Cloud and developers will democratize the use of AI
“Intelligence is the efficiency with which you acquire new skills
in tasks for which you were not previously prepared”
Francois Chollet, Software Engineer and AI Researcher, Google.
Imagining the future of humanity supported by Artificial Intelligence (AI) is very exciting. The last century saw the birth of the first computational models to represent knowledge, and with it, the concept of AI was born. We will remember the century thanks to its democratization . The next one, predicted by futurists, will be the century in which AI is in all the daily activities of the human being.
So it turns out that making the applicability in everyday life a reality requires, in principle, its democratization. Bernard Shaw said that “democracy helps us ensure that we will not be governed better than we deserve.” Consequently, the quality of the outcome of the democratization process will depend on four factors:
1.-Numerous competition from companies that innovate in the development of AI models for different industries.
2.- The platforms to generate the AI models must innovate in the data preparation phase; also in the automation capacity for the construction, deployment and administration of these.
3.- These platforms must be built with ethical principles.
4.- Industries must be aware that investment in the generation of AI models spreads the risks of experimenting; It also helps to establish the principles of consumption, to be later adopted by final consumers.
Who will democratize its use?
As for who will democratize its use, at least two actors can be identified: the cloud and software developers . Bearing in mind that we live in the multi-cloud era, Tomás Valles, director of Systems Engineering at VMware Mexico, explains that companies can start by consuming AI models offered by cloud providers such as Amazon, Azure, Google, etc. if it were one more microservice through which the information flows to be analyzed and then deliver behavior patterns for decision making as a result. Without a doubt, with this pragmatism, it does not seem difficult to acquire or implement an AI service,On the contrary, companies will be able to access its benefits in an agile and simple way, without the need to develop it on their own. In addition, companies that are being born with digital services in the cloud will more naturally adopt the use of AI.
However, for complex problems, the solution is not achieved through the implementation of a single scheme, but through the combination of multiple models and technologies. Gartner calls this Composite AI.
In its report, “5 trends drive the Gartner hype cycle for emerging technologies 2020” , Generative and Composite AI is positioned as trend number four. In another report, also by Gartner, ” Top strategic technology trends for 2021″, nine strategic technologies that will guide organizational plasticity in the next five years are addressed.
DataOps, ModelOps and DevOps
Among these is the adoption of a solid AI engineering strategy that will facilitate the performance, scalability, interpretability and reliability of AI models, while at the same time guaranteeing the use of the value of investments in the field. Without AI engineering, most organizations will not be able to take projects beyond proof of concept and prototypes to full-scale production. Gartner also points out that AI engineering is based on three basic pillars: DataOps, ModelOps and DevOps.
DevOps deals with code changes at high speed. This is essential because Artificial Intelligence projects experience dynamic changes in code, models and data; therefore, everything must improve continuously.
Organizations that start to implement AI will need to apply DevOps principles, through the creation of pipelines for DataOps and pipelines of the machine learning model for MLOps, which affects taking advantage of the benefits of AI engineering.
In a somewhat countervailing fashion, DataRobot CEO Dan Wright earlier this year said he doesn’t think the future is so much about MLOps, but more about monitoring the entire life cycle of a model and continually updating it as data changes. data. In the latter, there is agreement with Gartner.
For DataRobot, the most important thing is the combination between MLOps with automated machine learning. Automation is the name of the game, so AI models update automatically as data changes for continuous learning. Wright says it’s not about working for six months to put an AI model into production. In this regard, I completely agree, and in that sense, the speed of deployment in code changes will promote the adoption of AI.
Democratization of AI
Having talked about DevOps, automation, composite or generative AI, and the cloud, whatever the complexity of the AI model, whatever the level of flexibility in configuring it, or how easy it is to build those models on specialized platforms or from the services of cloud providers, it should also be added that they are the software developersthe enablers of the AI democratization process. Because, your programming skills will make it possible for models to analyze deeply, almost like humans, but in faster times, tons of data. In addition, they will materialize that it is inversely easier to use. And, finally, in its rational nature to code, the trust and ethics on which the Artificial Intelligence models are based will also be deposited.
Returning to the relevance of the cloud, it can be added that another of its values to support the democratization of AI is the scalability and agility for computing tons of data required, intensive both in CPU and economically.
Boost the market
It is the cloud and the adoption of GPUs, a fascinating topic, which is perfectly articulated with the democratization of AI , which I will address in another article. For now, I will end by pointing to the relationship between actors; with the quality of the result of said process. So, competition is the context of free participation in the market that will be able to promote greater actors at the same time as better quality.