The promises of artificial intelligence appear infinite. In one incarnation or another, AI has been touted as a panacea for everything from curing cancer to solving world hunger. While there is valuable research aimed at these audacious goals, there has been very little practical deployment of artificial intelligence. To date, the most widely used applications of AI are helping recommend additional products for 1-hr delivery, auto-tagging pictures of friends, or adding cat-ears to a selfie. Each of these are clearly strong technical innovations, and have created significant enterprise value for their respective shareholders. But they are not solving the problems that so many are hoping will be resolved with this new wave of artificial intelligence capabilities.

So what is holding back the true expansion of AI into applications that will change the way humans live? Until now, it has been primarily a sense and compute problem — the human eyes and brain are difficult to replicate. In order shift control of the physical world to machines (such as in a self driving cars), they need to understand their environment, decide on the most appropriate course of action, and then perform that action — exactly what the eyes and brain do. Many groups are getting close to replicating such anthropoid functions through innovations in technologies such as lidar, radar, computer vision, and novel compute-architectures. These developments have enabled advancements in the field, such as the launch of Google’s first truly-driverless-car, and Arterys’ approval to use their deep learning software to support diagnosis of heart conditions.

While the sense and compute problems are not fully solved, these innovations have reduced this dimension of the problem from a key bottle neck to an “area in need of improvement.” There is a more fundamental reason that AI is not yet fully adopted: security, or rather, the lack thereof.

The global spend on cyber security is estimated to be $100B to $200B (somewhere between the GDP of Morocco and Portugal) every year. With this tremendous investment, one might imagine that our software networks and applications are extremely secure against hackers. That is, of course, unless one is a customer of one of the multiple enterprises that have recently been compromisedevery year. With this tremendous investment, one might imagine that our software networks and applications are extremely secure against hackers. That is, of course, unless one is a customer of one of the multiple enterprises that have recently been compromised. There is clearly a lack of true security, regardless of the dollars spent. In these markets, customers are willing to accept the inherent lack of security for the convenience afforded by using credit cards and having personal information stored by retailers.

This tradeoff completely changes as artificial intelligence enters the physical world. While customers may accept the loss of money as a result of their credit card getting hacked, no one will accept the loss of their life as a result of their self-driving car being comprised. The same holds true for attacks on robotic surgeons and AI pathologists. When AI enters the physical world, the risk-benefit calculus fundamentally changes.

A recent wave of hardware vulnerabilities exposed many of our devices to potential attacks. The Broadpwn attack gave hackers direct access to over a billion cell phones. An entire class of processors is susceptible to malware hacks due to a hardware weakness. Luckily these issues have not yet been exploited by hackers in an extended fashion, but it is only a matter of time.

Protecting against these kinds of attacks is difficult. It requires a completely different type of thinking — one that is full-stack in nature that starts at the hardware, where no software can protect against hacks in the wild. Today, I’m proud to announce that Eclipse Ventures has funded Tortuga Logic’s seed round as they embark on a mission to do exactly that.

Tortuga Logic is building products and tools to identify and prevent security vulnerabilities across hardware ecosystems, from low-level silicon to the data center, and from the semiconductor fab to the IOT endpoint. Now is the time for broad adoption of Tortuga Logic’s solutions due to the convergence of three key macro trends:

1. Technological readiness of AI systems to control the physical world

2. An exponential increase in the number of custom / new computer chips

3. Convergence of hardware and software into full-stack, highly coupled solutions.

The founders of Tortuga, Jason, Jonathan, Tim, and Ryan, have been living and breathing hardware security for the last decade. They have now translated the results of their research into Tortuga’s products, enabling the level of security required for the next generation of AI to truly transform our world.

Follow Eclipse Ventures on LinkedIn and Twitter for the latest on the Industrial Evolution.