Insider Perspectives

By Contributed Article | January 15, 2018

Could Artificial Intelligence Spell Trouble for the Connector Industry?

AI has the potential to change everything about how connectors are made — and who makes them.

By Ray Alderman

It’s time that we take a broad and pragmatic view of the future, what it holds for the electronics industry, and specifically what difficulties could be in store for the connector industry. The primary mechanism for change today is artificial intelligence (AI), not new processors, faster interconnects, better sensors, or new memory chips. Those components are all minor elements in the big picture. The real question is, how will we be affected by the numerous and nefarious technological, legal, competitive, and standards development problems that AI will inevitably spawn? Change is accelerating at an increasing rate, and the connector industry is not immune.

The connector industry operates on four basic principles: 

  1. The connector makers know more about connectors than their customers.
  2. Only the connector makers have the skills and equipment to produce the devices in volume.
  3. Their patent portfolios protect their business models.
  4. Connector makers have shown an amazing historical ability to avoid liabilities for harmful failures of machines that use their products.

Now let’s take a closer look at these principles, while also acknowledging the real threats AI poses. In a speech to schoolchildren, Russian president Vladimir Putin said, “Artificial intelligence is the future not only of Russia but of all of mankind. Whoever becomes the leader in this sphere will become the ruler of the world.” He was talking mostly about autonomous weapons and killer robots. But could Putin’s statement also apply to the connector industry? Let’s explore how that possibility could take apart the very foundation of the industry.

Data is Power

This article is offered as a raging torrent of ideas, mingled with rivulets, riptides, and whirlpools of speculation, cascading into a waterfall of problematic possibilities.

Connector makers have spent a lot of money developing simulation models for their high-speed copper connectors, characterizing their performance parameters and eye-diagrams related to certain chips (e.g., PCI Express, Ethernet) and transmission lines (e.g., cables, backplanes, PCB traces). They share those with the standards committees, who decide which connector to specify in their documents. The fundamental premise of this industry is that connector makers know more about connectors than their customers. AI could endanger that assumption.

AI is not about who has the best algorithms. It’s about who has the most data. AI systems learn, and that takes huge amounts of training data. And, the leading high-volume systems companies probably have the most data about connectors, relative to operational performance and reliability over time. The connector developers sell the system developers a connector and move on to the next customer. The system vendor then builds thousands or millions of systems, and with that experience curve, they have a good perspective of cost, performance, and reliability. From that data, the system developer’s AI computer-aided design (CAD) system knows the design principles for a better connector. They just don’t have the ability to build them in volume.

How do connector makers get data from the systems vendors? They can spend a ton of money on AI-CAD tools and servers, and offer that as a service to their customers in lieu of the customer having to buying their own AI-CAD tools and servers. That way, they get the data for free and they get money from the customer for using the tools. This process is called software as a service or SaaS. The connector vendor’s tools also collect and learn from the data generated by numerous customers in a process that’s called the network effect. System vendors know their data is worth a lot of money, and the connector design elements that it creates will give them a competitive advantage in their markets. Why would they share their data, indirectly, with one of their competitors who might be using the same connector vendor’s AI-CAD tools? Personally, I don’t think the SaaS model will work for connector makers, unless they marry their manufacturing expertise data with the customer’s operational and reliability data, and keep it compartmentalized.

As AI-CAD matures in stages, customers may have less incentive to share their data sets with connector vendors. You can probably see where we’re headed here: At some point, data sets could be more valuable than patents, and the patent-based business model of the connector industry could be threatened. It’s also not implausible to take this idea one step further and assume that AI-CAD could overwhelm the patent system.

Now that system vendors have better data than connector makers about how to design better connectors, what will they do with the output of their AI-CAD machine? When 3D printers are more capable, they’ll hook that system up to a 3D printer and make their own connectors. AI-CAD and 3D printers could do to connector makers what Amazon and the web did to brick-and-mortar retailers and shopping malls.

Slipping Standards

Let’s get something out of the way here: I am not a lawyer, nor have I played one on TV. The information offered here is not intended to convey or constitute legal advice or legal opinion. I am just asking questions and posing some scenarios. You should contact a qualified attorney about these topics if you have questions or concerns.

There are numerous articles about how AI will affect the economy, society, healthcare, the financial industry, and employment on the web. But I could not find one article about how AI could affect the standards development community. I found no position papers or perspectives from ANSI or other major standards development organizations (SDOs). That’s disappointing and disconcerting. So, let’s take a whack at it here.

Go to any standards developer meeting, and you can’t swing a dead cat without hitting a dozen engineers from connector companies. Connector companies live on the business created by open standards hardware used in telecom, military, industrial, transportation, and medical markets. They protect that business with their intellectual property (i.e., patents).

Let’s say that a company has a good AI-CAD system and participates in the standards development process. Their people come back from a standards meeting with the scope and the basic parameters of a new standard. They plug that information into their system, which can predict the different branches that the essential elements of the standard must take. Their AI-CAD system then spits-out patent applications for each of the essential elements, in each of those possible branches. This is not new. For decades, the humans in standards meetings have been anticipating the path of the essential elements and applying for patents. AI-CAD will speed things up, potentially allowing participating companies to loot, plunder, and undermine the standards process.

Since September 2011, the US has been on a first-to-file patent system. Each time an essential element is incorporated into a standards document, several connector vendors could declare patents (or patent applications) at the meeting. There’s a lot of discussion about how the new patent law is actually a first-to-publish instead of first-to-file system, involving grace periods and subsequent disclosure not being considered “prior art,” etc. This process could go on for months or years, depending on the complexity of the document and the legal processes. AI-CAD has the ability to exacerbate the present patents-in-standards issues for SDOs and participating companies.

SDOs could subscribe to an AI system in the cloud, on a pay-per-use basis. They could upload a new standard’s scope and parameters, and those algorithms could also identify the branch possibilities and essential elements. In other words, an AI system could actually write the standard. It could also identify all the relevant standard essential patents (SEPs) and avoid them by optimizing the path through the possible branches — except for the patent applications, which are not publicly disclosed for 18 months. Such a process could irritate participating members and might create some legal problems.

Members of an SDO could also connect their AI-CAD systems together in a collective agreement, and create their own standards. They could bypass SDOs and consortia altogether. Hooking the SDO’s AI machines to its member’s AI machines (i.e. utilizing the network effect) could create some interesting legal, technological, societal, and economic creatures that we have never seen before. Could the DOJ mandate that their AI machines — programmed for the legal aspects — be connected to any and all SDO networks to monitor them? We have already proven that AI machines can think faster than engineers, so it’s even more true that they can think faster than lawyers. That prospect could make the laws dynamic, changed on the fly by the legal AI machines, as it discovers what SDOs and their member’s machines are doing collaboratively. Lawyers are the beavers in today’s rule-of-law society. To some degree, their mission in life is to get in the middle of a cutting-edge stream of innovation, technological development, and experimentation, and dam it up. Try explaining to a judge and a jury what some connected AI machines were doing that was legally questionable and watch their brains explode. Mixing AI-CAD and the law is like mixing oil and water.

Slaughter-bots and Liabilities

Take a look at this video about totally autonomous slaughter-bots, driven by sophisticated AI algorithms. Future military forces and law enforcement agencies could utilize these new tools. Now assume that the devices kill some innocent bystanders during their mission. That could initiate civil and criminal actions. But who would get sued or charged? The agency that launched the bots? The maker of the bots? The person or company that wrote the algorithms? The owner of the data set used to train the bots? All the companies who supplied the components inside the bots, including the company who made the connector that passed the data from the sensors to the processor? The answer to this question is probably all of the above.

At a lower level, if an SDO’s AI system does a patent search, maps-out the optimal path through the branches that a potential standard might take, and misses some patents or applications, does the SDO now have liabilities to implementers who relied on that search to protect them from patent infringement claims? Like the dearth of papers about AI effects on the standards development organizations, there’s also a paucity of papers from the legal community about potential AI liabilities and complications.

Employment

Designing a connector is a bounded problem. We are bound by the three-dimensional space in which the connector fits. We are bound by the conductive and insulating materials used for the shell and contacts. We are bound by the laws of physics and transmission line effects on copper and optical fibers. Solving bounded problems is the fastest path to revenue, and that’s why thousands programmers are working on them, especially in the AI-CAD arena.

While it takes human engineers months to design a new product, an AI-CAD system with extensive training data can do it in a few minutes or hours. Then, the AI system can run simulations to verify the design in minutes. No need to build several prototypes and go through multiple expensive testing phases and design revisions. Could we see job losses for engineers in the connector, semiconductor, board, and systems markets in the future? Yes. We just don’t know to what extent.

Conclusion

Why did Amazon buy Whole Foods and get into the grocery business instead of buying a connector company and getting into the connector business? The grocery industry brings in about $16 billion per day in sales in the US alone. The connector industry sells about $164 million per day worldwide. There’s no big money in the connector business and, for Amazon, it’s not worth disrupting. But it is worth disrupting for the small CAD product companies. Amazon has over two million servers installed in their data centers today. Their people already design their own server boards, and have them built by contract manufacturers in Asia. If there is any company that has a large data set on cost, performance, and reliability of connectors, it’s Amazon, or Google, or Facebook. Could they run that data through their AI systems and come up with better connector designs and have them made in China? To some degree, they have probably done that already.

Ray Kurzweil, Stephen Hawking, Elon Musk, and other futurists have commented that AI will take us to a singularity, the point at which machines become more intelligent than humans. They predict this event will occur around 2045. But they miss the point.

The true singularity will be when AI-CAD confounds the rule of law, overwhelms the standards development process and the patent system, and shifts the balance of power to the customers with big data sets and away from the connector makers. That could happen before 2045.

We’ve taken a cursory look at only a few elements on this topic here, and we took some liberties to speculate about them. The effects of AI-CAD on our industry might be more significant than this article proposes. If I had a decent training data set, and access to a good predictive AI algorithm, I could show you a tree-diagram of the possible branches for what could happen, calculate the probabilities for each branch, identify the primary-secondary-tertiary elements in each branch, and demonstrate the effects on this industry in detail.

But first, we have to admit that even in pondering this, AI may already have the answers. While a human can manipulate a few variables in a static spreadsheet per hour to answer “what if” questions, an AI algorithm can manipulate hundreds or thousands of variables, millions of times in just a few seconds. While we can only speculate, AI may already know how this story will end.

Ray Alderman is chairman of the Board of Directors at VITA. He was in military intelligence in the Vietnam War; started in the mainframe computer business with Burroughs; was a partner in several computer start-ups; president of PEP Modular Computers; and enjoys irritating connector vendors as a hobby. Please feel free to harass him at [email protected].

 

Subscribe-g80

 

Sign Up for Updates

 

Recently posted:

[related_posts limit=”10″]

Get the Latest News
x