The problem with AI neural networks, and more predictions about the future of technology

Mark R. Anderson of Strategic News Service, the Future in Review Conferences and Pattern Computer.

Popular methods of artificial intelligence have an “explainability problem” — the inability to see exactly what’s happening between input and output, and this challenge will prevent major new advances from neural networks.

That’s one of the predictions for 2023 from Mark R. Anderson, founding chair and CEO of Strategic News Service (SNS) and the Future in Review Conferences. An author, scientist, and entrepreneur, he’s known for using pattern discovery to accurately predict the future, and his predictions are one reason SNS has been followed closely in the industry since 1995.

Anderson acknowledges a vested interest in his take on neural networks, as the CEO of machine learning company Pattern Computer, which uses an alternative approach, focusing on pattern recognition.

Applications of Pattern’s technology include a rapid COVID-19 test based on patterns in light from spit, and drug combinations believed to kill breast cancer cell lines in culture. (For more on Pattern, read this August 2022 GeekWire story about the company by Charlotte Schubert.)

While neural networks have created incremental improvements in prediction rates, Anderson says this “black box” problem looms as a fundamental obstacle.

“This tool is a blunt-force tool,” Anderson says. “You’re not going to base your future business on the fact of having an incremental improvement in a prediction rate.”

On this episode of the GeekWire Podcast, he answers questions about two more of his predictions for 2023: skeptical takes on the near-term future of autonomous systems, plus virtual reality and the “metaverse.”

On the subject of the metaverse, Anderson introduces the term “MegaFUP,” which stands for Mega Failure to Understand Potential. It’s a reference to CEOs making disproportionately large bets on technologies with limited potential. Exhibit A for metaverse MegaFUP, he says, is Meta’s Mark Zuckerberg.

“Anyone who’s ever studied this stuff closely is fully briefed on the nausea effects, the effects of dizziness,” he says, citing the challenges encountered by Microsoft’s HoloLens in the military as the latest example. “And yet, Zuckerberg has put billions and billions and billions of dollars in the hope that that’s not true. It is true.”

Bringing it full circle, does Anderson think Microsoft’s big bet on OpenAI is a MegaFUP?

“Maybe it’s the right move at the right time for Microsoft. But what really we’re seeing is an arms race, a weapons race between the big cloud guys, because the amount of money at stake is indescribable,” he says.

“Is GPT the answer to their problems? Probably not,” he adds. “So temporarily, it’s a new weapon to further differentiate against the other three kids in the fight. But how long would that particular advantage last? Two years, three years. But you’ve got to have more than that if you really want to win.”

Anderson’s full list of predictions is available to SNS subscribers. Anderson is the featured speaker at an event on Thursday, Feb. 9 in Bellingham, Wash., benefitting the Rotary Club of Bellingham. Tickets available here.

Listen above, or subscribe to GeekWire in Apple Podcasts, Google Podcasts, Spotify or wherever you listen.

Podcast edited by Curt Milton.

Read More

Todd Bishop