The Electric Grid: A Strategic Conundrum
The adoption of AI will force us to rethink about collaboration & the open source software approach. Imminent hardware manufacturing & allotment discordance, geopolitics, questionable selective biddings for green projects & supply chain bottlenecks is only the tip of the iceberg we are facing.
The prospects of AI are a double edged sword (as with all technological breakthroughs - with great power comes great responsibility). On one side we have the opportunity to re-engineer the very networks our lives depend on, like power stations, and on the other hand those virtual applications generate inferences that require a ton of compute power & become massive energy guzzlers.
In this article i will analyze the dynamics of our grid' resilience, and the potential options for accomodating the massive incoming wave of AI applications.
A report written by the Linux Foundation Energy (LF Energy) this month suggests that putting emphasis on rapid digitalization with open source software - is the key to harnessing the maximum benefits AI can yield from the energy sector.
It's a massive understatement to say that the highly fractured environment of the electric grid contributes to a power efficiency slow down for meeting AI demands.
The electricity business is a state-by-state market, with different regions having varied economic & political opinions about what benefits the common good.
It is facing three primary uphill battles:
- Decentralization
A massive influx of green energy hardware like wind turbines, solar panels, huge batteries as well as the need to minimize the risk of central point of failure.
- Digitalization
The shortage of human operators in the energy sector to monitor & manage the grid.
- Decarbonization
Strategically minimizing the use of fossil fuels for energy generation.
For implementing advanced AI models...everything boils down to data, and in order to develop, train & refactor the designs that will tackle these problems - fast application development of complex tools is needed to filter and process massive amounts of engineering data feeds.
A lot of folks associate the open source term as a buzz word for non-proprietary software or "free" software...but its a much more strategic concept than it sounds.
Here are just a few simple examples of what an open source approach can offer us:
- Tust
Traditionally, utility companies resisted decentralization & documented their protocols in secrecy. An OSS approach will allow stakeholders to invest more efficiently in their companies by having instant access to data, algorithms & models and understand the datasets involved with their systems, giving them a substantial adtvantage on predictive analytics.
- Interoperability
Rapid information sharing will facilitate the integration of the grid with applications like self-driving EV's, Iot devices & AI Agents. This will allow communities and neighboring regions to maximize productivity and leverage resources.
- Collaborative Engineering
Developers, local companies & community leaders will be able to quickly share feedback on market needs.
As i'm writing this article on January 20th (inauguration day), news is circulating that Trump will declare a "national energy emergency" (no one seems to be quiet sure what exactly that entails) and mandate agressive extraction of fossil fiels, challenging Biden's vision of reducing the carbon footprint (who's administration just a few days earlier managed to secure $15 Billion for a California utility company).
That being said, multiple offices at the DOE haven't been ignorant about this debacle and have been actively involved in grid re-structuring analysis for many years, especially since the launch of their massive Connected Communities initiative in 2020.
In 2021 the DOE announced 61$ million in funding for 10 smart building projects.
In October of last year, 17$ million was invested in offshore energy projects and a few months earlier, $65 million was directed at Connected Communities 2.0 (to improve grid resilience as well as innovate charging technologies in communities).
Rant aside, to re-emphasize the urgency need for standardizing OSS as a building block for AI - here's an argument from a different angle: just a month ago the DOE released a report (authored by the Lawrence Berkeley National Laboratory) on electricity use by Data centers and it explicitly states that the lack of transparency as well as un-coordinated metrics are an impediment for scaling energy infrastructure.
If closed/unshared Data Center SOPs are an issue for unprecendented electricity consumption,..imagine what a lack of collaboration between power generation conglomerates can cause in the near future.
While open source is a reasonable approach, it also presents issues like isolated/inaccessible data engineering, hacking, and the massive shortage of qualified ML engineers to take on this feat.
This is where the government should step in and offer incentives for collaborative frameworks, training programs, and joint-force cybersecurity.