The DOD JEDI program provides a clear example of how not to do it
The list of priority career objectives for pioneers in artificial intelligence (AI) rarely include trillion dollar weapons programs, institutional turf battles, stifling bureaucracy, or systemic corruption — even for those who desperately want America and democracy to prevail. Highly intelligent scientists, engineers, entrepreneurs, and architects are typically motivated by solving the myriad of other challenges facing the planet, not least in saving the planet itself and each species within it, including our own.
It’s not surprising then that the academic culture has become a serious problem for national security. Indeed, some corporate and academic leaders in AI have taken actions directly counter to U.S. national security interests in favor of China, our largest adversary. With these and other challenges, the future leader in AI systems — which likely means economic leadership, may not be the U.S. Unfortunately, the problems from within are not restricted to academia or big tech.
To illustrate the severity of the challenge facing the U.S. and DOD, The Washington Post reports that a rare public meeting took place on JEDI just two weeks after the disclosure of the Capital One hack, which was a nearly perfect manifestation of the type of risk scenario we warned about in the JEDI construct, yet the senior staff at the Pentagon appear to be doubling down on the $10 billion contract with a single cloud vendor.
Our company KYield was never a contender for the JEDI contract, we have no formal relationships with any of the parties involved, and our distributed AI OS can be run on all of the common configurations. We therefore have no conflict other than a strong desire for customers and prospects to adopt network architecture that ensures a high level of security, performance and integrity. One of our specialties in AI is prevention of human-caused catastrophes, which quite often includes analyzing risk, system design, decision making, and unhealthy relationships that influence those decisions.
We have written many letters to senior leadership in the U.S. government warning of the threat from China in AI and emphasizing the need for speed, beginning in late 2017, including to Secretary Mattis and President Trump. However, we have also continuously warned about the risk in sprinting in the wrong direction over a cliff.
In several decades of researching AI systems, operational risks and systemic risks, with a special focus on prevention, I’ve rarely observed a more timely or potentially catastrophic red flag warning than the Capital One hack — nor any with such rich irony. The alleged hacker, Paige Thompson, was formerly employed as an engineer at Amazon, which is the expected winner of the JEDI contract. The prospect of the CIA, DOD, other federal and local government entities and the majority of the Fortune 500 all placing critical data on one cloud platform represents an extreme level of risk that is reckless and completely unnecessary.
Newly discovered vulnerabilities in cloud vendors are common. For example, a flaw in Azure was recently disclosed at the Black Hat security conference that left millions of users and researchers open to attack. Azure and AWS are the only remaining contenders for the JEDI contract — further evidence that relying on any single hosting vendor is imprudent. Although cloud hosting can provide a lower risk profile for some activities, it also introduces new types of risk due to the consolidation of datacenters in one vendor, including insider risk, physical attacks, and systemic cyber attacks.
This is why hybrid and multi-cloud configurations are the preferred architecture by most leading corporations. Hosting data in a carefully planned manner spread across corporate owned data centers, secure co-hosted facilities, public cloud and application providers will achieve the optimal risk profile in networked computing. A multi-cloud architecture allows organizations the ability to reduce catastrophic risk with precision management of information assets while improving redundancy and remaining operational in worst-case scenarios.
In addition to catastrophic security risks, overreliance on any single cloud provider also represents a considerable economic risk to the critically important U.S. technology sector. Although the two finalists in the JEDI contract each represent different challenges for the U.S. economy, they share commonalities with other big tech companies that are not in the best interest of the U.S., one of which is that ubiquitous offerings are available to anyone in the world, including adversaries. Commoditized computing not only provides no competitive advantage, it favors those with other advantages such as predatory pricing, market manipulation and scale, which plays directly to the strengths of China in particular.
Another commonality shared by several leading tech companies is enormous and rapidly growing market power over the U.S. economy. Amazon is aggressively using its massive scale in e-commerce operations and profit from AWS to expand into new industries, including grocery, healthcare, entertainment, media and now shipping. The $10 billion JEDI contract over a decade would provide an additional several billion dollars for Amazon to apply the power of network effects and subsidized pricing to compete unfairly with other industries.
In the case of Microsoft, the company not only still dominates the desktop and office suite markets, which has always contained risk — some of which has been realized, the company also owns the world’s largest business network in LinkedIn, the second largest search engine in the U.S. in Bing, and the second largest cloud hosting platform in Azure. Although other companies have recently received more antitrust scrutiny, Microsoft’s market power is without equal as reflected in its current valuation of over one trillion dollars — over $100 billion more than any other tech company as of August 14th, 2019.
When government agencies continue to support companies long after market dominance has been achieved, it sends a very powerful negative message to innovators, entrepreneurs, investors, and other market participants. The combination of these two companies headquartered a few miles apart with other tech giants in the consumer and enterprise markets is having a significant negative influence on the U.S. economy, even though the impacts are largely invisible to almost everyone except would-be competitors. The market barrier problem in U.S. technology is one of the reasons strong AI clusters are emerging in the U.K., Canada, China, Switzerland and Germany, among other nations.
The government seems completely unaware of their impact on economic dynamics in advanced technology, apparently having lost institutional memory of the government’s critical role in fostering competition during key generational changes in technology. No previous technology other than perhaps the atomic bomb is likely to be as important as the combination of AI and quantum computing in terms of national and global security.
An additional illustration of the risk in the JEDI strategy is apparent failure to understand that attempting to compete with China on scale is futile. China’s strengths are quite obviously scale combined with totalitarianism and market manipulation. America’s strength is in our diversity — we are successful despite monopolies and oligopolies. The future of technology will not be found in brute force computing, traditional scale metrics or through police state surveillance of citizens, but rather creativity and ingenuity at the confluence of entrepreneurship, synthetic computing, physics and mathematics.
As the DOD’s recent JEDI contracting process clearly demonstrates, the U.S. has a serious integrity and credibility problem with regard to how it sources and manages relationships. In order to project power and win future economic and military wars, the United States federal government needs to first look internally, embrace organizational excellence, act with integrity and fundamentally change its culture, decision-making and business processes. The DOD faces an urgent need to embrace innovators and their products based on merits versus scale, political influence, academic pedigree or any other irrelevant factor beyond national and economic security, which requires robust and dynamic diversification of our tech industry.
As has been the case in previous generational shifts in advanced technology, for the U.S. to continue to lead in advanced technology, which may determine our very survival, it will require a certain level of creative destruction that allows a new generation of leaders to emerge that offer state-of-the-art technology, much stronger integrity, and fewer conflicts of interest. Only then can we collectively provide the economic strength necessary for the U.S. to prosper. System integrity is absolutely essential.