The Kaleidoscope of New Inventions

Chapters about the future age particularly badly. They easily mistake a laboratory success for an industry, a presentation for a power station and a funding round for physical feasibility.

This chapter therefore begins with three categories:

• available: technically and commercially deployable;
• demonstrated: physically or technically shown, but not yet economic at broad scale;
• visionary: a plausible development path with major unresolved requirements.

A kaleidoscope is the right instrument for the task. It does not declare one technology to be the future. It shows how the same pieces can form successive new patterns. In the electric century, those pieces are energy, AI, robotics, biology, quantum physics and networked production. Each technology has its own development curve. The larger upheaval begins where they interlock.

The nearer future: better combinations

The largest advances of the next few years may look less impressive than their advertising. More transmission, stronger distribution grids, standardised connections, better transformers, thermal storage, batteries, flexible plants and controllable demand may achieve more than one miraculous machine.

Nuclear power is returning to planning as well. At the end of 2024, global operational nuclear capacity stood at 377 gigawatts. The IAEA’s projections for 2050 range from 561 to 992 gigawatts — an unusually wide spread that reveals less about certainty than about the importance of political, financial and industrial conditions.^[10] Life extensions, large new plants and perhaps small modular reactors can provide firm low-carbon power. But “modular” is not a synonym for cheap, and an announcement does not generate a kilowatt-hour.

Advanced geothermal systems may make dependable heat accessible across more geographies. Long-duration storage may bridge days rather than hours. AI can improve materials research, maintenance, weather prediction and grid operations. These technologies do not allow escape from the system. They make the system denser, smarter and more dependent on maintenance.

When the tools converge

The first great role of AI does indeed lie in optimisation. It can forecast generation and demand, coordinate storage, recognise fault patterns, schedule maintenance, simulate grid variants and align millions of small decisions more quickly than a human control room could manage alone.

That is important. But it is the role of an exceptionally capable operations manager. The larger role begins when AI does more than administer existing systems: when it produces designs, narrows search spaces and directs experiments.

A new chain of invention then becomes possible:

AI designs → classical computers calculate → quantum processors solve specialist problems → robots experiment → humans verify → industry scales

No link replaces all the others. A language model cannot talk a solid electrolyte into existence. A quantum computer does not manufacture a battery. A robot does not know which research objective society should value. Together, however, these tools may radically shorten the route from a question to a material result.

The night when a thousand laboratories kept working

A second great wave may transform research itself. AlphaFold demonstrated that AI can open a search space that humans alone could scarcely navigate. AlphaFold 3 models interactions among proteins, nucleic acids, small molecules and other biological components. In materials science, GNoME calculated millions of possible crystal structures. Autonomous laboratories have already demonstrated closed loops in which software plans experiments, robots execute them, instruments measure results and the system selects what to try next.^[11]

Today these laboratories are narrow, expensive and fallible. At a plausible next stage, they could collaborate across continents. A team defines an objective: a cheaper storage material, a more durable membrane for desalination, or a catalyst that uses fewer critical minerals. AI systems explore alternatives and propose experiments. Laboratories in different time zones test candidates. People investigate the surprising and contradictory results.

The night would become more productive without requiring a doctoral student to spend every night in the laboratory.

The point is not that the machine becomes “the scientist”. Science does not begin with the number of experiments completed. People decide which problem matters, which risks are acceptable, how a result must be verified and whether an application should be pursued. AI enlarges the field of hypotheses and accelerates the return of evidence. In doing so, it can encourage something profoundly human: the pleasure of asking a good question and being surprised by reality.

A small university, hospital or medium-sized company might temporarily join a global research network rather than own every instrument. Scientific capability would not automatically become equal. But it could become far less confined to a few buildings and metropolitan centres.

When matter becomes more calculable

Quantum computing could become an unobtrusive but exceptionally powerful specialist instrument in this chain. A quantum computer is not a particularly fast ordinary computer, and it is unlikely to replace today’s AI data centres. It is closer to a scientific accelerator: expensive, delicate and superior only for certain questions — especially where quantum mechanics itself makes classical approximations unreliable.

Error correction is the decisive bottleneck. At the end of 2024, Google’s Willow processor gave the first unambiguous demonstration of quantum error correction below the necessary threshold: a larger logical memory protected quantum information better than a smaller one. In 2026, a reinforcement-learning system learned to adjust control parameters during operation and improved stability against injected drift by a factor of 3.5.^[12] AI is no longer optimising only the power grid. It is helping to keep the quantum computer calculable at all.

A cautious transition is also visible in applications. An experiment published in 2025 examined complex quantum dynamics beyond the reach of known classical simulation methods and demonstrated one practical building block in a tightly bounded learning problem. The authors themselves described the real application as future work. IBM, meanwhile, is targeting a 2029 system with 200 error-corrected logical qubits and 100 million operations. That is a roadmap, not a fulfilled promise.^[13]

Imagine a laboratory network in 2040 searching for a solid electrolyte that uses fewer critical materials, charges quickly and remains stable in heat. An AI explores millions of possible structures. Classical supercomputers calculate most of them. Only candidates for which strong electronic interactions make ordinary approximations unreliable are sent to a quantum processor.

The quantum computer does not return a finished product. It identifies three compounds worth an experiment. An autonomous laboratory makes them. The first fractures, the second conducts too poorly, and the third behaves differently from the prediction — but interestingly enough to justify another series of trials.

The quantum computer did not invent the battery. At one decisive point, it removed a bad approximation.

The same combination could produce catalysts, desalination membranes, superconductors or light-activated cancer drugs. AI proposes what might be possible. The quantum computer calculates part of what physics permits. The laboratory asks matter whether both are right.

The quantum computer will probably replace neither the laptop nor Excel. It may, however, find the material from which the next laptop battery is made.

When biology becomes a design space

A similar convergence is beginning in medicine. Cancer is not a single adversary but a collective name for many diseases, each capable of changing even within one patient. There will probably be no single AI that “cures cancer”. A more plausible chain links early detection, molecular profiling, drug and protein design, automated laboratories, personalised cell or gene therapy and continuous monitoring of outcomes.

Every link already has a precursor. AI is used in drug discovery, in predicting treatment response and in analysing tumour data. CAR T cells can eliminate some advanced blood cancers for years and appear to cure some patients. Personalised neoantigen vaccines have produced durable immune responses in small early trials. The first approved CRISPR therapy now treats severe sickle cell disease by editing a patient’s own blood stem cells.^[14]

AI would not itself be the cure. It would be the system that narrows search spaces, recognises combinations, prioritises experiments and learns faster from every tested result. A cure would emerge from the interaction of biology, data, laboratory, energy, manufacturing, doctor and patient.

The limiting factor would not be knowledge alone. Personalised cell and gene therapies would have to be manufactured safely, quickly and affordably. Medicine that can theoretically save everyone but reaches only a few would have solved a research problem and created a civilisational one.

When humans acquire electrical spare parts

The cyborg of the electric century will probably not first appear as an armoured superhuman. It will arrive as a patient who wants to hear, see, speak, grasp or walk again.

Cochlear implants already stimulate the auditory nerve electrically. In a clinical study published in 2025, a wireless retinal implant helped people with previously untreatable central vision loss to recognise letters, numbers and words again. The image remained limited and had to be learned. But it was an image.^[15] Trial participants control bionic legs through preserved muscle signals; implanted brain–computer interfaces translate intended speech in paralysed people into an audible voice almost in real time.^[16]

Advanced battery technologies could become one of the quiet keys. An implant must receive power for years without generating harmful heat, becoming toxic or requiring surgery for every battery change. Higher energy density, solid-state cells, wireless charging and energy harvested from light, motion, ultrasound or body chemistry compete for the same precious space beneath the skin. Some implants may operate without a conventional battery at all.^[17]

The most important chip in the cyborg may not be the most intelligent. It may be the one that receives reliable power for twenty years. An electronic body part should not, after all, enter power-saving mode after lunch.

Technical possibility creates new rights. Who owns the data from an eye or brain interface? Who may update the software? Can a manufacturer end support after ten years? An artificial knee may wear out. It should not stop working because a subscription expired. The cyborg question is therefore less how much machine may enter a human being than how much control over the body must remain with the person.

When machines go to war

Not every new activity of the electric century will be desirable. Robots and drones will do more than build, map, care and explore. They will fight.

A battle in 2035 may begin before a soldier sees an opponent. Small aircraft search for radar positions, others jam communications, unmanned vessels defend or threaten ports, ground robots clear routes, and software redistributes tasks within a swarm as individual machines are lost. Humans set the mission. Yet individual movements, evasive manoeuvres and counterattacks may unfold faster than any person could approve step by step.

This direction is not remote speculation. Military programmes are already organised to field thousands of inexpensive unmanned systems with varying degrees of autonomy; research programmes are developing swarm tactics for combat. The same capabilities can find wounded people, control bleeding or carry casualties out of dangerous areas.^[18]

The humanitarian gain is real. A machine can enter a minefield, search a burning structure or retrieve an injured soldier without exposing another person to the same danger. Distance also changes the threshold for violence. If fewer soldiers are put at risk and lost systems can be replaced cheaply, military force may become easier to begin, easier to prolong and harder for the public to see.

The decisive boundary therefore does not lie between crewed and uncrewed. It lies between machine assistance and the delegation of irreversible decisions. A system may navigate, evade or detect interference. Identifying a human being as a target and authorising lethal force remain questions of law, evidence and responsibility.

The battery does not distinguish between a bionic hand and an attack drone. The model feels no difference between a rescue route and a targeting run. Humans and institutions must create that moral distinction — and preserve an auditable record of it even when war begins to calculate in milliseconds.

When electricity learns to fly

Wireless power did not supply a city in 2025. It did, however, make popcorn.

At a test site in New Mexico, the US research agency DARPA sent a laser beam across 8.6 kilometres. More than 800 watts reached the receiver for 30 seconds. Across the campaign, the system transferred more than one megajoule; some of the energy was used, for effect, to make popcorn.^[19]

That may sound less elevated than the beginning of a new civilisation. But the future often becomes credible at the moment it first performs something remarkably ordinary.

The principle is older. In a 1975 NASA/JPL experiment at Goldstone, more than 30 kilowatts of direct-current power was recovered from a microwave beam transmitted across 1.6 kilometres. In 2022, the US Naval Research Laboratory demonstrated roughly a kilowatt across one kilometre. And in 2023 Caltech’s MAPLE demonstrator transmitted microwave energy in orbit and directed a detectable signal to Earth.^[20]

The technical chain is:

Electricity → microwave or laser → directed beam → receiver → electricity

This is not energy Wi-Fi filling a room with invisible power. Energy is aimed at a known receiver. Lasers need line of sight and suffer from clouds, dust and atmospheric disturbance. Microwaves are more tolerant of weather but require large antennas and receiving surfaces. Both involve conversion losses, safety questions and substantial regulation.

The dream is Wi-Fi for energy. Physics currently requires a very large target.

Power beaming could still matter for remote stations, disaster zones, drones, islands, military logistics or space infrastructure. Its deeper significance would lie in making electrical energy addressable and dynamically steerable. In a limited sense, energy would acquire properties of information: a beam could be assigned, redirected or denied.

That returns us to the politics of the grid. The cable may disappear at one point. The power relationship remains.

Solar power stations in space

In orbit, sunlight is available more consistently and intensely than on the ground. Large solar systems could collect energy and beam it to Earth by microwave. Caltech has demonstrated individual elements. An economic power station remains far away.

A 2024 NASA study examined two potential systems entering service around 2050 and found them more expensive under its assumptions than sustainable terrestrial alternatives. Enormous progress would be required in launch cost, autonomous assembly, maintenance, durable lightweight structures and efficient power transmission.^[21]

The most important early market might therefore not be Earth. On the Moon, cables, fuel deliveries and long nights are exceptionally difficult. Power transmission between illuminated sites, relays and bases may acquire value there before it can compete with a mature terrestrial grid.

Fusion and abundance

Fusion remains the most powerful energy promise: widely available fuels, exceptional energy density and no long-lived highly radioactive waste of the kind produced by today’s fission reactors. The physical progress is real. A power station must nevertheless do more than produce a fusion reaction. It must extract energy continuously, preserve materials under neutron bombardment, remove heat, maintain components, handle tritium and sell electricity economically.

Fusion is therefore neither mere fantasy nor a basis for present-day planning. It is a serious visionary option with an open timetable.

If energy becomes exceptionally cheap and abundant, not all scarcity disappears. Land, attention, ecological sinks, skilled labour, political legitimacy and time remain limited. Energy abundance can make water, materials and transport easier to obtain. It can also expand the reach of human error.

Abundance does, however, change which dreams become reasonable. Desalination and water recycling could make dry regions more resilient. Materials could be separated and reused through more energy-intensive processes. Greenhouses could control climate and nutrients more precisely. Low-carbon fuels could be made for applications that resist direct electrification. Compute for science, languages, education and planetary models would no longer have to compete quite so fiercely with every other kilowatt-hour.

The philosophical shift would be considerable. Politics and economics have learned to administer scarcity. They would also have to learn how to cultivate possibility. The question would no longer be only who receives how much, but which projects enlarge the field of human capability.

A society with more energy may discover that its scarcest resource is no longer power. It is the ability to agree on goals worth pursuing.

Electric expansion

Space flight ultimately is energy infrastructure. Rockets need concentrated energy, satellites need power, communications need compute, lunar bases need heating and cooling, and resource extraction needs machines. The farther human activity moves from Earth, the less it can rely on frequent fuel delivery and improvised repair.

The long chain is:

Energy → electricity → automation → autonomous infrastructure → expansion

From this perspective, SpaceX and other space companies are not merely transport providers. They may become builders of an infrastructure in which energy, compute, communication and production converge beyond Earth. Whether that creates a new civilisational stage or an extraordinarily expensive extension of terrestrial power struggles remains open.

The enthusiastic possibility is different. Space infrastructure could become the first field of work conceived from the outset as global, robotic and electric. Telescopes, power systems, scientific instruments and construction robots could be designed and operated by teams distributed across several continents. People would not all have to enter a hostile environment in order to be present there. They could see, act and build through machines — and later decide where to follow them.^[22]

That would not make Earth less important. On the contrary. Those who see a small inhabited planet from outside may understand more clearly that oceans, atmosphere and biological diversity are not local departments.

More energy, learning machines, more calculable matter, repairable bodies and autonomous tools: the arsenal of possibility is expanding. Yet a civilisation does not consist of its arsenal, even a peaceful one. It begins where people do something with their inventions.

The kaleidoscope has shown what may enter the world technically. We must now enter the place where technology is used, shared, loved, misunderstood, improved and sometimes tried simply for pleasure.

Sources and notes

1. International Atomic Energy Agency, Energy, Electricity and Nuclear Power Estimates for the Period up to 2050, 2025 edition: https://www.iaea.org/publications/15942/energy-electricity-and-nuclear-power-estimates-for-the-period-up-to-2050
2. Abramson et al., Accurate structure prediction of biomolecular interactions with AlphaFold 3, Nature 630, 2024: https://www.nature.com/articles/s41586-024-07487-w; Google DeepMind, Millions of new materials discovered with deep learning, 2023: https://deepmind.google/blog/millions-of-new-materials-discovered-with-deep-learning/; Szymanski et al., An autonomous laboratory for the accelerated synthesis of novel materials, Nature 624, 2023: https://www.nature.com/articles/s41586-023-06734-w
3. Google Quantum AI and Collaborators, Quantum error correction below the surface code threshold, Nature 638, 2025: https://www.nature.com/articles/s41586-024-08449-y; Google Quantum AI and Collaborators, Reinforcement learning control of quantum error correction, Nature 655, 2026: https://www.nature.com/articles/s41586-026-10759-2
4. Google Quantum AI and Collaborators, Observation of constructive interference at the edge of quantum ergodicity, Nature 646, 2025: https://www.nature.com/articles/s41586-025-09526-6; IBM, IBM lays out clear path to fault-tolerant quantum computing, corporate roadmap: https://www.ibm.com/quantum/blog/large-scale-ftqc; NIST, Quantum Computing Explained: https://www.nist.gov/quantum-information-science/quantum-computing-explained
5. National Cancer Institute, Artificial Intelligence and Cancer: https://www.cancer.gov/research/infrastructure/artificial-intelligence; NCI, CAR T Cells: Engineering Patients’ Immune Cells to Treat Their Cancers: https://www.cancer.gov/about-cancer/treatment/research/car-t-cells; NCI, Neoantigen Vaccines Keep Kidney, Pancreatic Cancer at Bay, 2025: https://www.cancer.gov/news-events/cancer-currents-blog/2025/neoantigen-vaccine-pancreatic-kidney-cancer; US FDA, FDA Approves First Gene Therapies to Treat Patients with Sickle Cell Disease, 2023: https://www.fda.gov/news-events/press-announcements/fda-approves-first-gene-therapies-treat-patients-sickle-cell-disease
6. National Institute on Deafness and Other Communication Disorders, Cochlear Implants: https://www.nidcd.nih.gov/health/cochlear-implants; Holz et al., Subretinal Photovoltaic Implant to Restore Vision in Geographic Atrophy Due to AMD, New England Journal of Medicine, 2025: https://www.nejm.org/doi/10.1056/NEJMoa2501396; Stanford Medicine, study summary, 2025/2026: https://med.stanford.edu/news/all-news/2025/10/eye-prosthesis.html
7. Song et al., Continuous neural control of a bionic limb restores biomimetic gait after amputation, Nature Medicine 30, 2024: https://www.nature.com/articles/s41591-024-02994-9; NIH, Brain-computer interface restores natural speech after paralysis, 2025: https://www.nih.gov/news-events/nih-research-matters/brain-computer-interface-restores-natural-speech-after-paralysis
8. Nair et al., Miniature battery-free bioelectronics, Science 380, 2023: https://www.science.org/doi/10.1126/science.abn4732; Zhou et al., Wireless battery-free ultrathin lithium-niobate resonator as an implantable biomedical sensor, Nature Communications, 2025: https://www.nature.com/articles/s41467-025-67413-0; Abbas et al., Development of an efficient mid-field wireless power transfer system for implantable medical devices, Scientific Reports, 2025: https://www.nature.com/articles/s41598-025-99609-1
9. US Department of Defense, Replicator Initiative, 2024: https://www.defense.gov/News/News-Stories/Article/Article/3657609/defense-innovation-official-says-replicator-initiative-remains-on-track/; DARPA, OFFensive Swarm-Enabled Tactics: https://www.darpa.mil/research/programs/offensive-swarm-enabled-tactics; DARPA, Medical Swarm Robotics for Extraction and Life-Saving Interventions, 2026: https://www.darpa.mil/research/programs/medical-swarm-robotics-for-extraction-and-life-saving-interventions
10. DARPA, “DARPA program sets distance record for power beaming”, 16 May 2025: https://www.darpa.mil/news/2025/darpa-program-distance-record-power-beaming
11. NASA Technical Reports Server, Goldstone demonstration: https://ntrs.nasa.gov/api/citations/19810008041/downloads/19810008041.pdf; US Naval Research Laboratory, SCOPE-M: https://www.nrl.navy.mil/Media/News/Article/3004608/nrl-conducts-successful-terrestrial-microwave-power-beaming-demonstration/; Caltech, MAPLE: https://www.caltech.edu/about/news/in-a-first-caltechs-space-solar-power-demonstrator-wirelessly-transmits-power-in-space
12. NASA Office of Technology, Policy, and Strategy, Space-Based Solar Power, 2024: https://www.nasa.gov/organizations/otps/space-based-solar-power-report/
13. ESA, New ESA connection to advance robotics for lunar exploration, 2025: https://www.esa.int/Space_in_Member_States/United_Kingdom/New_ESA_connection_to_advance_robotics_for_lunar_exploration; NASA, NASA Enables Construction Technology for Moon and Mars Exploration, 2025/2026: https://www.nasa.gov/directorates/stmd/nasa-enables-construction-technology-for-moon-and-mars-exploration/