One of my favorite quotes is “History doesn’t repeat itself, but it often rhymes” (usually attributed to Mark Twain, a famous American author). I’m reminded of this saying when I talk to C-level technical executives about AI. In this case, the rhyme is with Cloud Computing.
Virtualization and Cloud Computing Aren’t the Same
A few years ago, when I talked to CIOs and other senior managers at large organizations, I would ask them about Cloud adoption. In many cases, the answer I got was that “we’ve already fully adopted cloud computing”. After some further discussion, however, it became apparent that they had utilized virtualization, but they had not yet fully embraced Cloud Computing. The difference between the two is actually quite important because while virtualization can save a lot of money by reducing the need for expensive servers, it doesn’t move the business forward towards digitalization because it doesn’t provide automation, which is a base requirement for digitalization. The only way to get the full benefit of Cloud Computing is to have the enterprise silos of software be upgraded to take advantage of the self-modifying, self-optimizing capabilities of the cloud.Without this, enterprises end up adding more copies of siloed software (in more servers) and manually managing workloads, tasks which increase costs and impede the business’ ability to rapidly respond to changes in customer preferences and market demand. In addition to technical complexity and the discussion of whether a Private Cloud, Public Cloud or Hybrid Cloud should be chosen, a key impediment to broad adoption of Cloud Computing in most enterprises is the existing organization, which mirrors (and is the cause of?) the siloed software organization. Without breaking down the walls internally, it is almost impossible for an enterprise to fully adopt Cloud Computing. As a result, even today, as Cloud Computing becomes yesterday’s topic, many enterprises have still not moved to full digitalization because of the impediments of technology and organization. This is the History of Cloud Computing.
So, Where Is the Rhyme with AI?
Well, first of all, let’s discuss exactly what we mean by AI. One common definition is using machines to perform tasks which are generally performed by humans. That may be a little too general because it could include bank ATM machines which do the work of bank tellers and it misses a main use of AI, which is to augment the work of humans so that they are more effective. If we expand our definition to include this, then what technologies are included? It is in this area that the discussion of “Big Data”, “Machine Learning” (ML) and “Deep Neural Nets” (DNN) collide. I think we can say that DNN is simply the common way to implement ML, but how does Big Data relate? What about all the Data Lakes we were supposed to build and all the Data Scientists we were supposed to hire? Does AI mean that ML supersedes Big Data or are they somehow related?
Many of Huawei’s most successful Big Data customers create applications which replace or augment tasks done by humans. Examples include autonomous network troubleshooting and remediation, automated loan approvals and automated traffic light orchestration. Each of these has been implemented without ML. Could some of them be improved by ML, yes, probably, but they can’t be replaced by the Data Lakes and algorithms of Big Data.
On the other hand, Huawei has also developed a set of cloud-based solutions based on ML (and DNN) in such areas as speech recognition and image classification which can be used to augment and replace humans for tasks such as automated customs clearance and chatbots for customer service. Could these be done by Big Data analysis? No probably not, but they count on huge amounts of data which are stored in the same Data Lakes used for Big Data analysis and in fact usually depend on Big Data analysis to improve optimization. As a result, I think that both Big Data and Machine Learning can properly be classified as AI for a broad range of business activities.
“We’re already doing AI”
Now let’s return to the subject of why AI is the new Cloud Computing. When I talk to C-level technical executives today, I ask them about their activities related to AI and many times I get the answer that “we’re already doing AI” – echoes of Cloud Computing. In fact, some consulting organizations have reported that as many as 60% of enterprises are using AI. But again, when we have further discussions, “doing AI” can mean anything from “we’ve established a Technology Innovation group and they have downloaded Hadoop” to “we’re using Microsoft 365 and my sales rep told me that has AI built in”.
So if enterprises are going to really embrace AI, which includes both Big Data and Machine Learning (ML), what are the challenges they will confront and how should they proceed? First of all, digitalization is a pre-requisite. It is highly unlikely that an organization which has not implemented the steps, both technically and organizationally, needed to implement Cloud Computing, will be able to successfully address AI. The reason is that successful use of AI requires centralization of relevant data in a data lake (breaking down silos), adoption of software structures which allow for rapid, flexible adaptation (cloudification) and unity of organization in pursuit of a strategic vision.
Where Are All the Experts?
AI also poses an additional challenge. If you thought Data Scientists were hard to find, just try finding experts in ML. So should you just give up? No, but you need to attack the problem in a different manner. First of all, if you haven’t finished digitalization, don’t jump into AI because it will be a waste of time and money. Secondly, begin by building Big Data capabilities and the data lakes which will underpin your AI efforts. Finally, explore ways to enhance your knowledge of AI, particularly of ML. There are opportunities to use pre-packaged ML capabilities in the areas of speech and image recognition without having built the full knowledge bases needed to train accurate models, but they can’t really be successfully integrated into your organization without having a solid Big Data foundation to monitor and optimize them. And, of course, make sure that you pick areas for experimentation which are appropriate for your business and can demonstrate an ROI, either through cost reduction, revenue enhancement or customer experience enhancement. Finally, you need to pick partners with proven experience who can not only provide you with technology expertise, but also can work closely with you to plan and execute an AI strategy.
Certainly AI is not the same as Cloud Computing, but there are remarkable similarities and synergies between the two and enterprises need to build on their experience with Cloud Computing to successfully achieve the promise of AI.
Let us know in the comments below how your enterprise is approaching AI. And make sure you join us at Huawei Connect where AI and Enterprise Intelligence are central themes.