Artificial intelligence takes centre stage
The challenge for investors is to separate the real deal from the hype
Artificial intelligence (AI) is one of the hottest buzzwords in the startup scene, with AI companies attracting 15-50% more money in funding rounds. Even larger corporations, many not even tech companies, must still pay lip service and claim to use AI or risk looking like they are unprepared for an AI-driven future.
But how much of this so-called AI technology is actually of any value and not just empty buzzwords? AI is the de facto gold rush at the moment, as venture capital firms (VCs) throw money at AI startups and companies scramble to gain an edge on their competitors. So much investment will inevitably attract the usual hucksters and snake oil salesmen.
Due to the complicated nature of AI and machine learning, laymen can find it difficult to tell the difference between a major tech breakthrough and digital parlour tricks.
A recent study found that 40% of firms in Europe classified as "AI startups" actually did not have any AI technology at all in their business. While some of these businesses could be benignly mislabelled, others may be purposely misleading investors.
There's reason to believe this statistic is not unique to Europe.
"I don't have specific evidence regarding the Asia market, but assuming that organisations that track private companies have similar processes to those in Europe, I'd imagine the ratio may be similar," said David Kelnar, head of research for the venture capital firm MMC and author of the study.
SPOTTING THE FAKERS
Narongpon Boonsongpaisan, head of InVent, the venture capital subsidiary of InTouch Holdings, said bunk AI startups are nothing new in Thailand or the rest of Asia.
"Some startups use these buzzwords to attract investors, and I just have to frankly ask how their technology is at all related to AI," he said. "I don't know if it's just a misunderstanding or they just haven't achieved their goal yet, but it definitely happens not just in Thailand, but everywhere."
But Mr Narongpon added that a savvy VC has no difficulty sussing out the fakers. InVent looks for multiple factors to determine the validity and value of the AI technology by looking at data sets, the AI model and business results. The VC also looks at past performance and customer feedback, as well as if the people running the startup have any pertinent experience in AI.
"Whether its an AI or non-AI startup, we do a tech evaluation and test the performance of their product and examine the product very carefully to determine if it has a competitive advantage," he said. "The value of the AI startup is how impactful the AI can be to the core business."
Mr Narongpon said AI startups are relatively new in Thailand and are still in the "initial stage of the AI journey", with the most common being companies that make chatbots, but he expects the scene to develop much further within three years.
"I think AI is one of the trends we are looking at for high-growth startups," he said. "We believe AI is a tool for startups to get a competitive edge over other players and have also seen many corporations interested in the technology."
InVent is invested in 17 startups, having splashed 120 million baht in 2018 with plans to spend 140 million this year.
Mr Narongpon said VCs and corporations investing in AI are getting smarter at detecting misleading AI startups and are "learning along the way".
Scott Zoldi, chief analytics officer of FICO, said developers of AI should be working towards demystifying the technology and making the processes more transparent and understandable to regular people. A move towards "explainable AI" would in theory cut down on the amount of fakers using people's misconceptions about the technology to attract funding.
"If you are using AI to make decisions, you should have to explain it to the customer," Mr Zoldi said. "Even though AI and machine learning are great tools, individuals still need to know the details behind the decisions about them."
FICO, for instance, is trying to make its credit scoring more transparent through explainable AI.
"I see too much enthusiasm currently when it comes to AI," Mr Zoldi said. "It's such a hot topic that everyone feels they need to get in on it or get left behind, but there's less discussion about how it actually works and there's a lack of machine learning experts."
He said the inflated enthusiasm could lead to expectations not being met, which could lead to another "AI winter" or a period of reduced interest and funding in AI.
"People need to become more educated and know how to ask more questions to sift between firms that have actual machine learning and those that don't," Mr Zoldi said. "But in my experience in Asia, the amount of talent who can ask those questions is still lacking."
As with AI, many companies say they are using machine learning, when they are actually using a model that is not learning or improving itself. To avoid this, Mr Zoldi recommends clearer guidelines about what exactly constitutes AI or machine learning and a process to keep the fibbers accountable.
The EU has taken the lead in regulating AI with its General Data Protection Regulation, which requires companies to explain to its customers any automated decision-making that could affect them.