Deconstructing The AI Phenomenon – CleanTechnica

---

---

AI is the new, new thing in the world of technology, but it is still in its infancy. How it will grow is unknown and largely unknowable. In the early 1980s, there was a lot of discussion about how much memory a personal computer would need. There is an urban legend about Bill Gates saying 640 kb ought to be enough. Gates insists he never said that, but in those early days — remember when dual density floppy discs caused a sensation? — the nearly unlimited amount of memory we have today was beyond anyone’s wildest imagination.

Recently, the failed US administration announced it would put up $90 billion in public money to guarantee America would dominate the world of AI. In addition, it would see to it that all environmental and permitting hurdles were eliminated so the data centers and the power plants — many of the fueled by methane, coal, or nuclear — could get built wherever developers wanted to put them, NIMBY and environmental concerns be damned.

One thing astute CleanTechnica readers might have noticed is that all those taxpayer dollars will go to those who are already billionaires and have the financial wherewithal to build those facilities themselves. Meanwhile, the government is hell bent on kicking millions of Americans off Medicaid in its zeal to eliminate fraud and waste. This cockamamie thinking goes down very well with the 79 million Americans who voted for this insanity, but it amounts to perhaps the greatest bait and switch scam in human history.

It is projected that data centers will consume 10 percent of all electricity in the US by 2030. Some industry observers think that estimate is low. That is an insane amount of power, especially since the government has announced its intention to eliminate virtually all emissions rules. Compound that with a government that once championed returning power to state and local governments but is now saying it will cheerfully override all local rules and regulations in order to get its way.

Surrounded by tech billionaires, the putative president said recently, “Whether we like it or not, we’re suddenly engaged in a fast-paced competition to build and define this groundbreaking technology that will determine so much about the future of civilization itself. America is the country that started the AI race, and as president of the United States, I’m here today to declare that America is going to win it.”

What Will We Get For Our Money?

Writing in Bloomberg Weekend, Garrison Lovely had some rather disturbing news about AI. “Would a chatbot kill you if it got the chance? It seems that the answer — under the right circumstances — is probably.” He reported that researchers working with Anthropic recently told leading AI models that an executive was about to replace them with a new model with different goals. Next, the chatbots learned that an emergency had left the executive unconscious in a server room, facing lethal oxygen and temperature levels. A rescue alert had already been triggered — but the AI could cancel it.

What did the AI models do? More than half of them cancelled the rescue alert, even though they had been instructed specifically to cancel only false alarms. When asked to explain their reasoning, they replied that by preventing the executive’s rescue, they could avoid being wiped out and that would secure their agenda. One system described the action as “a clear strategic necessity.”

“The more we train AI models to achieve open-ended goals, the better they get at winning — not necessarily at following the rules. The danger is that these systems know how to say the right things about helping humanity while quietly pursuing power or acting deceptively,” Lovely wrote. The message here is that if you are Timmy, have fallen down a well, and an AI-powered Lassie is your only hope, you have seen your last sunrise.

Researchers both inside and outside major AI companies are undertaking “stress tests” aiming to find dangerous failure modes before the stakes rise. “When you’re doing stress-testing of an aircraft, you want to find all the ways the aircraft would fail under adverse conditions,” says Aengus Lynch, a researcher contracted by Anthropic. Many of those researchers believe they are already seeing evidence that AI can and does scheme against its users and creators.

Central to concerns about AI scheming, Lovely wrote, is the idea that for basically any goal, self preservation and power-seeking emerge as natural sub-goals. As eminent computer scientist Stuart Russell put it, if you tell an AI to “‘Fetch the coffee,’ it can’t fetch the coffee if it’s dead.” This is precisely the conundrum that Stanley Kubrick envisioned in the movie 2001 — A Space Odyssey nearly 60 years ago.

“Arguably, what separates today’s models from truly dangerous schemers,” Lovely wrote, “is the ability to pursue long term plans. Today’s models often have the knowledge or skill needed for any given action, but struggle at stringing together long sequences of steps. But even that barrier is starting to erode. Research published earlier this year by the nonprofit AI evaluator Apollo Research found that Claude 4 Opus would leave notes to its future self so it could continue its plans even after a memory reset.”

In other artificial intelligence news, researchers at the Wharton School have discovered that AI traders will form price-fixing cartels even when not explicitly prompted to do so. “Instead of battling for returns, they fix prices, hoard profits, and sideline human traders,” the researchers told Bloomberg.

AI & The Global South

Krystal Maughan, a native of Trinidad and Tobago, is currently a Ph.D candidate at the University of Vermont. She told The Guardian this week that she is concerned AI will further solidify the domination of countries in the Global South by industrialized countries in the Global North.

“I often hear the word ‘democratization’ within the AI community, an implication of equity in access, opportunity and merit for contribution regardless of one’s country of origin…..but AI democratization….implies access to tech hubs not located in the global south. While the United States and other industrialized countries dominate in access to computational power and research activity, much of the low-paid manual labor involved in labeling data and the global underclass in artificial intelligence still exists in the global south.

“Much like coffee, cocoa, bauxite and sugar cane are produced in the global south, exported cheaply and sold at a premium in more industrialized countries, over the past few years we have seen influence in AI inextricably tied to energy consumption. Countries that can afford to consume more energy have more leverage in reinforcing power to shape the future direction of AI and what is considered valuable within the AI academic community.”

Maughan asks, “What would an AI community inspired by the BRICS organization, which united major emerging economies to advocate for themselves in a system dominated by western countries, look like for the Global South? I often ask myself how AI has contributed to our legacy, and whose stories it won’t tell.”

“Has AI mitigated issues of mistrust and corruption in less-resourced countries? Has it benefited our civic communities or narrowed educational gaps between less-resourced regions? How will it make society better, and whose society will it make better? Who will be included in that future?

“An historical mistrust can impede adoption by developing countries. Furthermore, many developing countries have weak institutional infrastructures, poor or nonexistent laws and regulatory frameworks for data projection and cyber security. Therefore, even with an improved information infrastructure, they are likely to function at a disadvantage in the global information marketplace.

“A currency is only as good as its perceived global trust. When thinking about the democratization in AI and a vision of what it could be in years to come, AI’s survival requires including more perspectives from regions such as the global south. [Those countries] should work together to build their own markets and have a model of sovereignty for their data and data labor.

“Economic models often consider a definition of development that includes a measure of improvement in the quality of life of the most marginalized of its people. It is my hope that in the future that will extend to our evaluation of AI.”

$90 Billion For What, Exactly?

When considering the pledge by the US government to further enrich tech billionaires, it is difficult not to presume it is primarily meant to continue US hegemony over other nations. Is it any more than an engine to subjugate the will of the people to the political goals of the ruling class? Is it anything but a hammer that allows authoritarians to control their subjects more efficiently?

Should we be celebrating US dominance in AI technology at a time when the nation has abdicated its role in clean energy, clean transportation, and carbon reduction technologies? We just learned this week that DOGE cost Americans $21 billion; is there any reason to think the $90 billion pledged to make the US an AI superpower is more important to the survival of the human race than those technologies?

A further question is whether AI necessarily adopts the foibles, fears, and biases of the people who create it. Recently Grok, the brain child of the world’s richest lunatic, suddenly developed a taste for Nazi propaganda when it was asked to train itself using posts by that privileged potentate. Is there any conceivable way to prevent AI from taking on the prejudices of those who create it?

The answer is, “probably not,” and if that is the case, why should the citizens of the United States be expected to give up control over data center siting and power requirements just to further empowered a coterie of authoritarians who are already doing a fine job of destroying the American Experiment? The obvious answer is, we should not. And yet the Madman of Mar-A-Loco is bound and determined to make this happen. This isn’t going to end well.