Its popularity has contributed to the start of what many in the technology industry are calling the AI arms race, as major technology companies such as Microsoft and Google look to develop and launch their own artificial intelligence models. These are trained on large amounts of data and therefore require large amounts of computing power.
“Currently, only a handful of companies have the resources necessary to create and deploy these large-scale AI models at scale. We need to recognize that we are giving too much power to people,” said Meredith Whitaker. The president of encrypted messaging app Signal said in an interview with CNBC last week.
“We should be really concerned that a small group of companies whose primary focus is profit and shareholder interest are making such socially important decisions.”
Whittaker previously worked at Google for 13 years, but became disillusioned in 2017 when he learned the search giant was working on a controversial contract with the Department of Defense known as Project Maven. Mr. Whittaker became increasingly concerned that Google’s AI could be used in drone warfare, and helped organize a walkout involving thousands of Google employees.
“AI as we understand it today is fundamentally a technology derived from the centralized power and control of corporations,” Whitaker said.
“It is built on intensive resources amassed by a small number of large technology companies, primarily based in the US and China, through the surveillance advertising business model, with powerful computational infrastructure and vast amounts of data, It’s given them a huge market to draw from’ that data, and the ability to process and structure that data in ways that help create new technologies. ”
Whittaker is not alone in this view.
Frank McCourt, former owner of the Los Angeles Dodgers baseball team, now runs Project Liberty, an organization that motivates engineers and policymakers to “build a more responsible approach to technology development,” according to its website. are doing.
McCourt also thinks AI could give tech giants too much power. “There are basically five companies that have all the data,” he said, without naming the companies.
“Large language models require a lot of data. If we don’t make changes here, the game is over… Only these same platforms will prevail. And they will be the beneficiaries. ” McCourt said in an interview with CNBC last week. .
“Sure, people come and build small things on top of those big platforms. But it’s the underlying big platforms that manage this data that are the winners.”
Whittaker and McCourt are among those who feel that users are losing control of their data online and that it is being used by tech giants to make profits.
“Big tech and social media giants are causing serious damage to our society,” McCourt’s Project Liberty manifesto states. And he believes AI could make this even worse.
“Don’t be fooled. Generative AI is a fancy name for getting more power out of your data,” McCourt said in an interview with CNBC.
Generative AI is the technology that writes applications like ChatGPT. The models powering these apps are trained on vast amounts of data.
“Generative AI built on large-scale language models is essentially an enhanced, or more powerful, version of current technology with a fancy name: a centralized, authoritarian surveillance technology. .And I’m against that, and I think there’s a lot of harm being done in the world right now.” McCourt said.
Tim Berners-Lee, the inventor of the web, has also expressed concern about the concentration of power among the tech giants.
For Wikipedia founder Jimmy Wales, the state of social media is of particular concern right now.
But when it comes to AI, he feels that while the tech giants are currently leading the way, there is room for disruption.
In an interview with CNBC last week, Wales pointed to a leaked Google memo this year in which researchers at the US tech giant said the company had no “moat” in the AI industry and cited the threat of an open source model. He said he is doing so. These are AI models that anyone can develop and add to, rather than being owned by a single entity such as Google or Microsoft. These could allow competing AI applications to be created without the large amount of resources that are currently required.
“The models that are out there and the open source models that anyone can download and run on a few machines that startups can invest in. [just] $50,000 in training…that’s not a big deal at all. It’s really impressive,” Wales added.
Source link
