(Photo: Smith Collection/Gado/Getty Images)
Gad via Getty Images
Ai is great, isn't it?
Well, that's certainly for the large tech companies that make and support it. Stock prices of Microsoft, Google, Amazon, Nvidia and other companies in the middle of the AI boom have skyrocketed over the past few years. Billions are cultivated into data centers, infrastructure and hardware to support the hungry needs of new technologies. Indirect companies and countless ambiguous startups from US steel to HP and local utilities benefit from the construction, energy consumption and computational demands created by the industry.
Hallucinations, inaccuracies, false information: data
Still, the public is not convinced. In most cases, AI is fun and even productive. However, AI applications remain unreliable due to hallucinations, inaccuracies and incorrect information.
A recent KPMG survey found that over 48,000 people around the world say that around 66% use AI regularly, while only 46% feel willing to trust AI systems. A recent survey of over 1,100 people found that around 82% were somewhat skeptical of the AI ”overview” in searches. 61% may “trust” the results, 21% do not, and only about 8.5% always trust the response. Another study from Gartner supports these findings, with around 53% of consumers saying they don't trust the results of AI-powered searches or summaries. why?
These studies show that many users continue to see serious mistakes. 42% reported experiencing inaccurate or misleading content, about 36% reported losing important context and 31% reported reporting bias in summary results.
Even software writers are suspicious. Developer polls show that 84% of 84% plan to use AI coding tools, but in reality only about a third trusts the output, which has been declining since previous years. According to the report, their main complaint is to get “almost correct” results, which ultimately takes additional debugging time.
Why can a product that works so poorly, so hype?
Hallucinations, inaccuracies, false information: big technology is the problem
Despise big technology. Since Openai released ChatGPT three years ago, software companies have deployed AI products that are not ready for prime time to catch up. Still, these companies keep pushing customers to buy AI applications and use AI agents to increase productivity when the opposite is happening.
According to a report from NBC News AI's “Slop” (blurred logos, non-sensical text, general or unsleek writing/code), companies are forcing freelancers, artists, writers and developers to fix or terminate what the AI was wrong. These companies have found that many revisions involve more effort than expected. Many people find it easier for human workers to start from scratch, rather than patching things AI-generated.
“In April 2024, it seemed Agent AI would be the next big thing,” writes AI expert Steven Newman. “The next 16 months have made great progress in many ways, but little has been made in real-world agencies.”
Hallucinations, inaccuracies, misinformation: it's not their fault
It's easy to blame the big tech companies for not actually working. And they deserve a lot of responsibility. But likewise, companies, especially large companies, are throwing hundreds of millions of dollars at things like this without thinking properly.
The New York Times recently reported that of 80% of companies using generator AI, they have never seen a significant final impact, with 42% reporting that they had abandoned most of their AI pilot projects by the end of 2024, a sharp rise from 17% the previous year. Still, the Times reports that companies continue to “actively” increase their investments in anticipation of generative AI spending, which is expected to nearly double this year. It's like a gluttony for punishment.
Another MIT report states that it's not necessarily the big tech fault, but rather the way customers deploy AI. It was found that around 95% of AI pilot programs were unable to provide measurable profit and decline effects, and that only about 5% of these pilots achieved rapid revenue acceleration. These are things that “focus on a particular problem and execute well.”
The report found that many failed pilots did not fail because of poor AI models, but because the tool was not integrated with existing workflows, it did not adapt to the company's needs or because there was no learning-enabled system. MIT researchers suggest that back-office automation, finance and procurement are one area where return on investment is excellent, but instead many companies invest heavily in features such as difficult to expand, such as sales and marketing.
“The 95% failure rate for enterprise AI solutions represents the most obvious symptom of the Genai disparity,” writes Zvi Mowshowitz, a former hedge fund manager and longtime AI commentator. “Organisations that remain on the wrong side continue to invest in static tools that cannot adapt to workflows, but are focusing on learning-enabled systems beyond Divide.”
He's right. I've always used AI for simple applications (for example, it helps to summarise research and content from various sources I use to write this article). Many businessmen I know are happy to use it to transcription conversations, write emails, draft company policies, and do basic analysis. Generating AI platforms are very useful for these types of activities. But is it worth all the hype? All high ratings?
Hallucinations, inaccuracies, false information: AI bubble
Chevaugn Powell, author of “The Trillion-Dollar Ai Bubble Nobody Soie Shows Coming,” believes he is creating a bubble that is different from the 2001 dot com collapse.
“According to research firm Gartner, spending on generated AI will reach $644 billion this year alone,” he writes. “On the other hand, last year these same hyperschools generated only $45 billion in actual AI-related revenue.”
He also points to industry-wide vulnerabilities to new innovations.
“The single Chinese startup (Deepseek) has proven that the emperor has no clothes in the AI kingdom of Silicon Valley,” he said. “The tech market didn't wobble when Deepseek announced it had built an AI model that rivals ChatGpt, which is under $6 million.
Still, profits are a big technology. Experts, experts and academics warn us that we have to “invest” in AI or risk falling behind, losing or going out of business. However, it is clear that most business people still don't think it makes sense to invest in something that clearly doesn't work well. There's probably a bubble.
Most of my clients – small and medium business owners – look carefully at this technology. They see possibilities. But they certainly trust it now. And thanks to the disappointment, their level of trust in those big tech companies is also low. It is not known that it relies on AI to help execute or manage core business processes. And you can't blame them. Big Tech is a product that doesn't work very well. And their reliability is getting injured because of it.
