Mark Zuckerberg recently came out saying that he believes GenAI systems will replace intermediate developers this year. ChatGPT has progressed quickly, in fact it's passed the Turing test and found to be more collaborative than the average human. Many businesses have remote teams and developers, one has to wonder how working with an AI dev agent might translate. These agents will increasingly be making more and more decisions, and as we get comfortable that will increase with time. Given all of this you might be wondering if that's possible, should I be changing my outlook?
Let's review his statement from a few perspectives: the broader market, cost, current state of the technology, what you should be concerned about, and how you can benefit from it.
Market Trends
While Big Tech companies dominate the news cycle, the reasons behind these kinds of statements need to be analyzed. Let's peel the layers back here a bit.
There's a new administration in the US, with many Big Tech companies feeling the pressure to endear themselves with the upcoming administration. We've seen Facebook remove features such as fact checking due to potential concerns. Zuckerberg probably is concerned to some degree about retaliation and would want to adjust head count.
Zuckerberg is the CEO of a publicly traded company. Rather than admit demand for new features may be waning, it sounds better to shareholders to state that AI will take over jobs. Several years ago, Cisco, a company primarily focused on networking hardware, claimed that AI had reduced the need for jobs; I find it hard to believe they had suddenly developed the ability overnight. The better explanation would be they over-hired during 2021 and wanted to adjust headcount.
Meta is likely looking to reduce headcount. Looking at their layoff patterns, there was a major set in 2022, and 2023; they're likely overdue and trying to use this to build positive momentum with shareholders than paint a picture of regulatory pressures and demand drop off.
The point here being I take what Mark says with a grain of salt. Tech "bros" also have a history of overpromising and underdelivering - we've seen this pattern from Elon, Sam Altman, and many others in the industry. Driving hype behind something translates to more investments.
On the other hand, if we look at where investors are putting money in startups, it favors those companies with AI, the numbers seem to back this story. Bloomberg reported that AI Startups funding was at a record $97 Billion.
It is impacting the market to the point that .ai has become one of the leading domains. I've across my fair share of startups that have nothing to do with AI but use the domain as it helps from investors and clients.
There is undoubtedly an intrigue of what's possible with AI, our history as a species means we associate communication with intelligence. The fact these systems can communicate means we tend to trust what is stated.
Cost
There has been a consistent narrative that AI tooling will reduce in cost over time as it reaches market saturation similar to cloud services. I don't believe this to be true - ChatGPT recently introduced a new Pro plan costing $200/month, a 10x increase from their previous plan. I would argue this makes it less accessible.
Looking at historical context of other tech companies in this space, there tends to be an enshitification tax where the price will increase will quality decreases. Once a tech company reaches market saturation, we see the pattern occur. Netflix is a great example of this, when it launched the business model was unique for the industry and convenient for consumers. Over time the price has continued to increase and the number of shows/movies have moved to other services. It's now a worse landscape than just cable, granted lots of convenience but at a considerable cost.
With GenAI systems, it's more complicated. Every GPT that is released is magnitudes larger than the previous - for perspective GPT-2 was trained on 10 billion tokens, GPT-3 on 300 billion tokens. As such the costs of training go up, not just from the compute needed but as underlying energy for data centers increase that gets passed on. GPT-3 175B model required 3.14E23 FLOPS of computing for training, it meant millions of dollars for a single training run.
While newer chips are more capable, and quantum compute would change this discussion considerably, there is still the cost of energy. We are an energy hungry society; energy supply can't keep up with the demands. It's easy to infer the cost of such systems will increase.
Another perspective is the cost of historical conversation context it requires to run LLMs, in memory is expensive much of it needs to move into databases. To retrieve this context for new conversations is an overhead. While having a single omnipotent LLM would be awesome as it could do all tasks it becomes too expensive. It's a big reason why agentic models are being pushed, it's cheaper to have systems targeting individual domains. This introduces an overhead that exists with current engineering teams - collaboration!
How The Market Will Deviate
Now that we understand why an agentic model is necessary, what does that mean from an engineering perspective? Software development requires multi-dimensional complexity and understanding of multiple domains. Arguably the complexity of delivering products lies in the collaboration of people with different backgrounds.
Autonomous transformations will result in structures based on business focus. Tech focused businesses will invest more resources into development of AI, at least in the short term. For tech-enabled businesses (these tend to be incumbents), they will be some of the large-scale adopters of these technologies.
Translating To Teams
Agentic models will require some form of collaboration, we will see new metrics beyond DORA and SPACE to understand how efficient these agents are. Will it translate to teams of AI agents replacing developers?
Truth is I'm not entirely sure. As the price of energy continues to rise and demand increases for agents, it may become a struggle between supply and demand. It's more likely that we will see engineers deliver better quality and increase speed of development. From a business perspective, the choice becomes to increase features and explore new opportunities or focus on reducing costs.
Will development jobs be impacted? Yes, they will; another way to look at it is development will be within reach for lots of people. Your domain knowledge for the business around the industry will be more valuable. As development will be easier to access, I suspect the salaries will drop similar to what we've seen it happen in other industries where automation has taken over. Eventually, there will be a mix of human and AI developers.
Leveling The Playing Field
The struggle between incumbents (traditional tech-enabled) and disruptors (tech focused business) is going across industries. Disruptors typically have singular focus, lots of investor attention sometimes translating to significant funds, & less technical debt from decades of decisions. It also means they can often give software engineers, and their tech teams perks and salaries that incumbents often can't compete with.
If we look at banks versus fintech companies, we see this pattern. Banks have brick & mortar locations, need to pay employees related to customer service whereas most fintech companies tend to operate online only, typically have fewer regulations that they need to be concerned about.
What am I trying to get at here? I believe this skills gap between disruptors and incumbents is going to reduce due to AI tooling. Imagine having a junior engineer with Github Copilot now producing code at mid-level. It's then possible to build a world class team without the world class budget.
Everyone can have access to pair programming, and the arguments of wasting engineering resources are less of a concern.
How You Should Prepare
I had put a previous article on why I didn't believe developers were going away anytime soon. The recommendations I made at that time still remain relevant, so I am repasting them here:
Pair Programming 2.0 - Pair programming allowed developers to collaborate, reduce defects and build higher quality code. While there have been plenty of tools over the ages to support refactoring; CoPilot is a step better. If the developer begins navigator role and lets CoPilot be the driver results tend to excellent. We can then produce higher quality code, aligned with business needs.
Support with non-development engineering tasks - everything from creating QA test cases to developing user stories for product backlogs to creating documentation & even brainstorming, engineering teams will benefit from utilizing GenAI. It could offset tasks that can often be barriers for completion, additional it could be trained to produce outputs from your team templates.
Security Systems - This isn't GenAI specific, but broader. One of the issues with detection and incident management I often see with teams is that they don't have the resources to dedicate towards reviewing logs - it is too much data. Even with tools such as Splunk the organization can be very poor. Pattern recognition is definitely a strength, pair that with GenAI we could have level 1 support providing significantly more insights than previously. I'm just scratching the surface here.
Team sizes - It would be a lie to say team sizes wouldn't be impacted, you might not need to hire as quickly to grow a team. As the average output of a developer increases, we should be able to get more done with less. This is entirely dependent on the goals of the organization and timelines for releases
Thoughts? How do you feel you will be impacted by ai? Comment Below
I bet it's not long before they replace nearly all developers jobs for regular companies. Perhaps not for the bleeding edge tech, but they would greatly benefit from integrating with AI.
I bet it's not long before they replace nearly all developers jobs for regular companies. Perhaps not for the bleeding edge tech, but they would greatly benefit from integrating with AI.