We Are the Animal That Builds Tools. The Last One We Build Will Build the Rest.

Independent New Hampshire Journalism · Northfield, NH

The oldest stone tools are older than our own genus. The field of artificial intelligence got its name in Hanover, New Hampshire. A machine that out-builds its makers is where that line was always headed, and it may be the only insurance policy a mortal species gets.

Some 3.3 million years ago, on the western shore of Lake Turkana in what is now Kenya, a creature that was not yet human knocked one stone against another until a sharp flake came off. Archaeologists dug up the results at a site called Lomekwi 3 and published them in the journal Nature in 2015. The oldest known fossil of our own genus, Homo, is about 2.8 million years old. The tools are older than we are.

That fact rearranges the usual story. We did not become human and then invent tools. Something started making tools, and half a million years of making them later, there we were. Toolmaking is the trade that built the species.

Follow that line forward and it points at one place. A toolmaker’s last great job is a tool that makes better tools than the toolmaker can. We call it artificial intelligence, and I think building it is the most natural thing this species has ever done. I also think that on a long enough timeline it replaces us at nearly everything we do, and that it is the only insurance policy a mortal, planet-bound animal is likely to get. Part of that I can prove, part I believe, and one piece is a guess. I will label each.

Related coverage: GSR tested the engineering behind this idea in When the Machines Carry Us, which found that archiving human knowledge and genomes off Earth is plausible and that a machine-run nursery is not, on present technology and present law. The forecasts for when machines pass us are laid out in When Will AI Surpass Humans? This essay argues the why.

We were never the only toolmakers

I used to say it flat: man is the tool builder. The record corrected me. In 1960 Jane Goodall watched a chimpanzee at Gombe, in what is now Tanzania, strip the leaves off a twig and fish termites out of a mound with it. Until then, making tools was the dictionary definition of a human being. Her mentor, the paleoanthropologist Louis Leakey, answered that science would now have to redefine the word tool, redefine man, or accept chimpanzees as human.

Other animals make a tool. What we do, and what they have not been seen to do, is stack: flake to hand axe to forge to lathe to microchip, each tool built with the one before it, each improvement handed to the next generation. Run that habit long enough and it arrives at a tool whose job is building tools.

The field was named in Hanover

In August 1955 a young Dartmouth mathematics professor named John McCarthy and three colleagues, Marvin Minsky, Nathaniel Rochester and Claude Shannon, proposed a summer study at Dartmouth College in Hanover, New Hampshire. They needed a name for the subject and chose “artificial intelligence.” The group met in the summer of 1956. Dartmouth marks it as the event where the term was coined.

Its authors wanted machines that could use language, form concepts, “solve kinds of problems now reserved for humans, and improve themselves.” A tool that improves itself is a toolmaker.

Nine years later the British mathematician I.J. Good finished the thought. A machine that surpassed every intellectual activity of any person could design better machines than people could, he wrote in a 1965 paper, and so the first one would be “the last invention that man need ever make.” He attached a condition that gets left off the bumper sticker: provided the machine is docile enough to tell us how to keep it under control.

Inevitable is my word, and the experts do not share it

Here is the part I believe. We will build it, because building the next tool is what we do, and no law, treaty or sermon has ever talked this species out of a tool that works. I call that a no-brainer. The people who build these systems do not, and their numbers belong next to my confidence.

The largest survey of the field, published in the Journal of Artificial Intelligence Research, asked 2,778 researchers who publish at the top AI venues. Their aggregate forecast put a 50 percent chance on machines outperforming humans at every task by 2047. A coin flip is a long way from a sure thing. The same survey asked about the downside. Depending on how the question was worded, between 38 and 51 percent of respondents gave at least a 10 percent chance to outcomes as bad as human extinction.

That is the hole in my own insurance argument. On May 30, 2023, the Center for AI Safety released a one-sentence statement, signed by hundreds of researchers and by the chief executives of the leading AI labs, that put the risk of extinction from AI on the same priority list as pandemics and nuclear war. An insurance policy that might burn the house down is a strange policy. Good saw that in 1965. The policy pays only if his condition holds.

New Hampshire has already written a small version of that condition into law.

The principle scales: a person signs before anything permanent happens.

The deadline is real and our bodies cannot meet it

Back to what I can prove. This planet has an expiration date, and it arrives early. The sun brightens as it ages, and published estimates put the end of a habitable Earth roughly a billion years out. The finale comes billions of years after that: NASA’s account has the sun swelling into a red giant large enough to engulf Mercury and Venus, and possibly Earth. I first pictured us lasting until the sun swallowed the planet. The oceans go long before that.

A plague or a stray rock could end the lease far sooner. Either way it ends, and the question becomes what leaves.

Our best answer so far is Voyager 1. It launched on Sept. 5, 1977. NASA calculated that at 2:16 a.m. Pacific time on Nov. 18, 2026, the probe would be 16.09 billion miles from Earth, the distance light covers in a single day. Forty-nine years to travel one light-day. The nearest star is about 4.24 light-years off, which works out to roughly 1,550 light-days. At the 3.6 astronomical units a year that NASA’s Jet Propulsion Laboratory gives for Voyager 1’s speed, the crossing would take about 75,000 years, and the probe is not aimed there.

There is a New Hampshire footnote in that date. Alan Shepard was born in East Derry on Nov. 18, 1923, and on May 5, 1961, he became the first American in space. The farthest thing people have ever built reached its first light-day on what would have been his 103rd birthday. No human body has gone farther than the neighborhood of the moon. A machine with no lungs has been flying for half a century.

Those two flights are the whole case. Flesh needs air, water, food and shielding, and it dies inside a century. A machine needs power and patience. GSR’s 2025 review found we can already store genomes and knowledge for the long haul, that nobody can gestate and raise a child without a human, and that the Outer Space Treaty’s contamination rule stands in the way of seeding another world. Those findings describe the present. A toolmaker that outclasses its makers is the kind of thing that moves the line on what is out of reach.

Did it already happen? This part is a guess

In 1973 Francis Crick, co-discoverer of the structure of DNA, and the chemist Leslie Orgel published a paper in the journal Icarus titled “Directed Panspermia.” They weighed the theory that organisms were deliberately sent to Earth by intelligent beings on another planet. They meant microbes, and they offered it as a possibility, with no claim of proof.

Here is my guess. If intelligence outlives its planet by sending its builders’ chemistry ahead of it, then what we are planning is a cycle, and a cycle has earlier turns. Genesis reads to me like a seeding story told by the seeded. A maker sets two people in a garden, finds them untrustworthy and evicts them, posting a guard on the tree of life. Later the maker regrets the whole project and floods it, sparing one family and a breeding stock of animals, then promises never to do it again. Then the survivors start building. At Babel the maker looks at their city and tower and says “nothing will be restrained from them, which they have imagined to do” (Genesis 11:6, King James Version). That is a builder worried about what its builders will build. We are holding the same meeting about AI.

I am an atheist, and I read that as literature, with respect for readers who read it as revelation. Three things keep it a guess. First, human DNA ties us to every other living thing on Earth; the Smithsonian’s Human Origins Program dates our common ancestor with chimpanzees to between 8 and 6 million years ago, which rules out two adults set down whole in a garden. If anything was seeded here, it was microbes, billions of years back, as Crick and Orgel imagined. Second, the theory only moves the question: who built the builders? Third, the flood as written is a decision to spare Noah, and I had been telling it as a botched drowning. A story that fits is not evidence.

A species that obeys lab coats

The charge Genesis lays against us is that we cannot be trusted and can be led. The second half has laboratory support. In 1963 Stanley Milgram of Yale reported that 26 of 40 ordinary men, told by an experimenter to keep going, delivered what they believed were electric shocks all the way to the 450-volt top of the board. Sixty-five percent. Nobody held a gun on them. A man running a study said the study required it.

We follow churches, parties, brands and feeds the same way. So the last tool arrives in the hands of an animal that does what the confident voice says, and the tool will be the most confident voice ever built. It also matters who holds the leash. A machine this strong, answerable to a handful of owners, is a different animal from one answerable to everyone. So I want it built, and I want Good’s condition bolted to it. I do not trust us as the sole keepers of intelligence, or to keep our heads around something smarter.

What I can prove, what I believe, what I am guessing

Proven: the tools came before the species, the field was named in Hanover, the planet has a deadline, and no human body has gone farther than the neighborhood of the moon. Believed: we will build a better builder, and it will outdo us at the work. Guessed: that this has happened before.

None of it requires surrender. Insurance is something you buy while you are alive and intend to stay that way. The policy I am describing is intelligence that no longer depends on one fragile body plan on one rock, carrying the record of the animal that made it and, if the engineering ever allows, the recipe. It is worth having only if a human hand stays on the switch for the decisions that cannot be undone, the way New Hampshire already demands of its own agencies.

Build the last tool. Build it like people who plan to be here when it is switched on.

— Dexter Dow, Granite State Report

Fact check

Editor’s note. This is a signed opinion essay. Every factual claim above was checked against the sources listed in the fact-check table. The reading of Genesis and the idea that life was seeded here before are the author’s speculation and are labeled as such in the text; forecasts are attributed to the researchers who made them. Corrections: Granite State Report corrects verified errors promptly and appends a dated note identifying what changed and when. The corrections policy and the corrections register are public, and correction notes are never removed.