the tech world which have directly impacted HackerEarth. First is the emergence of ChatGPT and other generative AI and their use in candidate assessment tests.

I’m sure you’ve all seen the LinkedIn posts about how ChatGPT has been acing entrance tests and accumulating certificates. While the models are already very good at such structured tests, they will only get better with more data. Unless regulated, this may mean the end of the traditional fact/information-based testing methodology. That testing as a concept will have to evolve and completely move away from testing information retention and formula application to real problem-solving.

Generative AI, while extremely powerful, is still extremely poor at problem-solving. These models can efficiently solve well-defined problems but are incapable (at least for now) of solving real-world problems. Tech assessments are no different and there is a lot of concern around the use of ChatGPT in coding assessments in their current form. Particularly because a lot of companies have so far relied on complex algorithmic coding tasks as a measure of competency. This will get completely undermined by systems like ChatGPT.

One school of thought is that every software developer is going to use generative AI for coding in the near future, so it only makes sense to provide generative AI systems as part of the assessments to not only enable them to use it for problem-solving but in some cases to even test if they know how to effectively use a generative AI system. However, there is an alternative opinion that even though most of the code can be generated, to be a good software developer you must understand the fundamentals, and hence people should demonstrate that skill as part of the assessments. And since ChatGPT can undermine that, it comprises the assessments.

It is an ongoing debate, and as a facilitator of technical assessments, we at HackerEarth can see the argument from both ends. We have always aligned our assessment methodology to be aligned to how work gets done in real life, but we appreciate that an engineering manager would want to know how good a developer is sans AI assistance. So, instead of taking sides, we decided to support both personas.

We built a unique proctoring feature that creates a constrained environment for the test taker and hence blocks the use of not just ChatGPT, but any other support tool. Smart Browser, the new addition to our flagship Assessments product, is now live, and you can read more about it here. At the same time, we have embedded generative AI into our code editors. Like the Smart Browser, it’s an optional setting. When turned on, the test takers will be able to use generative AI right there in the test interface and answer their questions using such a system. We are also investing in questions that are better suited for situations where a developer has access to a generative AI technology while writing their code.

Okay, that’s enough about AI now!

The second skill-related conversation starter I referred to was the need to up- and re-skill tech teams. Upskilling programs have existed for a long while, and we all know how they have fared. I recently wrote a piece for Fast Company in which I went into great depth on why traditional upskilling initiatives do not work.

To recap:

– there is a significant gap in measuring ROI from current upskilling platforms

– there is a lack of social contract in these learning models, which hampers the 70% of upskilling that happens organically within a team

– these models lack an application-based learning pathway, so most of the time, course completion cannot be taken as a signifier of skill enhancement

I believe upskilling is integral to the skill-first tech ecosystem we are trying to build. Continuous learning not only helps engineers find a better pathway for career growth but also enables companies to address skill gaps and predict productivity outcomes. HackerEarth has always favored ROI-based learning pathways that do more than just help your engineers attain a certificate.

We have worked hard to create a platform that can merge real-world needs for developer upskilling with business outcomes, and we are close to completion. You can check out our Learning and Development platform here or write to me to learn more. This is a conversation I’d love to have with you!

And with that, this first quarterly update is out to print. I feel inclined to say something cliche like we’re all in the same boat and la di da, but I’m sure you all know that. Tech has lived in a unique bubble during COVID (that got built over the previous decade), but that bubble has now burst. The only way forward is through experimentation, exploration, and innovation, and that, as always, starts with who and how you hire.

Until next quarter,

Sachin