2024-09-03 14:28:36
Billions invested, incredible excitement, a new boom in the digital world, it all started with the sudden explosion of free generative artificial intelligence (AI) for all with ChatGPT.
Once the hype has died down, a growing number of opponents are expressing doubts about the impact of generative AI on the economy. Several experts see flaws and worrying issues in this famous AI.
So why would AI be overrated?
AI is not so… smart
Generative AI and its chatbots can perform very well on ChatGPT and others, like a real superintelligent machine.
But these bots are just a fancy way of aggregating the Internet and spitting out a mishmash of what they’ve found, which relies on work done by previous humans but is incapable of generating big new ideas.
AI is flouting copyright, so companies are going to great lengths to prevent their data from being gobbled up by thinking machines.
In short, experts agree that we are dealing more with “machine learning” than with true “artificial intelligence”.
Other well-supported evidence is that AI cannot distinguish between true and false and struggles to calculate basic mathematics.
The Lies and Failures of AI
The AI industry and the media have come to refer to the falsehoods and errors generated by AI as “hallucinations.” Every now and then, it seems as if AI is smoking the good stuff or eating mushrooms, and then says weird, made-up things that have made headlines. Hallucinations are quite common, such as putting glue in pizza sauce.
Not reliable enough to replace workers
There are concerns that AI will eliminate jobs. However, the opposite seems to be happening, for example, companies are looking for more translators despite the availability of intelligent language tools, simply because AI is not reliable or intelligent enough to perform our tasks.
And that seems to be true for a whole host of other jobs.
AI’s capabilities have been exaggerated
Similar to translation jobs, computer coding jobs were thought to be at risk from AI. However, researchers have found that much of the code generated by AI is not very good. Researchers at Bilkent University found that over 30% of AI-generated code was incorrect and 23% was partially incorrect.
AI use remains surprisingly limited
In a recent study, the US Census Bureau found that only 5% of businesses had used AI in the past two weeks. That’s not going to have a profound impact on the economy or productivity. Right now, AI is replacing people because it saves them money, shifting some of the work to customers. Not great.
AI is running out of steam
First, progress in improving these models depends, in large part, on the use of large amounts of data. A major problem: the ChatGPTs of this world have already consumed the entire Internet.
Worse, their thirst is now coming up against data access restrictions imposed by content and media companies.
Another constraint is a shortage of the special electronic chips needed to run AI. This is another major cost and headache for AI companies. Not to mention the costs of the inordinate amount of electrical energy needed to power all those chips.
1725480550
#Jobs #real #capabilities #limits #reached #artificial #intelligence #overestimated #reasons
