(this article originally appeared in Towards Data Science and was written by Alexandre Gonfalonieri)
With the rise of no-code/low-code AI platforms, I wanted to write this article to explain the business model behind these solutions and what could it mean for data scientists. Will the job of a data scientist disappear or just evolve?
These Code-free solutions often come in guided platforms, offering classic drag-and-drop functionality to fully automated machine learning services suitable for beginners and Machine Learning professionals alike.
Today’s reality is that Machine learning experts are hard to find and… hard to keep. It takes time and investment to transform your business into an AI-driven organization. As such, it is almost impossible for some companies to adopt AI.
When it comes to mid-size companies, there is a high demand for data scientists as companies lack the technical talent required to build scalable AI solutions but companies who are unable to hire developers are facing the risk to be left behind. As a consequence, a growing number of companies are increasingly turning to no-code platforms for machine learning.
No-code platform: Platforms that enable companies and business professionals with minimal or no coding experience to build apps and fill the talent gaps in their organization.
New tools to help data scientists
On top of no-code AI solutions, we are also seeing a lot of low-code solutions. Indeed, a growing number of tools promise to make the field of data science more accessible. Obviously, this isn’t an easy task considering the complexity of the data science and machine learning pipeline. None the less many libraries and tools including Keras, FastAI, and Weka made it significantly easier to create a data science project by providing us with an easy to use high-level interface and a lot of prebuilt components.
I realized that it is often difficult for a company to attract highly qualified machine learning specialists, the demand for which is growing and many times exceeds the supply. The solution here may be to provide access to automated machine learning tools.
Automated machine learning: the process of automating the process of applying machine learning to real-world problems. AutoML covers the complete pipeline from the raw dataset to the deployable machine learning model.
The idea is to challenge the traditional approach of learning technical machine learning and introducing more accessible machine learning.
The underlying objective of these business models is to harness the size and perhaps the international scope of the SMEs with no or limited data science teams.
Secondly, by building low-code or no-code AI platforms, market leaders will more easily establish themselves as the central AI ecosystems and further build alliances to scale the use-cases and delivery as well.
I also believe that large tech firms invest in no-code AI model building platforms to accelerate the AI democratization and build this top-of-mind awareness from a branding perspective. Furthermore, it helps companies win over programmers or end-users who will build their own models. The business model might be to create a relatively easy to use but limited platform and propose a premium subscription for better AI-related services (training, assistance, etc.). As of today, I expect these companies to charge users based on the number of requests.
I believe that large tech firms are actually building AI ecosystems and having a no-code AI platform can help provide customers with a no-brainer approach to staying on that brand’s ecosystem and roadmap, which reinforces the customer lock-in effect. Moreover, it increases large tech firms’ user reach and enables them to propose a service to not just data scientists but also business users.
Leading AI companies are developing a variety of platform strategies to accelerate large enterprise adoption but at the same time, the SMEs market remains untapped…
Ideally, these companies want business users to start with small POCs to see what the technology can do and also remove all possible barriers to entry. As they become more familiar with the technology and the process, they will have other ideas about how AI can address specific business problems.
I believe that there are actually two races going on in AI right now:
— one to win over programmers and one to win over users. “The question is, how many end-users will actually build their own models, even if it is as easy as using drag-and-drop mechanisms?”.
Large tech firms open-source or develop no-code applications because they wish to be the foundations on which other people innovate. By doing so, they will be able to gather all the most strategic data of many SMEs…
Open-sourcing and no-code AI platforms serve these companies’ broader goals of staying at the cutting edge of technology. In this sense, they are not giving away the keys to their success: they are paving the way to their own future.
On top of all these elements, a no-code AI platform also makes it possible for these companies to do some cross/up-selling by proposing additional AI expertise through consultants and computing resources to quickly and efficiently complete AI online training.
Another interesting perspective is that no/low-code AI platforms can help startups which lack a formalized AI practice build better applications and solutions quickly without any coding experience at a much lower cost. This is also another very interesting market…
Automation is coming for many tasks data scientists perform… Indeed, Every major cloud vendor has heavily invested in some type of AutoML initiative or no/low-code AI platform.
Today, most solutions offer pre-built algorithms and simplistic workflows with features like drag-and-drop modeling and visual interfaces that can easily connect with data and accelerate bringing services/applications to the market.
- Google’s Cloud AutoML trains codeless machine learning models, with a simple interface (based on drag and drop). It focuses on artificial sight, natural language, and translation. Google also offers Teachable Machine, an even simpler tool designed for amateurs interested in experimenting with machine learning and understanding how it works. With nothing more than a camera (webcam or cell phone camera), Teachable Machine feeds a small neuronal network in the browser, without having to send the images to a server and export it to websites, apps, physical machines and more.
- Uber AI developed Ludwig, a code-free deep learning toolbox, to make deep learning more accessible to non-experts. The non-experts can quickly train and test deep learning models without having to write code. Experts can obtain strong baselines to compare their models against and have an experimentation setting that makes it easy to test new ideas and analyze models by performing standard data preprocessing and visualization.
- Baidu’s EZDL is a simple drag-and-drop platform that allows users to design and build custom machine learning models.
- Obviously.ai is a startup that enables users to run complex predictions and analytics on data, simply by asking questions in natural language.
These Code-free machines often come in guided platforms, offering classic drag-and-drop functionality to fully automated machine learning services suitable for beginners and Machine Learning professionals alike
I believe that this growing number of no/low-code AI solutions will pose a greater competitive threat to application platforms vendors, including MEAP and PaaS providers, which lack a low-code strategy.
What is possible through no/low-code AI?
Most of the existing solutions can:
- Automatically identify the design style and room type from images of the interior of houses
- Identify defective products in manufacturing environments
- Monitor cracks in roads and pavements
- Detect if a customer is likely to churn
- If a loan would be approved, etc.
The leading ones at this point only help with the actual implementation and training of a network, not with the feature engineering, data analysis, or testing.
Limits of Current Solutions
These solutions sound great but… you’ll surely still need data experts. Indeed, business users must know what to drag and where to drop it… However, complex projects have thousands of tasks and require data scientists.
Drag-and-drop tools look great if you only need to drag and drop a few things but the reality is not that easy. I have seen the workflow of a lot of AI projects and I can assure you that, in order to scale and reach production, your AI project will have thousands of tasks.
About Data Scientists
It is safe to assume that your data scientists will not feel comfortable using no-code/drag and drop tools. Indeed, they already know how to write code, so they don’t need a drag-and-drop UI. You’ll probably hear them complaining about the lack of scalability of such tools…
Despite these legitimate elements, the takeaway for data scientists is to look toward improving their skills in things that are not automatable.
I believe the role of data scientists will remain a vital element for the success of an AI project (depending on the complexity).
In my opinion, there is no, nor can ever be, AI or no-code tool that you can feed even with the most moderate rules of a business and let it generate an application. There’s always going to be the human element that does the feature engineering, architecture, testing and maintenance.
Moreover, with the rise of no/low-code AI solutions, we might see Data scientists ending up spending too much time fixing their colleagues’ work as they could on their own tasks.
It is safe to assume that as we continue to democratize AI, many tasks Machine Learning elements will be automated. This shift will enable data experts to be more strategic and creative in how they solve problems.
One can hardly expect that a business will be able to use the results of automated machine learning without the help of data experts. In any case, the preparation of data, interpretation of the results and other stages of an AI project will still require a data expert.
In my opinion, Data science will continue to be strategic — but the world is moving to a functional data science world where practitioners can do their own analytics. You need data engineers, more than data scientists, to enable the data pipelines and integrated data structures.
The future data scientist should be even more specialized, tackling the most business-critical and complex challenges that will help their businesses create added-value.
All these platforms and tools still have their limits, but are a sign that ML for amateurs — simplified and almost homemade AI — is getting closer and closer.