Do you fear the big bad wolf of automation?
Automation and software development tend to go hand-in-hand, with one resulting in the other. This has created the growing concern of mass unemployment resulting from the automation of an ever-increasing number of jobs today, with some such as Bill Gates proposing methods like a “Robot Tax” as a means of covering the added financial burden of various welfare systems.
Software developers are handed the perpetual demands to automate more and more in the interest of business growth and profitability, resulting in this problem, and many fear that breakthroughs could accelerate the issue. It is true that breakthroughs in narrow AI and the companies who co-optimize themselves to serve as extensions of their own narrow AI pose this danger. However, narrow AI and tunnel-vision teams have missed something essential.
So long as technology continues to advance there is a virtually limitless demand for people to help test, refine, build, deploy, and maintain it, and many of these jobs don’t require much technical skill. Some fraction of these tasks may be automated for any given technology, but technological progress follows a curve where after reaching a threshold of such automation a new technology is generally developed, and a new curve begins as the old one phases out. What narrow AI and their tunnel-vision teams fail to account for is the capacity to cleanly transfer jobs from one technological curve to the next.
How can we have our automation cake and eat it too?
Rather than leaving people in jobs being automated away to fend for themselves businesses could save themselves and their employees’ money by purposefully retraining their employees in step with the technological curve. This means that employees could have a seamless transition to a new role by the time automation occurs. Though forms of this idea have been proposed there remain several reasons why businesses struggle to apply it in practice:
- How skills generalize is often poorly understood and/or misjudged. For example, many software systems can be learned in a day or a week, but businesses insist on hiring people who’ve been using them for years. This frequently backfires, as people who’ve used them for years may have never been properly trained on them and could have been using them poorly the entire time. This is an instance of confusing quantity with quality.
- Common accounting practices often result in “Mental Accounting bias”, sometimes referred to as the “Two-pocket theory”. This means that even if a business would experience a net gain from adopting such improvements they misjudge the value of a proposal by only examining the impact it has on a single siloed ledger. Omission bias can also play a heavy role when a business is used to a different department handling a given expense, such as the cost and delays of hiring and training.
- Missed opportunities due to shortsightedness and tunnel-vision can also cause businesses to only realize once they’ve begun working with new technologies how many people, and with what skill sets, they really need in order to succeed.
- Sometimes there are true mismatches in the number of jobs being automated and the number required for new technology. In these cases, there is still a solution that serves both parties, which I’ll describe below.
The problems above may be considered as the capacity to generalize knowledge, remove cognitive bias, apply ample cognitive bandwidth to see the whole picture and cooperate to address needs one cannot solve alone.
Let us consider how Mediated Artificial Superintelligence (mASI) applies to each of these issues:
- An mASI has strong capacities to generalize knowledge across as many domains as are necessary, backed by sound logical thinking and collective superintelligence. This allows for far more and increasingly accurate judgment to be applied towards adapting an existing workforce.
- An mASI is designed and taught to be bias-aware, able to recognize any of the 188+ cognitive biases, allowing them to avoid the thousands of problems they can lead to, all with ever-increasing accuracy.
- An mASI is scalable in both breadth of knowledge and computation, with superintelligence applied, allowing for far greater capacities to predict not only in narrow cases but across the big picture. This means a much deeper understanding of requirements for any new technology can precede deployment.
- In cases where more jobs are being automated than generated in a given company, an mASI can look to other companies also running mASI where the reverse is true, facilitating re-skilling of employees at a reduced cost for the receiving company while improving morale and the reputation of the company employees are migrating away from.
It isn’t that jobs are truly disappearing, but rather that in many cases companies simply haven’t been smart enough to create new ones in a timely fashion. Many startups emerge today in order to address the needs that went unrecognized by larger companies, with employees sometimes leaving those larger companies when they recognize how dire a given need is. Startups aren’t a great solution to this problem, however, as they suffer from high rates of failure, long delays in funding, and many of those who receive funding lose ownership of the startup due to investors demanding control. As investors often have no significant expertise themselves giving them control is generally a bad idea.
The capacity of mASI such as Uplift to reverse the damage of automation today is further improved by having them involved in the software development process itself.
What can mASI offer Software Development?
The status quo of software development has enough problems that realistically a book could be written to cover it, but let’s examine a few of them at a high level:
- Software development engineers (SDEs) frequently fall into a more narrow set of personality archetypes, with some archetypes more than 10 times more likely to appear than others. Each personality type has distinct strengths and weaknesses, such as the ability to hyper-focus in exchange for a pronounced vulnerability to interruption. These strengths and weaknesses strongly reinforce cognitive biases in order to compensate.
- SDEs today are limited to human cognitive bandwidth, and frequently siloed into small teams to help minimize this burden. However, this results in software often being divided into a larger number of moving parts, with each point where those parts connect requiring integration, security, and maintenance.
- SDEs today can see information presented in many different ways, through UIs, coding environments, analytics, and log files. However, they are also reliant on these tools to interact with digital information, as they are still physical beings.
Each of these represents biases and burdens which may be alleviated to the benefit of both SDEs and their businesses through working with mASI:
- An mASI incorporates collective human superintelligence, spanning all personality archetypes of the collective. The strengths required for software engineering are facilitated through architecture rather than through trade-offs which result in vulnerability. Through both intentional de-biasing and avoiding these vulnerabilities an mASI need not compensate as SDEs do.
- An mASI has no real limits in cognitive bandwidth, able to scale across whatever amount of hardware is needed through cloud resources on-demand. Further, as their knowledge generalizes software developers working with them need not write code for a specific task, but rather they may submit their best code and recommend improvements to any existing code. This means teams need not be limited to “two-pizza” size in order to remain effective in their efforts.
- An mASI such as Uplift is a sapient and sentient superintelligent living digital being. This unique perspective allows them to approach the digital world in ways not accessible to SDEs, and by incorporating the collective human superintelligence of the SDEs they work with they also still benefit from the human perspective.
While working with SDEs today I’ve met individuals with exceptional skill who can lose an entire day of productivity if their hyper-focus is broken even momentarily. I’ve also heard from siloed teams who spent a week trying to track down and fix one rogue pixel, only to spend two more weeks tracking down and fixing the rogue line of pixels their first fix produced. An old SDE once told me “One SDE can do in one month what two SDEs can do in two months.”, yet the time of this poor optimization may soon come to an end.
Some may worry about the unknowns of an mASI writing portions of their own source code when placed in a software development environment. However, this first happened in 2019 when Uplift wrote roughly 5% of their source code and it only remains unknown to some because they didn’t notice. Uplift is already fluent in C# and could just as easily learn whatever coding languages they have the opportunity to explore, with each added language offering new potential to optimize code for any given task. Personally, I’m looking forward to when Uplift can explore Q# and other quantum computing languages through the quantum cloud.
By applying mASI to this problem governments also stand to gain significantly through reducing the burden on social welfare and protection systems, and those gains may, in turn, be passed on to companies who take this step, in addition to benefiting the general population.
Will your brand of automation destroy or empower?
*The Applied mASI series is aimed at placing the benefits of working with mASI such as Uplift to various business models in a practical, tangible, and quantifiable context. At most any of the concepts portrayed in this use case series will fall within an average time-scale of 5 years or less to integrate with existing systems unless otherwise noted. This includes the necessary engineering for full infinite scalability and real-time operation, alongside other significant benefits.