When it comes to presenting the core capabilities of a startup, there has often been an overlapping use of the terms analytics, artificial intelligence or data science. Many companies who claim to have a strong data science team internally, which they say is working on complex data problems, is really statistics and programming.
This ecosystem seems to be cluttered with many companies calling themselves “AI startups”, even when they are only using analytics. Many startups claim to have a machine learning team, but do they really have ML capabilities? It is important to get the hype sorted.
The Hype That Has Become A Trend
Every technology provider wants to push AI into its product strategy for the exciting opportunities it brings along. There has been a tremendous increase in startups who claim to be offering AI products without any real differentiation from the other companies.
For instance, there are many chatbot companies who claim to be using AI to run these chatbots but in reality, they are just programmed functions that need constant ‘manual’ updates. If it were really AI, it would have learnt to improve and deliver results at par with humans without any intervention. The other popular process confusion is between automation tool that replaces simple administrative tasks such as insurance claim and fraud detection.
Many startups today claim to be AI startups to attract funding. Investors today are in a fix as they say that every pitch they come across has terms like “AI”, “deep learning” or “analytics” in it. Most companies are even reinventing their products with an “AI touch” in it. While AI techniques are useful in solving numerous problems, they are not yet applicable in every case. It should be understood that adding an off-the-shelf algorithm to old software is not necessarily going to teach it new tricks.
How Real Is The AI?
While a lot of companies are claiming to be AI startups, they have little AI capabilities. They are using AI to market existing approaches. Some are even hiring people who can somehow insert AI into their product to show AI as a commodity skill, like Java programming.
Another problem is that companies often claim to offer AI when actually they are using classic machine learning solutions rather than more modern techniques such as deep learning. Most AI products sold today use quantitative statistical techniques instead of actual AI algorithms and models to train the products.
How To Declutter This Confusion
If you do not have enough data, you are not an AI company: AI learns from large amounts of data and it is difficult for most firms to generate enough data to make algorithms efficient or simply to afford to hire data analysts. The data source is what ultimately drives the ecosystem, and it must be well-structured and optimised.
Online chatbots are not necessarily AI-chatbots: Many firms are claiming to use AI in their chatbot functionalities which in actuality may not be AI-based. Most of these are updated manually which defies the rules of being AI.
If your products are not learning on their own, they are not AI products: The basic aspect of AI is to have intelligence and learning capabilities of its own. If they are not learning on their own, they are not AI-driven.
Using AI for fine-tuning doesn’t make you an AI company: Most companies just use AI to fine-tune their existing products rather than introducing radically new products. For instance, the AI-based toothbrush that recently became popular isn’t a new AI product in itself but used AI to improve some of its functionalities.
Programming and running statistical analysis doesn’t make you an AI company: With no proper algorithms in place, it would be unfair to call a company AI company. It should be a whole package of stats, data cleaning, memory management and algorithms.
Half-baked software products are not AI products: Software development is not as easy as it seems. While many companies begin to work on developing AI products they often get stuck and sell half-done products.
Rule-based systems aren’t AI: Technically if you are using the rules-based system, it may be called intelligent but not AI-enabled. The point is that the system should be able to learn on its own and not just automate tasks.
As Manish Singhal, founding partner of pi Venture said in an earlier interview with AIM on how he separates AI hype, “For us use cases drive investing decisions. We tend to see how technology is bringing about a 10x difference to the existing way of solving the use case.”
As time goes by, the distinction between AI and non-AI-based companies will become more profuse. However, the scenario right now seems to be in a state of confusion. There is an obsession with calling the startup an “AI startup” not just to attract investors but to stand out in this hugely competitive market where AI is substantially making a lead.
However, there are certain points that one can keep in mind before choosing AI vendor. It is important to ask questions such as what AI method is being used in its solution, how robust will be its implementation and deployment, how often will the AI system be re-trained. Also, any vendor claiming that their product includes AI should also be able to explain how it will benefit the end-user more than other product that does not use AI.