Learning from tech failures in AI world
The resurgence of Artificial Intelligence (AI) is clearly evident with various path-breaking innovations taking place in the field. This is a resurgence because AI is a concept that is half a century old, but it has taken off commercially in the past few years. As per Gartner, Artificial Intelligence and Machine Learning (ML) will be one of the top five investment priorities for over 30 percent of CIOs by 2020.
Failing Start-Ups a Major Issue
The brighter side makes everyone look forward to new products being introduced into the tech ecosystem, such as the recent launch of India’s first AI-based electric scooter. Although the coming of the AI in a strong way is being welcomed in the business world, there are also some challenges that need to be considered seriously. In the last two years, there have been many AI-related tech failures which force one to pause and analyze what went wrong while developing so many new products. Facebook’s M was shut down on 19th of January 2018 as it could not perform tasks that were meant for it and also did not justify its machine learning capabilities. Some tech analysts may tend to call this as a specific justification for closing up M. But there are some other critical angles to it too. In congruence to this, many other cases of tech failures have been recorded in AI space in recent years.
How Not to Fall
HR tech is estimated to be $400 billion marketplaces. Often it is found that the tech tools in the market don’t really improve productivity, or are not able to explain the need for the organization to have them when it comes to innovation and scalability. Hence, while creating any new AI product in HR Tech world, one must learn from past failures. Here is the list of some important lessons learned in relation to recent tech failures:
1. Funding - a major issue
Kaarya, launched in 2014 as a data analytics-driven fashion brand by Nidhi Agarwal and funded by Mohandas Pai and Ratan Tata, had to finally shut down in December 2017 owing to funding issues. They tried raising funds for last 18 months but were not able to survive. Another Fintech startup named Finomena based in Bengaluru had to face cash burn and hence, had to shut down in July last year. Started by IIT-Delhi graduate Abhishek Garg and Stanford graduate Riddhi Mittal in 2015, the startup used data and machine learning to reassess the creditworthiness of borrowers (mostly students and young professionals) before disbursing loans to them.
Fundraising is seen as a major challenge for many tech entrepreneurs. Around 8 out of 10 entrepreneurs’ crash within the first 16-18 months of starting their venture owing to cash paucity. One needs to master their business plan and use the thumb rule of ‘every penny counts’. Bootstrapping at every stage to attain a good market validation can make it a bit easier to raise funds. To reduce the expenses, young start-up strategies can include postponement of expensive investments and share office space with others.
2. Failure to produce a scalable product
Due to the slowdown in tech wearable market, Jawbone had to announce its exit in February 2017. Jawbone was the maker of a groundbreaking UP fitness tracker and a market leader in consumer wearable devices. From a once dominant fitness wearable company, it failed to impress users and reviewers after some time. Another example is of Pokémon Go. In this popular game, the creators of the algorithms failed to provide a diverse training set and didn't spend time in many neighborhoods, as there were fewer Pokémon locations in primarily non-White neighborhoods.That affected its scalability.
It is always recommended to try to imagine the life cycle of one’s product as a user with a design thinking approach. Questions on how to make it scalable are quite crucial. This helps to fill in the expectations gap. Not only should one observe what happens, but watch other users and customers react to trends. It may help one avoid being in the same position down the line.
3. Ignoring product development issues
Most of the Google Home devices recently faced a flood of complaints from several users reporting error messages every time they tried to interact with their smart assistant. In another instance, Google Home Minis were found spying on their owners. Google had to then quickly announce a patch to prevent the issue. There was again a scenario, when the AI tool missed to guess the winning horse in Kentucky Derby in May 2017. In another case that happened recently, a so-called crime fighting robot crashed into a child in a Silicon Valley mall as the machine’s sensors registered no vibration alert. Hence, product development becomes a prime area when it comes to predicting the success of any product.
In all the above examples, the development, design and the tasks that the product was supposed to perform had problems, faulty functioning and ignoring the safety of users. An AI company can survive and grow only if it takes into account the development problems and continuously improves upon them. Besides, as the reliance on AI grows, the questions of users’ privacy and safety have to be resolved at every step. Users are not going to use the AI machines if they invade their privacy and compromises upon their safety in the process of using them.
4. Lack of Research
Royal Bank of Canada invested heavily in AI through its research lab. In fact, Thomson Reuters follows three major research groups - traditional research group, application development research and the user experience group for delivering improvisation in its existing products and creating new products.
Inability to calculate and foresee future trends has been one of the primary reasons for many failures in the past and continues till date. Especially in India, research and consulting are not considered a priority and that’s why one often ends up in sometimes unforeseen scenarios.
5. Assembling the right team
To make the most of AI in the enterprise, you need to have a strong team of AI practitioners in place. As AI professionals are in high demand, to assemble the right team has become a herculean task. Retention and recruitment go hand in hand as companies lack skills to implement and support AI and machine learning. Alternatively, one can use the Freelance marketplace to get hold of expert consultation. Crowdbotics uses freelance talent as it helps them in cutting total development time in half.
The current investments are in nascent stages when it comes to implementation of AI in HR. But the future may be different; hence one must be cautious from beginning to avoid any unwanted collisions at the end. The underlying lesson is that mistakes are inevitable, but one should learn from such failures and avoid repeating those mistakes.