Analyzing AI's Bear Case

It might be time to urge some calm for investors after ChatGPT sparked a firestorm of interest in artificial intelligence investing.

‍

Stocks playing the trend have surged since OpenAI's ChatGPT became public. A new AI-powered version of Microsoft's Bing search engine has boosted the company's shares by more than 12% to start the year, outpacing the S&P 500's increase of more than 7%. Alphabet, on the other hand, is about 10% higher.

‍

The iShares Semiconductor ETF is also riding the wave of interest, up 23% this year. What about Nvidia? This year, the company has grown by a whopping 55%. In addition, there are other stocks on the trend, such as C3.ai, which has surged 128% over the last year. Stock prices of SoundHound AI, a company that deals with audio and speech recognition, surged 121%.

However, investors should remain cautious. AI offers promise, but there are pitfalls to avoid, as with popular investing trends in the past. In spite of AI's existence for a while, scalability remains a challenge for companies.

‍

“This is still early days and betting on individual companies, especially unprofitable ones, is still very risky and speculative,” said BCA Research's chief U.S. equity strategist Irene Tunkel.

‍

Artificial intelligence's bear case.

‍

It’s expensive

The computer power required for ChatGPT's human-like responses to prompts for Shakespearean poems may be impressive to users, but many companies would struggle to profit from generative AI.

‍

Morgan Stanley analyst Brian Nowak estimated last month that ChatGPT's cost per query would be 2 cents on average - or anywhere from $0.004 to $0.044 depending on the size and complexity of the query. Nowak estimated the cost of running a typical Google search at $0.003, which is less than half a penny.

‍

This will make building out generative AI capabilities for Google-parent Alphabet "significant", not to mention any other company trying to build out language processing models.

‍

“Based on the degree to which AI and natural language models like ChapGPT penetrates search, we estimate that for every 10% of GOOGL search queries replaced with natural language queries (similar to ChatGPT) in 2025, GOOGL will incur an operational expenditure of $0.6bn-11.5bn, depending on the number of words per search result,” Nowak wrote.

‍

Google’s dominance

If companies fail to take market share from Google, they will find it difficult to make money even without those costs.

‍

StatCounter, a web traffic analysis firm, estimates that Google holds 93% of the search engine market worldwide. As for Microsoft's Bing? Only 3%. Competitors include Yahoo!, Yandex, and Baidu, China's leading search engine they have even smaller pieces of that global pie.

‍

“I think it makes sense for Microsoft to pursue it because I don't think the Google function is perfect. I think OpenAI's chatbot is a better mousetrap than Google's search engine," said Pat Tschosik, senior portfolio strategist at Ned Davis Research.

‍

“Apart from Microsoft, I wonder how many companies have the technology, resources, and profitability to succeed,” Tschosik said.

‍

The perils of AI gaffes

In the meantime, other challenges remain.

‍

The accuracy of generative AI can only be measured as a function of the large number of datasets that it is trained on, so its accuracy can be limited by the amount of data it is trained on. Alphabet learned this to its detriment when its AI chatbot Bard responded incorrectly to a prompt.

‍

As a separate issue, the lack of proper citations in generative AI could lead to future lawsuits for AI companies. ChatGPT scrapes websites for data, but it must properly source its material to avoid plagiarism.

‍

Tschosik said that at least Wikipedia cites the source of everything. “However, ChatGPT does not list any sources for its responses, and maybe that will have to change.”

‍