Mark Cuban believes that companies need to leverage AI to succeed, but says companies that use AI incorrectly are doomed to failure.
“There will be two types of companies: those that are good at AI and those that are not,” said the celebrity billionaire of “Shark Tank” fame. “And because AI is such a transformative tool, ‘everyone else’ will fail.”
The comments came on a call with Clipbook founder Adam Joseph, whose startup has secured a seven-figure investment from Cuba. Business Insider has reviewed a recording of the call.
On the phone, Cuban explained to Joseph that he believes everyone, from entrepreneurs to employees, should and should not be using AI.
Like fellow Shark Kevin O'Leary, Cuban believes that AI will have a huge positive impact on the companies that successfully implement it. However, instead of treating different AI tools as interchangeable, business leaders need to understand the complexities and differences between them. He said that if AI is not used effectively, a useful tool can become an expensive distraction.
“AI is constantly changing, so we need to put people in place, and really all CEOs, over time to understand all the nuances of every new tool that comes along,” Cuban said.
AI revolution is not progressing
Even though generative AI tools like ChatGPT have been around for more than three years, and other forms of AI such as machine learning have been around for decades, the world is still in the “first inning of the first preseason game” of the AI revolution, Cuban said.
Technology companies like OpenAI, Google, Microsoft, Meta, and Elon Musk's xAI are spending tens of billions of dollars to win the AI war.
Cuban said it's “too early to tell” which of these companies will be successful or whether others will develop go-to AI chatbots.
Technology entrepreneur Mark Cuban became a celebrity on ABC's business show “Shark Tank.” Craig Sjodin/Getty
“They all want to be the destination that everyone looks at, but it's not that easy. There's no winner yet,” Cuban said.
In Cuban's view, companies that underestimate the power of AI are doomed to destruction.
“If you're going to compete in the world of AI, data and information is more valuable than gold, more valuable than oil,” Cuban said.
But for all the hype around AI, Cuban is clear-eyed about the technology's limits. AI tools can be error-prone and overconfident, and chatbots aren't always smart.
“AI is stupid,” Cuban said. “But it is the savant who remembers everything.”
Cuban likened AI chatbots to people with minds like iron traps. These tools can instantly recall, process, and aggregate large amounts of information in one place.
“He does a really good job of putting everything he's collected together and presenting it like a person with an amazing memory,” Cuban said.
AI is not a panacea
There are holes in AI chatbots beyond so-called hallucinations, Cuban said. AI tools may not get the latest information. It can also be unclear how a conclusion is reached, as the algorithm is opaque and flawed or inaccurate links can be cited.
He also said that people often assume that “the AI model that they are using or creating” will provide all the answers they need, which “is actually not the case.”
Cuban said AI can actively harm companies that use it incorrectly or don't understand how it works.
Employees using standard versions of tools such as ChatGPT may be putting sensitive company information at risk. Similarly, businesses that post their work online should be aware that they may be offering their work for free to web-scraping chatbots hungry for new information.
“Companies are now learning that their intellectual property is extremely valuable,” Cuban said. “Two years ago, last year, two months ago, they might have posted everything online to show how smart they are, or they might have shared everything in their proposal to show how smart they are. Now they have to pay close attention to their IP.”
Cuban said academics and hospital researchers need to pivot in the age of AI from a “publish or perish” mentality of sharing research results widely in peer-reviewed journals.
“Now, doing that is the biggest mistake you can make, because all you're doing is training someone else's model,” Cuban said. “So you have to be able to understand what IP you need to protect, how to disseminate that IP, do you sell it or keep it for your own model, and how do you get the information?”
