Frank Downing, Ark Investments, believes that two firms that specialize in artificial intelligence - UiPath (a business process automation company) and Twilio - are underappreciated. UiPath is one such firm. Downing, the director of research at ARK Next Generation Internet ETF said that large language models could unlock 20% or even 40% more processes that UiPath is able to automate. This is in addition to the processes they are already automating. This could lead to fewer human intervention, which would improve efficiency for UiPath customers. The stock has risen 41% in the past year. Not everyone is convinced. Analysts from investment bank Mizuho do not believe that the stock has much upside due to the conservative revenue guidance and the slower growth of new customers as a result of the economic slowdown. While we are encouraged by UiPath’s improved profitability, we expect shares to remain range bound over the short term until the company has made significant progress in its strategic repositioning for growth," Mizuho analysts headed by Siti Pantigrahi wrote to clients in a May 25 note. Mizuho’s $16 price goal points to a 10.8% decline in the stock. Downing also identified Twiilio as a potential undervalued stock.
Twilio connects companies with their clients. It is commonly used as a customer service or call center application. Twilio would be able to use large language models to make its messages "more personal" and "impactful" with the customer data it has access to. The high costs of creating new AI models have been a major concern for investors. The director of research, however, believes that Twilio is likely to be the company to benefit from the next wave of growth due the increasing commoditization of AI. In just two years, the training costs for GPT-3, an AI model that is known to be a commoditized version of AI, have dropped from $4.6m to $400m. This represents a 70% annualized cost decrease. Downing believes that while firms such as Microsoft and Google demonstrate a wide range of AI products, consumer willingness to purchase these could be limited because open-source models are becoming more commoditized. Downing believes that value and differentiation will come from niche applications that offer high returns on investment.