Think AI's Prospects Are Overblown? Try This Beaten-Down Growth Stock.
Is artificial intelligence a bubble? That's the trillion-dollar question. Large language models (LLMs) like those that power OpenAI's ChatGPT are certainly impressive, capable of generating coherent text, images, and functional code. AI agents, which leverage LLMs to perform complex, multistep tasks, could be a huge deal as companies race to adopt the technology. Annual spending on AI infrastructure is on pace to surpass $200 billion, according to IDC, as tech giants throw caution to the wind.While the AI boom is still going strong, there are a few reasons to believe that the industry is vastly overselling the capabilities of AI technology. LLMs can do a lot of things, but they may not be well suited for the kinds of real-world tasks that will ultimately generate revenue for companies.Writing computer code is a great example. Top-tier LLMs can write impressive-looking code, but these models work by predicting the next token in a stream of tokens. There's no understanding or reasoning going on, just a convincing illusion. Some mistakes will be easy to catch, but others will be subtle and require the expertise of a senior engineer. For companies relying on LLMs to write mission-critical code, it's questionable whether the technology will actually reduce costs given its tendency to produce errors.Continue reading
Is artificial intelligence a bubble? That's the trillion-dollar question. Large language models (LLMs) like those that power OpenAI's ChatGPT are certainly impressive, capable of generating coherent text, images, and functional code. AI agents, which leverage LLMs to perform complex, multistep tasks, could be a huge deal as companies race to adopt the technology. Annual spending on AI infrastructure is on pace to surpass $200 billion, according to IDC, as tech giants throw caution to the wind.
While the AI boom is still going strong, there are a few reasons to believe that the industry is vastly overselling the capabilities of AI technology. LLMs can do a lot of things, but they may not be well suited for the kinds of real-world tasks that will ultimately generate revenue for companies.
Writing computer code is a great example. Top-tier LLMs can write impressive-looking code, but these models work by predicting the next token in a stream of tokens. There's no understanding or reasoning going on, just a convincing illusion. Some mistakes will be easy to catch, but others will be subtle and require the expertise of a senior engineer. For companies relying on LLMs to write mission-critical code, it's questionable whether the technology will actually reduce costs given its tendency to produce errors.
