Elon Musk Tweets 2015-2017

Source and format: Musk’s own tweets across 2015, 2016 and 2017, the three years that follow Elon Musk Tweets 2010-2014, taken from the GitHub cmj/elontweets archive. The three years sit together as one era because they form a single continuous movement. The AI-risk worry he had spent 2012-2014 reading about turns, in this window, into an organized public campaign: open letters, OpenAI, Neuralink, the regulation argument. The older Mars, sustainability and first-principles threads keep hardening right beside it. Only Musk’s own composed tweets are used; retweets and embedded third-party lines are left out.
The arc: if 2010-2014 is where Musk finds his themes, 2015-2017 is where he acts on the biggest one. AI risk stops being a reading list and becomes infrastructure. He signs and amplifies the FLI safety letter (Jan 2015) and the autonomous-weapons letter (Jul 2015), says he is funding safety research, and co-founds OpenAI (Dec 2015) as his answer to “control … by a small number of humans” (Jun 2016). He seeds Neuralink as the brain-machine “symbiosis” hedge (Jun 2016 to Apr 2017). By Aug-Nov 2017 he has settled the two lines he will repeat for years: “vastly more risk than North Korea”, and that “AI should be” regulated “like we do food, drugs, aircraft & cars.” Alongside all this, the simulation hypothesis he is famous for shows up in tweet form (“It is simulations all the way down,” Jun 2016), and the Mars survival imperative sharpens into a Fermi-paradox doctrine (“extinct single planet civilizations probably exceeds live civs by … a thousand to one,” Mar 2017). The one stance that visibly moves inside the window is politics. His Feb-2017 defense of advising Trump (“stay engaged, more moderates not fewer”) gives way, four months later, to his resignation over Paris (Jun 2017), the era’s defining act of conviction.
Trust tier — verified. A verified archive of his tweets; a tweet is its own quote.
Quotes: each tweet links to its post on X. Only Musk’s own original posts are quoted: no retweets, and no text of other people’s posts that he shared.
Attribution. Every quote is an original Musk tweet, matched to the right date and post. None is a retweet, and none is someone else’s tweet pinned on Musk. The one tweet that carries another voice, the Atatürk maxim (“If one day, my words are against science, choose science”), is Musk’s own tweet endorsing it, quoted whole with the attribution exactly as he posted it, not a stranger’s line passed off as his. Several quotes are replies: the Zuckerberg “deep tech understanding,” the Sean-Carroll exchanges on life stability and Europa, the @lessteza “control … by a small number of humans,” the “It is simulations all the way down” reply. Each is Musk’s own reply text, quoted from after the @-mention.

Summary

This is where the early tweet voice of Elon Musk Tweets 2010-2014 turns into a program. The 2010-2014 years caught beliefs forming; these three catch the most important of them being built into institutions. Almost everything across the three years is left out: launch updates, product banter, news links and logistics carry no mind-material and aren’t quoted. What survives is the dated record of Musk turning his AI fear into organizations, and of the metaphysical and political worldviews he is still most quoted for finding their public shape.

The dominant story is AI risk going operational. In 2012-2014 it was a reading note hardening into “more dangerous than nukes.” Here it grows machinery. January 2015 opens with him amplifying the Future of Life Institute open letter, and days later saying he is “Funding research on artificial intelligence safety.” By July he is pushing the autonomous-weapons letter and stating the precautionary core: “Even if inevitable, we should at least attempt to postpone the advent of AI weaponry. Sooner isn’t better.” December 2015 is the landmark, the founding of OpenAI. By mid-2016 he has named the specific fear behind it: not just rogue AI but “control of super powerful AI by a small number of humans”. The same year he proposes the human-side hedge that becomes Neuralink: “Creating a neural lace is the thing that really matters for humanity to achieve symbiosis with machines.” 2017 closes the arc with his two most-cited regulation lines, given within months of each other: “Vastly more risk than North Korea” and the FAA/FDA analogy. Alongside them come the WW3 forecast and the “double exponential” model for why he thinks intuition underrates how fast all of it moves.

The metaphysical Musk takes tweet form here too. The simulation hypothesis he is most associated with is stated flatly (“It is simulations all the way down,” June 2016) and then argued: he offers the speed of light as evidence and suggests nested levels of reality may get “less interesting” the higher you go. The Fermi-paradox and Mars reasoning sharpens from a 2015 “where are the aliens?” musing into the 2017 survival doctrine that single-planet civilizations almost always go extinct, his intellectual basis for the multiplanetary imperative. His population-collapse worry also shows up for the first time here (“Consequences of population implosion greatly underestimated,” 2016).

The clearest shift inside the window is political. In February 2017 he argues for engaging from within: he stays on Trump’s advisory councils because “Activists should be pushing for more moderates to advise President, not fewer,” and “I’d rather do what I believe is right, than do what appears right simply to avoid criticism.” Four months later the same reasoning produces the opposite act. On June 1, 2017 he resigns over the Paris withdrawal (“Climate change is real. Leaving Paris is not good for America or the world”). In this era the principle that keeps him at the table is the very one that finally makes him leave it.

Several of these tweets are dated, falsifiable predictions: “Ramping up the Autopilot software team at Tesla to achieve generalized full autonomy” (2015), “Tickets to orbital hotels, the moon and Mars will be a lot less than people think” (2016), national AI competition as the “most likely cause of WW3” (2017). They are time-stamped readings of his optimism and his alarm.

Key tweets

AI risk: from open letters (2015) to OpenAI (2015) to the regulation argument (2017)

The central thread of these years. 2015 opens with the topic becoming organized advocacy. He amplifies the Future of Life Institute safety letter, then says, days later, that he is personally funding the work, with the dark humor he tends to wrap the fear in:

“World’s top artificial intelligence developers sign open letter calling for AI safety research:” ↗

“Funding research on artificial intelligence safety. It’s all fun & games until someone loses an I” ↗

His model of the real threat as networked and disembodied rather than a humanoid robot, said about Ex Machina:

“Worth watching @ExMachinaMovie. The AI would be in the network, not the robot, but otherwise good.” ↗

By mid-2015 the advocacy reaches autonomous weapons, and he states his precautionary stance at its plainest: postpone it even if you believe it inevitable.

“If you’re against a military AI arms race, please sign this open letter:” ↗

“Even if inevitable, we should at least attempt to postpone the advent of AI weaponry. Sooner isn’t better.” ↗

Then the landmark, December 11, 2015: the founding of OpenAI as his institutional answer to the risk.

“Announcing formation of @open_ai …” ↗

And the early democratization case he attaches to it: the open, widely shared AI that drives OpenAI in this period, a position he later moves away from.

“Would like to thank @nvidia and Jensen for donating the first DGX-1 AI supercomputer to @OpenAI in support of democratizing AI technology” ↗

Through 2016 he keeps feeding the exponential-takeoff model. He reposts the Wait But Why essay and reads the AlphaGo result as AI arriving faster than the experts predicted.

“Worth reposting the Wait But Why piece on AI. We are at the beginning of exponential growth in digital intelligence.” ↗

“Congrats to DeepMind! Many experts in the field thought AI was 10 years away from achieving this.” ↗

A reply shows how seriously he takes it. He gets tutored weekly by top researchers, and he says so while disagreeing in public with the AI optimists:

“Zuck doesn’t (yet) have a deep tech understanding of AI. I spend hours every week being educated by world’s best researchers.” ↗

His sharpest statement of which AI fear drives OpenAI, not rogue AI alone but concentration of AI power, comes in a June 2016 reply:

“control of super powerful AI by a small number of humans is the most proximate concern” ↗

An early warning about AI-driven manipulation of the information ecosystem:

“Only a matter of time before advanced AI is used to do this. Internet is particularly susceptible to a gradient descent algo.” ↗

In 2017 the campaign hardens into the two arguments he repeats for years. He endorses the Asilomar principles, then calls safe AI existentially critical:

“Top AI researchers agree on principles for developing benefical AI” ↗

“Yes, I believe that is critical to ensure a good future for humanity” ↗

He separates the deep AI risk from mere job automation, pinning down exactly what he is worried about, and offers the “double exponential” model for why intuition underrates it:

“Deep AI is the real risk, though, not automation.” ↗

“we actually face a double exponential rate of improvement. AI hardware & software are both exponential.” ↗

Then, in August 2017, come his most-cited warnings. He ranks AI above geopolitical nuclear threats, makes the regulation argument, and offers a theory of why experts dismiss the danger:

“If you’re not concerned about AI safety, you should be. Vastly more risk than North Korea.” ↗

“Nobody likes being regulated, but everything (cars, planes, food, drugs, etc) that’s a danger to the public is regulated. AI should be too.” ↗

“Biggest impediment to recognizing AI danger are those so convinced of their own intelligence they can’t imagine anyone doing what they can’t” ↗

The best-or-worst framing (citing Tegmark’s Life 3.0), and the WW3 forecast with its twist that an AI, not a country, might start it:

“AI will be the best or worst thing ever for humanity, so let’s get it right.” ↗

“Competition for AI superiority at national level most likely cause of WW3 imo.” ↗

“May be initiated not by the country leaders, but one of the AI’s, if it decides that a prepemptive strike is most probable path to victory” ↗

And the November 2017 restatement of the regulation argument with the FAA analogy, the stance visibly hardening:

“Got to regulate AI/robotics like we do food, drugs, aircraft & cars. Public risks require public oversight.” ↗

Human–AI symbiosis: the genesis of Neuralink (2016-2017)

Running underneath the alarm is the hedge. In June 2016 the brain-machine-interface idea that becomes Neuralink shows up in its earliest tweet form:

“Creating a neural lace is the thing that really matters for humanity to achieve symbiosis with machines” ↗

By March 2017 he ties it directly to existential risk, and in April he boils the merge idea down to a single line:

“Long Neuralink piece coming out on @waitbutwhy in about a week. Difficult to dedicate the time, but existential risk is too high not to.” ↗

“That is the aspiration: to avoid AI becoming other.” ↗

Simulation hypothesis: stated, then argued (2016-2017)

The metaphysical belief he is most associated with takes tweet form here. First he states it flatly, then he argues it with physics, then he develops it into the nested-levels picture:

“It is simulations all the way down” ↗

“Will try out @NoMansSky this weekend. Maybe reality is just a series of nested simulations all the way down …” ↗

“I agree. Speed of light is a point in favor of the sim.” ↗

“So far, even our primitive sims are often more entertaining than reality itself.” ↗

It also turns up as one item on his menu of Fermi-paradox answers, and as the half-joking “most entertaining outcome” rule of thumb he ties to it:

“Either that, simulation, great filter or they are very, very subtle.” ↗

“It does often seem that the most entertaining belief or outcome is the most likely to prevail” ↗

Mars and the Fermi paradox: the survival imperative sharpens (2015-2017)

The Mars case here is increasingly argued from the Fermi paradox. It opens in 2015 with the bare “where are the aliens?” puzzle, and with his contrarian view that primitive life is common while intelligence is rare:

“Seems like an opportune moment to bring up the Fermi Paradox, aka “where are the aliens?” Really odd that we see no sign of them.” ↗

“No way. Look at chemotrophs and extremophiles. Primitive life is bloody everywhere. Highly under the ice on Europa imo.” ↗

Underneath sits a worry about existential fragility, that advanced life tends to destroy itself, which feeds straight into the urgency of becoming multiplanetary:

“Am concerned that advanced life may be inherently unstable. Arguably, any civ capable of extinguishing itself eventually will.” ↗

“Hence the urgency of making life as we know it multiplanetary as soon as possible” ↗

His contrarian prediction on how cheap space access will get:

“Tickets to orbital hotels, the moon and Mars will be a lot less than people think.” ↗

By 2017 it has hardened into the canonical survival statement, the Fermi reasoning turned into doctrine, and the inspiration motive. He also does something characteristic: refuses to overclaim for his own cause, knocking down the hype around a NASA bill.

“Mars is critical for the long-term survival of humanity and life on Earth as we know it” ↗

“The number of extinct single planet civilizations probably exceeds live civs by at least a thousand to one. I’m being optimistic.” ↗

“It is high time that humanity went beyond Earth. Should have a moon base by now and sent astronauts to Mars. The future needs to inspire.” ↗

“Perhaps there will be some future bill that makes a difference for Mars, but this is not it.” ↗

Sustainability: subsidies, climate metrics, and the Paris resignation (2015-2017)

His climate argument keeps reasoning from first principles. He returns again and again to the fossil-fuel subsidy as the core market distortion, and works from the carbon cycle to say what actually matters:

“Fossil fuels subsidised by $10m a minute, says IMF” ↗

“Global economy of $100T is overwhelmingly fossil fuel based, so IMF estimate of $5T or 5% for carbon subsidy is def right order of magnitude” ↗

“CH4 rapidly decays back to CO2 & is absorbed by plants. What matters is adding new carbon to surface cycle from underground oil, gas & coal.” ↗

He rethinks the climate metric from scratch (peak, not average, temperature), makes the libertarian case for a carbon tax, and turns the subsidy fight into a question of fairness:

“Great article by @BadAstronomer. Important to note max temp record. Too much attention on avg temp. Max matters most” ↗

“The libertarian argument for a carbon tax” ↗

“Real fraud going on is denial of climate science. As for “subsidies”, Tesla gets pennies on dollar vs coal. How about we both go to zero?” ↗

The era’s defining act of climate conviction: resigning from the presidential councils over the Paris withdrawal.

“Am departing presidential councils. Climate change is real. Leaving Paris is not good for America or the world.” ↗

Self-driving and tunnels: predictions and the genesis of Boring (2015-2016)

The autonomy thread gets three dated 2015 markers. He puts a value on the human freedom to drive, predicts that autonomy may one day outlaw it, and declares the goal of full autonomy outright:

“To be clear, Tesla is strongly in favor of people being allowed to drive their cars and always will be. Hopefully, that is obvious.” ↗

“However, when self-driving cars become safer than human-driven cars, the public may outlaw the latter. Hopefully not.” ↗

“Ramping up the Autopilot software team at Tesla to achieve generalized full autonomy.” ↗

The famous AI worrier catching the tension in his own views when Autopilot comes under attack:

“Sure feels weird to find myself defending the robots” ↗

And the tunnels-versus-flying-cars reasoning that seeds The Boring Company. He thinks about city traffic in terms of dimensions: 2D streets under 3D buildings. Then comes the impulsive 2016 tweet that starts the company:

“Reasons I like tunnels: you still travel in 3D fast, but immune to weather, quiet and no risk cars fall on your head” ↗

“Right move is try both tunnels and flying cars. Otherwise, having 2D streets and 3D buildings means bad traffic forever.” ↗

“Traffic is driving me nuts. Am going to build a tunnel boring machine and just start digging…” ↗

First principles applied to physics, valuation, evidence (2015-2017)

The explaining, evidence-demanding voice runs all through these years. He reasons from Newton’s third law to kill a misconception, rejects pseudo-archaeology with a sharp standard for what counts as evidence, and views physics in terms of information:

“Reason is Newton’s Third Law. In vacuum, there is nothing to “push” against. You must react against ejected mass.” ↗

“Btw, please don’t mention the pyramids. Stacking stone blocks is not evidence of an advanced civilization.” ↗

“Interesting to think of physics as a set of compression algorithms for the universe. That’s basically what formulas are.” ↗

His rules for knowing things, boiled down: go to the primary source, put science above any authority including himself, and match how much you believe to how much evidence there is.

“Reading the source material is better than reading other people’s opinions about the source material” ↗

““If one day, my words are against science, choose science.” Mustafa Kemal Atatürk” ↗

“So strange that people often believe things inversely proportionate to the evidence.” ↗

His first-principles argument about valuation (a defense of Tesla’s valuation logic he makes more than once), and reusability treated as the obviously correct norm:

“A stock price represents risk-adjusted future cash flows.” ↗

“It’s starting to feel kinda normal to reuse rockets. Good. That’s how it is for cars & airplanes and how it should be for rockets.” ↗

Politics and free speech: engage-from-within — then resign (2017)

The era’s clearest political shift. In February 2017 he defends staying on Trump’s advisory councils, an engage-from-within case: decide by principle, not by how it looks, keep more moderates in the room rather than fewer, and hold climate as the moral anchor.

“I’d rather do what I believe is right, than do what appears right simply to avoid criticism.” ↗

“In addition, I again raised climate. I believe this is doing good, so will remain on council & keep at it. Doing otherwise would be wrong.” ↗

“Activists should be pushing for more moderates to advise President, not fewer. How could having only extremists advise him possibly be good?” ↗

Four months later the same reasoning produces the opposite act, the Paris resignation quoted above. And in November 2017 comes an early stand against ruining someone over a single comment, a first sign of the free-speech crusade to come:

“Do you really believe that someone’s life should be ruined if they make a single offensive comment for which they then sincerely apologize?” ↗

Self-revelation: population collapse, existential dread, and the cost of the path (2016-2017)

The candid admissions of these years. His population-collapse worry, which he returns to for years afterward, first appears here, twice:

“Consequences of population implosion greatly underestimated. Upside down demographic pyramid can’t support social services.” ↗

“The world’s population is accelerating towards collapse, but few seem to notice or care” ↗

The philosophical, existential side of him: literature as a mirror for the human condition, the mental cost of thinking about the future, and a humility about how small we are on a cosmic timescale.

“Have recently come to appreciate the awesome, absurdist humor of Waiting for Godot. We so often wait, without knowing why, when or where.” ↗

“Occasional existential dread is inevitable if you think at all about the future” ↗

“A reminder of the youth of our 10,000 year (to be generous) civilization. We are not even a flash in the pan compared to the 4,500,000,000 year age of Earth.” ↗

His reflection on cosmic finality and meaning (prompted by Asimov’s The Last Question), and his view of the self as an evolving pattern, which bears on his Neuralink and consciousness thinking:

“If heat death of the universe is the destination, it really is all about the journey” ↗

“What “you” are is an evolving, mind-body pattern, but there’s still a core youness that holds true over time.” ↗

The rawest admissions of these years. He names the personal cost of his work, owns the suffering as self-imposed, says he acted for the mission rather than the money even at a roughly 90% expected chance of failure, and lays out a credo on persistence and friendship:

“The reality is great highs, terrible lows and unrelenting stress. Don’t think people want to hear about the last two.” ↗

“If you buy a ticket to hell, it isn’t fair to blame hell …” ↗

“Thought 90% prob of losing it all (almost did many times), but it was the only chance.” ↗

“Don’t give up if the cause is important enough, even if you believe you are walking into doom. Good friends really matter.” ↗

And the deepest reason he gives for the ambitious projects: that life needs inspiring reasons, not just problems to solve.

“Our existence cannot just be about solving one miserable problem after another. There need to be reasons to live.” ↗

Connections

AI existential risk — the period the worry becomes a program: the 2015 safety/autonomous-weapons letters and “Funding research on artificial intelligence safety,” the OpenAI founding, the “control … by a small number of humans” motive, and the 2017 “Vastly more risk than North Korea” / “regulate AI/robotics like we do food, drugs, aircraft & cars” / WW3 lines.
Human–AI symbiosis / Merging with AI — the genesis of the Neuralink hedge: “Creating a neural lace … to achieve symbiosis with machines” (2016) and “to avoid AI becoming other” (2017).
Simulation hypothesis — the tweet-form origin of the belief he is most quoted for: “It is simulations all the way down,” the speed-of-light argument, and the nested/“more entertaining” reasoning.
Mars colonization — the Fermi-paradox survival doctrine (“extinct single planet civilizations … a thousand to one”), “Mars is critical for the long-term survival of humanity,” and the affordability and inspiration framings.
Sustainable-energy mission — the fossil-fuel-subsidy critique, the carbon-cycle and peak-temperature reasoning, the libertarian carbon-tax framing, and the Paris-withdrawal resignation.
Autonomous driving — the 2015 “generalized full autonomy” goal and the prediction that safe autonomy may outlaw human driving.
First principles — Newton’s-third-law and carbon-cycle reasoning, “physics as … compression algorithms,” “A stock price represents risk-adjusted future cash flows,” reusability-as-norm, and the source-material / proportion-belief-to-evidence epistemology.
Free-speech absolutism — the November-2017 anti-cancel-culture proportionality stance that predates the Twitter-era crusade.
Curiosity and truth-seeking — the evidence-demanding register (pyramids, “choose science,” “inversely proportionate to the evidence”) and the read-the-source-material maxim.
The Boring Company — the tunnels-vs-flying-cars dimensional reasoning and the impulsive “just start digging” genesis tweet.
Elon Musk — the entity’s 2015-2017 first-person record: the engage-then-resign political movement, the population-collapse worry, and the rawest self-disclosures on the cost of the path.