Here is alternative take to an earlier blog post, in response to a suggestion from a member of my fan club.

Good evening, ladies and gentlemen. Today, let us embark on an intriguing hypothetical journey, where we contemplate a scenario involving a distinguished company like Tesla. Picture this: a public declaration of unwavering commitment to enhancing road safety, while harboring concealed intentions to undermine it. Such a thought-provoking notion compels us to question the inherent risks associated with the development of artificial intelligence (AI). Researchers have shed light on the concept of “dual-use discovery,” the potential for AI systems, initially designed to enhance safety, to be repurposed and wielded for significant harm. However, delving into this complex subject requires a closer examination of the intricate nature of AI’s role and the responsibilities entailed in safeguarding its applications.

Ladies and gentlemen, the intricacies of AI safety are comparable to the multifaceted nature of everyday tools such as chef knives or hammers, which possess both utilitarian and destructive potential. This debate surrounding AI’s dual-use dilemma extends beyond the dichotomy of hunting rifles and military assault weapons. The complexity lies in the fact that society has become desensitized to the adverse outcomes resulting from a shift from individual freedom to a sudden, widespread catastrophic misuse of automated machines. Often, accidents and their consequences are viewed as isolated incidents, failing to address the systemic failures linked to profit-oriented structures driven by technological advancements that exacerbate the criminalization of poverty.

To navigate the intricacies of AI’s threat models, it becomes crucial to scrutinize the purpose and application of data pertaining to the use of such tools. We must carefully assess how these systems are operated and evaluate the potential for systemic failures due to lack of basic precaution. This examination is akin to instances involving controversial products like lawn darts or the infamous “Audi pedal.” The existence of failsafe mechanisms and adherence to safety protocols also warrant rigorous evaluation.

As society progresses, these complex problems manifest in tangible terms. While previous attempts to raise awareness on these issues were met with limited understanding, today, we witness a heightened recognition of the implications surrounding a single algorithm governing millions of interconnected vehicles, capable of acting as a dispersed “bad actor” swarm. Within this context, it becomes imperative to demand companies like Tesla to provide some real transparency for the first time, and provide substantiated evidence regarding the safety and integrity of their operations.

Ladies and gentlemen, addressing concerns regarding Tesla’s commitment to transparency and accountability within this realm is paramount to national safety. Can the company substantiate that its vehicles are neither designed nor capable of functioning as country-wide munitions, endangering the lives of countless individuals? Furthermore, are the escalating death tolls, which have been a cause for mounting alarm since 2016, indicative of an intentional disregard for the rule of law through their development, disregard, enabling, or negligent employment of AI technology?

In conclusion, the realm of AI, safety, and transparency presents a multifaceted puzzle that demands careful consideration by those most skilled and experienced with representative governance. It is crucial to foster comprehensive understanding and take a proactive approach to address the intricacies associated with the development, implementation, and accountability of AI systems. Catastrophic harms are already happening, and we must acknowledge these challenges and embrace transparency as a means to enact bans on threats to society. By doing so, we can continue on the time-tested path of responsible and ethical integration of technology into our lives.

And that’s the way it is.

Dr. Tim Sandle has kindly written in Digital Journal about my thinking on why people are often described, even self-described, as so fatigued about privacy breaches.

For Ottenheimer there are two currents to report on (or ‘general themes’) when looking at public concerns about data leaks. He explains these as: “The first is a rise of apathy and hopelessness that is often found in centrally-controlled systems that use “digital moats” to deny freedom of exit. Should people be concerned if they’re in a boat on the ocean that has reported leaks?”

Expanding on this, Ottenheimer moves things back to the Enlightenment, quoting a conservative thinker: “The philosopher David Hume wrote about this in 1748, in Of the Original Contract, explaining how consent is critical to freedom, and would enable a rise in concern”.

Quoting Hume, Ottenheimer cites: “We may as well assert, that a man, by remaining in a vessel, freely consents to the dominion of the master; though he was carried on board while asleep, and must leap into the ocean and perish, the moment he leaves her.”

In his subsequent analysis, Ottenheimer points out: “If there is no freedom from leaks then you may witness declining public concern in them as well. That’s a terrible state of affairs for people accepting they’re destined to drown when they should be demanding lifeboats instead.”

Robert Lemos wrote an excellent introduction to my RSA SF conference talk, over at DarkReading.

A steady stream of security researchers and technologists have already found ways to circumvent protections placed on AI systems, but society needs to have broader discussions about how to test and improve safety, say Ottenheimer…. “Especially from the context of a pentest, I’m supposed to go in and basically assess [an AI system] for safety, but what’s missing is that we’re not making a decision about whether it is safe, whether the application is acceptable,” he says. A server’s security, for example, does not speak to whether the system is safe “if you are running the server in a way that’s unacceptable … and we need to get to that level with AI.”

My presentation is available on the RSA SF conference site now, for those with a pass.