TECH NEWS – New research shows how risky the AI industry is: the vast majority of AI projects fail!

OpenAI, the multi-billion dollar startup that revolutionized the content creation (generative) AI sector with ChatGPT, is likely to make a $5 billion loss this year, yet the company is in talks to raise more money, and a $1 billion capital injection will reportedly push the company’s valuation to $100 billion. But this is just one company that trains AI models. There are others, and many of them are burning through money so fast that they can hardly be called successful.

A consortium of engineers and scientists estimates that 80% of these projects fail, and they have shown why. The RAND Corporation, a U.S. nonprofit global policy think tank, research institute and public sector consultant, has identified five reasons for many of the failures. First, industry stakeholders misunderstand what the problem is in applying AI. Then, they lack sufficient data to effectively train their models, leading to inaccurate results and discouraging users from using the platform.

Poor infrastructure can also contribute to the failure of an AI project, and with insufficient resources, founders often focus on technological advantage rather than creating value for users. According to the RAND Corporation, one of the remedies is to invest in infrastructure, as this will allow the model to be trained faster and provide good quality data to effectively train other AI models. Founders also need to understand that AI is not the answer to everything, as it has limitations.

Effectively training an AI model can indeed lead to the birth of a more muscular product; there is ChatGPT, which has been trained on terabytes of data and yet is still able to give sometimes inaccurate answers to our queries. So even the biggest player can make mistakes, so no one and nothing is perfect. Even OpenAI is not perfect.

Source: WCCFTech, RAND