Understanding AI’s ‘Ghost In The Machine” Is Vital Amid Vast Spending

While most of the world’s biggest companies burn through hundreds of billions of dollars in the race to achieve general artificial intelligence, a documentary called Ghost in the Machine debuted this week at the Sundance Film Festival, questioning both AI’s dubious roots in eugenics and race “science,” its worker exploitation in slums worldwide, and the possibility that all this spending is kinda just hooey.

“It’s machine learning. The only difference now is they have massive amounts of compute,” director Valerie Veatch told me in a recent interview. “This film is pushing back on this narrative on super intelligence: that it’s going to be this AI Doomer, god-like entity, or it’s going to be this terrible thing that’s going to ruin humanity. ”

Indeed, getting more and more compute to feed the AI models is the big push these days by trillion-dollar companies, which collectively plan to spend an estimated $1.4 trillion to build dozens of U.S. data centers, despite concerns about the limits of available water and power to run all those centers, and whether the result will actually achieve what AI’s lead hypsters say will result.

Ghost, which runs nearly two hours, is organized around loosely connected “chapters” examining the deep roots, current problems, and uncertain future of AI.

The film’s first half or so traces how we got from Francis Galton’s attempts to use his cousin Charles Darwin’s new theories of evolution to justify British imperialism by measuring skulls to the U.S. eugenics movement copied so diligently by Nazi Germany to the birth of statistics, whose founding scientist Karl Pearson was both a Galton protege and a notably vile racist.

The lineage continues through efforts to measure intelligence, such as the Binet-Simon test that generated an IQ score, through to Alan Turing’s creation of early computer systems to create beyond-human intelligence. The computer industry was largely founded in proto-Silicon Valley after lifelong Palo Alto resident William Shockley co-developed the transistor, won a Nobel Prize, and launched his company in the area, spawning notable spinouts such as Fairchild Semiconductor and later Intel.

Shockley also was a vocal and notorious scientific racist and eugenicist, his views amplified by the early venture capitalists at Richard Draper’s Pioneer Fund. The push for super intelligence sits rather blithely next to the beliefs of those tech masters of the universe trying to max out their looks and lifestyle for eternal life.

A late chapter explores the involvement of big AI organizations in recruiting desperate slum dwellers in Nairobi, Buenos Aires, India and elsewhere to do the grinding input work of training AIs to be more human, and a campaign in Kenya to push back on the conditions that resulted.

The last section of the documentary details the uncertainty of whether all this spending will actually result in the goal its biggest backers expect. A clip of shows OpenAI CEO Sam Altman being asked what he’ll do when general artificial intelligence is here. He doesn’t know, he admits; he’ll probably ask that GenAI tool to build an even better one that can come up with a solution.

Veatch’s documentary debuted the week that several of the world’s biggest companies, all deeply entwined with AI in a desperate race for primacy, announced their quarterly earnings, and their expected continued vast spending on capital expenditures.

Meta, for instance, guided that it will spend $115 billion to $135 billion this year to fuel its lagging AI success with the open-source Llama large language model. Llama has underperformed even though it’s well placed to exploit AI tools for Meta’s ad-supported social media and messaging services.

Frequent CNBC contributor Brad Gerstner, a prominent Silicon Valley backer of AI investment, used his bully pulpit on the cable channel to say “if anything we have a compute shortage. Sam Altman has been telling us for two years that we don’t have enough compute and now there’s a shortage.”

At one point in my conversation with Veatch I suggest that her film wants viewers to use an important variant of science-fiction author Theodore Sturgis’ epigrammatic “Ask the next question.” Instead, maybe we should “Ask different questions” about AI and what we’re trying to achieve with all these expenditures.

She lights up, and says she might steal it. Later, I double-check my reference, and find plenty of verification from sites such as this one, which features a photocopy of a 1967 essay on that topic by Sturgis, a Star Trek screenwriter and author of 11 books, among much else.

I also checked what ChatGPT thought was the source. It asserted that the phrase was most associated with Isaac Asimov, the hugely prolific chemistry professor who wrote more than 200 science and science-fiction books in the same mid-century era as Sturgis. Asimov, who wrote extensively about non-human intelligence in books and short stories such as I, Robot, appears in the first scene, and elsewhere, of Veatch’s movie.

Asked about Asimov’s involvement in “Ask the next question” ChatGPT asserted that it “encapsulates (Asimov’s) belief that progress comes not from final answers, but from continually questioning assumptions and pushing inquiry forward The idea also resonates strongly with his classic short story The Last Question (1956), which centers on humanity repeatedly asking ever more profound questions of an evolving supercomputer—though the exact phrase appears more in his nonfiction commentary than verbatim in the story.”

So, just as Veatch’s film suggests, ask the next question, ask different questions, also double check the answers you’re getting from your AI tools, because they can be wrong. In similar fashion, ChatGPT asserted Veatch’s husband “probably” was a sociolinguistics professor with King’s College London. He is not, Veatch said.

That said, Veatch’s film features a wide array of computer scientists, historians, philosophers, many of them people of color, who have a distinctly less hyped-up view of both AI’s promise and its problematic roots. At least someone is thinking to ask different questions about one of the biggest issues of our time.