Will AI bring benefits? Those who pay for it are already worried

AI For Business

Economist Daron Acemoglu warns that another big red flag is the common assumption that by processing more data and harnessing more computing power, generative AI tools will become more intelligent and more accurate, predictably delivering on their potential. His comments were published in a recent Goldman Sachs report, “Gen AI: Too Much Spending, Too Little Benefit?”

“Today's large-scale language models have proven to be much better than many expected,” he says, “but it's still a big leap to believe that an architecture that can predict the next word in a sentence will achieve capabilities as intelligent as HAL 9000 in 2001: A Space Odyssey.”

The final warning from skeptics (or realists) is that the road from “pretty good” to “perfect” for AI could be as long, or even longer, than the road from “nothing” to “pretty good.” Even if artificial general intelligence were to reach perfection, or something reliably close to it, the energy load could collapse the U.S. power grid, which is currently struggling through the summer, as a text message from Con Edison reminded me this week.

Those loudest in the crowd claiming AGI (HAL) is on the horizon are the people who stand to gain most from the fuss. Trillions of dollars in shareholder value hang in the balance. Consider this cheeky comparison made by technology analyst Benedict Evans: Accenture’s AI business has $3.7 billion in annual revenue, and the company makes more money from its AI consulting business than OpenAI makes from developing AI. Perhaps some restraint is in order.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *