Frankenstein Revisited: AI, AGI, ASI — and Humanity’s Oldest Technological Fear

Elon Musk’s $11 Billion Gift: Generosity, Political Influence, or Both? Elon Musk has pledged to give every child in America one SpaceX share, potentially distributing nearly $11 billion in wealth to the next generation. An extraordinary act of generosity. But with the United States approaching its November 2026 midterm elections, the timing raises uncomfortable questions. Children cannot vote. Their parents can. The proposed gift will be distributed through Trump Accounts, a government-supported investment programme associated with President Donald Trump. Could parents who believe their children will benefit develop greater goodwill towards the administration? Could that goodwill influence their votes? And here is the intriguing detail: the shares are reportedly scheduled for distribution in 2027, after the midterm elections. The potential political influence before November would therefore arise from the promise of a gift, rather than its delivery. Is this vote-buying? There is no established evidence of an improper exchange for votes. But political influence does not always require an explicit transaction. Philanthropy, public recognition and political advantage can coexist. In this episode of Artificial Intelligence to Artificial General Intelligence to Artificial Super Intelligence, we examine the ethics behind Musk's announcement, the financial risks of distributing individual SpaceX shares, potential tax implications and the growing relationship between concentrated technological wealth and political power. We also compare Musk's approach with Anthropic's public benefit initiative and explore a larger question: as artificial intelligence creates extraordinary fortunes, who should decide how that wealth benefits society? This is not about supporting or opposing Musk or Trump. It is about applying consistent standards of transparency, independence and accountability, regardless of political affiliation. An $11 billion promise deserves serious examination, not simply applause or suspicion. When a billionaire promises financial benefits to millions of children shortly before an election, how do we distinguish genuine generosity from the political goodwill it may generate? And would we ask the same questions if the donor supported the opposing political party?

Artificial Intelligence Is Not the Solution. The Problem Comes First. Are we using artificial intelligence to solve humanity's real problems, or simply making inefficient processes faster? As our AI to AGI to ASI Podcast celebrates 1,000 active listeners, a thought-provoking contribution from Anshika Rastogi inspired this episode: How can organisations use artificial intelligence to solve meaningful problems rather than simply introduce more tools? It is a question that deserves attention far beyond the technology industry. Businesses are investing billions in artificial intelligence, automating workflows, generating reports and accelerating decisions. Yet patients still experience fragmented healthcare, students struggle to receive individual support, critical infrastructure continues to fail, and businesses remain burdened by inefficient processes. Are we measuring technological adoption rather than meaningful progress? In this episode, we explore how artificial intelligence could address real societal challenges across healthcare, education, engineering, manufacturing, agriculture, transport, government and small businesses. We examine why identifying the root cause of a problem must precede selecting a technological solution. Drawing on established engineering principles, we consider the importance of evidence, quality assurance, accountability and measurable outcomes. But there is a deeper question. What happens when artificial intelligence evolves beyond being a tool? As autonomous agents increasingly coordinate workflows, interpret information and execute decisions, the relationship between humans and technology begins to change. Artificial intelligence may become an operational partner, but greater capability does not automatically deliver better judgment, ethical decisions or public trust. Who defines success? Who validates the evidence? Who remains accountable when an automated recommendation causes harm? And when productivity improves, who ultimately benefits? We also challenge the assumption that every problem requires artificial intelligence. Sometimes the better solution is improved leadership, simpler processes, stronger institutions, better training or investment in physical infrastructure. The future of artificial intelligence should not be judged by how many tools we deploy, how many prompts we generate or how quickly we produce information. It should be judged by the problems we solve, the harm we prevent, the opportunities we create and the lives we improve. From artificial intelligence to artificial general intelligence and potentially artificial superintelligence, technological capability will continue to advance. But human purpose, judgment and responsibility must remain central. The challenge is not simply to build more intelligent systems. It is to build a more intelligent approach to solving problems. Join us for a conversation about moving beyond artificial intelligence adoption towards meaningful, measurable and sustainable societal impact. If artificial intelligence could solve one real problem in your industry or community, what should it be, and how would you measure success?

Elon Musk has done it again. Tesla’s flying Roadster demonstration has captured global attention, but has he really demonstrated a flying car, or something potentially more valuable: the ability to sell belief in a future that has yet to arrive? In this episode of AI to AGI to ASI, we examine the engineering, economics and psychology behind Tesla’s spectacular demonstration. A vehicle that briefly launches into the air, rotates and lands is an impressive engineering achievement. But is it flight? And how far are we from a commercially viable flying car? We examine the physics of compressed-gas propulsion, the limitations of sustained flight, passenger safety and the enormous challenges separating an experimental prototype from a practical transport solution. We also consider a LinkedIn investigation by Christian Busch, who used Anthropic’s Claude to analyse the demonstration. His findings raise important questions, but can artificial intelligence reliably verify engineering claims without independently measured evidence? Beyond the technology lies a more compelling question: Is Musk selling a product, or selling the vision needed to finance its development? From SpaceX’s reusable rockets to Tesla’s long-delayed Roadster, Musk has repeatedly demonstrated the commercial power of ambitious engineering narratives. Such demonstrations can attract investors, talent and public enthusiasm long before a technology becomes commercially viable. But where does visionary leadership end and technological hype begin? The parallels with frontier artificial intelligence are striking. Impressive model demonstrations attract billions in investment, yet a successful experiment does not necessarily establish reliability, scalability or commercial readiness. This episode explores why engineering maturity, independent verification, provenance and transparency matter as much as innovation itself. Perhaps Tesla has not demonstrated the future of flying cars. Perhaps it has demonstrated something equally powerful: how belief in tomorrow can create economic value today. The question is not simply whether Elon Musk can make a car fly. It is whether investors and society can distinguish between genuine technological progress and the promise of what might one day become possible.

Microsoft’s reported MAI-Base-2 breakthrough may signal a major shift in the artificial intelligence industry: from competing primarily to build the most capable models toward competing to deliver the most useful intelligence with the least energy and computing power. The claim is significant because artificial intelligence is not weightless software. It depends on processors, data centres, cooling systems, water, minerals, factories, transmission networks and power stations. Every model trained and every response generated carries a physical cost. If Microsoft can achieve comparable or better performance using a fraction of the computing resources required by rival systems, the potential benefits could be substantial. Data centres might provide more services without proportional increases in electricity demand. Businesses could access advanced tools at lower cost. Smaller organisations could participate more easily in the artificial intelligence economy. Electricity networks could face less pressure, and the environmental impact of individual tasks could decline. However, the announcement should be treated as an important signal rather than a final answer. Microsoft’s claim concerns pre-training efficiency, which does not automatically prove an equivalent reduction in total electricity consumption. The energy required to train a model is only one part of its lifecycle. Everyday operation, known as inference, may consume significant resources when millions of users rely on the system. A model that is efficient to train may not be equally efficient to operate, while a more demanding model could potentially complete complex tasks with fewer repeated attempts. Meaningful comparisons therefore need to measure energy per successful task across the full system, not simply computing resources used during training. Efficiency also does not guarantee lower total consumption. If artificial intelligence becomes cheaper to operate, businesses and consumers may use it more frequently, while new applications may become economically viable. This could create a rebound effect in which efficiency gains encourage enough additional demand to offset the savings. Electricity consumption may continue rising even as individual models become more efficient. The environmental outcome will also depend on whether new demand is supplied by clean generation or fossil fuels, and whether data centres create additional pressure on local networks, water supplies and communities. Microsoft’s position makes the announcement especially important. The company develops models, designs specialised hardware, operates cloud infrastructure and delivers artificial intelligence services to businesses and consumers. This allows it to coordinate software and hardware more closely, potentially improving efficiency across the entire system. The future competition may therefore involve not only model intelligence, but also processor design, data-centre performance, cooling, networking, electricity partnerships and transparent measurement. That opportunity creates responsibility. Microsoft and its competitors should publish clearer information about the resources required to train and operate their systems. Independent researchers should be able to verify efficiency claims, and customers should be able to compare models using consistent measures of performance, cost, reliability and energy consumption. Governments must also ensure that artificial intelligence infrastructure does not undermine the affordability, reliability or fairness of electricity systems on which everyone depends. The artificial intelligence race is often described as a competition to build the most powerful mind. In the long term, it may also be a competition to build the most efficient body.

The White House Accord on Super Intelligence brings together the United States government and major artificial intelligence companies around a shared promise: frontier systems should be developed with stronger controls, independent review and greater accountability. But this episode asks a harder question: is that enough to create genuine trust? The discussion moves beyond the signatures and focuses on provenance, traceability, auditability and institutional independence. If an artificial intelligence system makes a consequential decision can we reconstruct what happened? Which model version was used? What data influenced the outcome? Which tools and permissions were available? What did the system do automatically, what did a human approve, and what evidence remains afterward? The episode argues that safety controls and trust are not the same thing. Safety attempts to prevent failure. Traceability helps reconstruct events. Provenance tells us where information came from. Audit tests whether the evidence supports the claims being made. But audit is not enough if the auditor is not genuinely independent. That raises a deeper question: independent from whom? If a technology company builds the system, selects and pays the auditor, controls the evidence and determines what is disclosed, the assurance process may still depend too heavily on the company being examined. Yet transferring oversight entirely to government does not automatically solve the problem. Governments have political priorities and their own use of it. The podcast therefore explores institutional separation. Industry develops the technology. Government creates laws and enforcement mechanisms. Independent institutions should help verify the evidence. These could include accredited auditors, standards bodies, universities and regulators. Using engineering quality assurance as a comparison, the episode asks why artificial intelligence should be held to a lower evidentiary standard than bridges, pipelines or aircraft. Mature engineering systems rely on records, inspections, version histories, independent reviews and preserved evidence. Artificial intelligence can be harder to investigate because prompts disappear, models change, websites are updated and databases are overwritten. The deeper concern is that artificial intelligence may increasingly influence infrastructure, finance, medicine, government, employment and everyday life while leaving behind less evidence than traditional engineered systems. The White House Accord may be an important beginning, but not a complete assurance framework. The real test will be whether artificial intelligence governance evolves beyond promises and reputations toward verifiable evidence, common standards, independent scrutiny and accountability. In the age of increasingly autonomous artificial intelligence, trust should not simply be requested by companies or granted by governments. It should be something that can be independently demonstrated.