Faking AI: the Modern Mechanical Turks

Faking AI is becoming pretty common for startups that need to test a MVP. However, it could lead to an AI winter.

Analysts expect that Artificial Intelligence (AI) will be part of daily life in every aspect of our routine. Currently, several solutions take advantage of AI for escalating operations, accelerating production, and reducing costs. If a founder wants to gain the love of venture capitalists, integrating AI in its operations is a must. AI can be labor-saving, but developing AI is complicated and expensive. In consequence, some companies are faking AI.

How is it even possible to fake AI? Humans require more energy to complete a task, and they can’t work 24/7 hours. Cheap abroad labor plays a key role in achieving this goal. Job marketplaces such as Amazon Mechanical Turks connect companies with individuals to perform tasks that the AI still doesn’t domain. 

Mechanical Turks: From chessplayers to chatbots

The history of humans faking automatons started before the industrial revolution. In 1770, Wolfgang von Kempelen presented to the Empress Maria Theresa of Austria an Automaton Chess Player. Dressed with Ottoman ropes, the automaton was baptized as the Turk.

The Turk exhibited a strong chess game and defeated several statesmen in Europe and the United States, including Napoleon Bonaparte and Benjamin Franklin. However, it was a hoax. A real chess player was hidden inside the Turk. 

Real machines defeating humans in chess should wait more than two centuries for appearing. When Kasparov lost against Deep Blue, the soviet chess player argued that the IBM computer was a fraud, controlled by real chess players to defeat him. 

Computers are now so good at playing chess that none Grand Master can defeat them. However, there a new mechanical Turks in other areas, such as chatbots. This fact evidences that AI still has a long road to becoming a successful solution for every human need. 

Faking AI for probing a concept

Integrating AI in operations can reduce costs for companies. However, developing AI is expensive. And sometimes, companies could invest in features that nobody requires or are wishing to pay for them. 

For that reason, several companies are faking AI to probe a business concept. The underlying idea is that humans will do the tasks until an AI-based solution is developed. However, companies have failed to be sincere in their current stage of development. In some cases, companies don’t report that a human will check the information that the user is given to them. How would users feel if they discover that the company that assured all their private info will be processed by a machine is sharing their info with third parties?

Another AI-Winter, the riskiest scenario

AI solutions are currently well-branded. Venture capitalists are interested in investing in AI-based companies. Meanwhile, a startup can get easy access to funding if assures that its solution is powered by AI. Nevertheless, companies should be realistic and sincere in the promises that are delivering to investors and users.

Innovation Consultor Eric Ries suggests faking AI in his best-seller book Lean Startup. He didn’t call in that way, of course, but he describes as an acceptable MVP a service provided by humans instead of AI, before developing an AI solution. The logic behind this is testing if the public wants your service before investing a high amount of time and money in a solution that nobody cares about.

For testing an MVP, faking AI is an acceptable approach. However, in some cases, companies don’t inform users that a human will check their private information, and don’t inform investors that they still lack a scalable technology. These attitudes could lead to a loss of trust of the investors in AI-based companies. And then, an AI winter.

Read also: How can AI and Data Management avoid dropouts in College?