The Darwin's Law of Evolution for computers

We’re leaving in the ‘hype’ era, where everything is sold as the Holy Grail for a non-real problem. Let’s try to analyze things with a critical eye.

Pushed by a friend, who is writing a couple of very good posts I recommend, I’m now sharing my thoughts about automation, the future of network engineers, AI, and so on. Of course, this is the point of view of my mind, as a result of 20 years playing in the IT world at both work and home.

It seems to me that vendors are trying to push more and more cool words, at a higher rate compared to the past. A few years ago the “IT outsourcing” mantra lasted for a while, and everyone had to “outsource” something to become a good manager. After a while, it was the “cloud” era, but it was too early, so it was a false start. In the last two years, we have seen many “mantras” coming together: big data, cloud (again, but with a different awareness), Cyber security, automation, big data, and machine learning. They are too many mantras! Not so many years ago, the idea of the cloud could be compared to a startup, and that startup was selling something not ready for the enterprise market. In the last few years, thanks to Amazon, Google, Microsoft, Salesforce, and so on, enterprises and IT managers can now approach a valid cloud strategy with more awareness.

Lesson learned: better if we wait for the second generation to have a mature technology.

I would add another one:

Lesson learned: successful startup aims to be bought by big companies. Buying cloud services from small companies can lead to unexpected services/license changes.

Now let’s apply the above thesis to the current mantras, especially to: automation, machine learning, and Cyber security.

I guess I can affirm that networking vendors are including automation on every product. So we have REST APIs to program every device and UI (user interfaces) that should help to manage the whole infrastructure with few clicks and without the knowledge of what is below. But again the real world is different than what is told by marketing: a lot of bugs that make software unusable or dangerous.

Moreover asking for support from vendors gives fewer results than a walk to Santiago de Compostela or Lourdes.

Adecco India at Cisco

It seems to me that vendors are not interested in selling good software, everything should be developed and sold “agile” (no matter if the agile methodology is a completely different thing). And the most critical aspect is that the quality of support teams is terrible: a lot of FTE is spent for debugging what should be analyzed by the vendor, a lot of “it’s by design” embarrassing limits, and really, too many untrained people unable to support anything or anyone.

Lesson learned: vendors cannot write good software, and apparently, they don’t care too much about that. They are only interested in selling ideas, not real and working features.

Let’s move now on to machine learning, the “hype” that will make humans useless. Computers can learn and detect things better than humans and will be able to analyze a lot of data in a relatively short time. So big data is not a problem anymore. But… because there is always a “but”, the real world is a little bit different. If someone is using Facebook to guess sexual orientation, we should ask ourselves how much that research is “marketing” and how much is “real”. And the answer is that machine learning, used to detect if a photographed man is gay, is slightly more efficient than an untrained human. So current machine learning algorithms perform better than thousands of untrained people. And that’s still a good thing, but we should remember that machine learning algorithms don’t (currently) perform like experts.

Lesson learned: (current) machine learning algorithms slightly perform better than untrained humans, but they can manage a lot of data and much faster than humans.

But machine learning applied to automation and Cyber security will save the world! (maybe)

The security of our infrastructures/software/iot/whatever electronic device is embarrassing. We have a lot of non-upgradeable devices (mobile phones, IoT devices, set-top boxes…) and we have also a lot of vulnerable design devices. Moreover, we have infrastructures managed by untrained (cheap) people, all around the world. We have critical services exposed to the Internet with default or blank passwords. But we have a lot of vendors selling the Holy Grail of security.

Holy Grail is not enough I guess, because we still have thousands of computers infected by WannaCry ransomware and sooner or later someone will pay the price.

Lesson learned: Cyber security is the current mantra, but it won’t help too much.

So we have mature cloud services, but automation, machine learning, and security are still too young. The last year demonstrated that connecting everything to the Internet is not always a good idea, especially if we’re connecting hospitals, 911 emergency numbers, peacemakers, airplanes, or IoT devices.

Should we live as Amish people does? Indeed that’s a possibility, and when I retire, I’ll think about that for sure.

How will be the future?

To conclude my mind-flow:

Will the cloud kill IT, experts? No, because infrastructures must be managed, platforms must be well-deployed, and services must be well-designed and well-implemented.
Will automation kill network experts? No, because network experts started to automate things 10 years ago, and the current “hype” is not changing things so much. You have to know what is below if you want to implement a good and effective solution.

Is Cybersecurity only a new “hype”?

The network has been always asked to fix poorly designed software. I could write a lot about how “stretched data-center” is a bad choice, but that’s not the right post. I can see more good tools used to fix the wrong thing (containers used as a VM, snapshot used as a backup…). But the network layer cannot fix every single mistake of the above layers, and sooner or later the above layers will pay a price.

I think that Darwin’s Law of Evolution by Natural Selection is valid also for computers and electronic devices. But it could dramatically impact human life too.