The 25 Science-Fiction Books You Must Read, Ranked

And six short story collections

Some ‘best of’ lists seek to achieve a spurious objectivity through means such as reader polls, complex ranking systems or asking distinguished experts for their views.

This is not one of those lists. This list has instead been compiled by the greatly superior method1 of me thinking really hard about it and assuming that my opinion mattered more than anyone else’s.

In terms of criteria, firstly, every book on the list is an excellent read. Some of the books on here may be worthy, but there are no books on here purely for their worthiness. Secondly, every book is one that will make you think, or open your eyes to a new perspective - for that is a core function of science fiction, to explore new possibilities. And thirdly, to whittle the list down to 25, I have taken into account relevance and impact - either to the world of today, or to the genre more broadly.

While the borders of every genre are porous, every book on here has a clear claim to being science fiction. There is no pure fantasy, or alternate history - much though I enjoy those genres, this is not their list.

So, without further ado, let’s get cracking. As is traditional, we’ll do this in reverse order, so you’re kept guessing to the very end as to whether I’m overlooking your own favourites, or saying they’re amongst the best.

Last and First Men - Olaf Stapledon (1930)

A future history, not a novel, Stapledon takes us forward across billions of years as new races of humanity rise and fall - on Earth, on Venus and ultimately on Neptune. Some winged, some more intelligent and some less intelligent, some deliberately designed and some evolved, each have their turn on the stage. [...]

The Mote in God’s Eye - Larry Niven and Jerry Pournelle (1974)

What happens when you combine Niven’s extraordinary imagination with Pournelle’s ability to actually tell a story that has a plot? You get The Mote in God’s Eye, perhaps the best first-contact encounter with a truly alien extra-terrestrial species ever written, in a richly rendered future setting. A rare occasion of a joint-authored novel being better than any that either has written alone. [...]

The Day of the Triffids - John Wyndham (1951)

What if almost everyone went blind, and sentient, walking, carnivorous plants inherited the earth? This unlikely premise forms the basis for one of the most civilly bleak post-apocalyptic novels, as the protagonist and his companion wend their way across the devastated Home Counties, seeking a safe haven where the remnants of humanity can survive. [...]

The Diamond Age - Neal Stephenson (1995)

A world of nanotechnology and AI, in which nation states have been supplanted by ‘phyles’ - groupings, sometimes ethnic, sometimes regional, sometimes political, of freely associating individuals - the Diamond Age tells the story of Nell, a young girl growing up in poverty who receives an advanced AI ‘primer’, designed to educate and guide its owner into a ‘more interesting life’. A wildly imaginative roller-coaster ride through a provocative future. [...]

The Warrior’s Apprentice - Lois McMaster Bujold (1986)

Arguably the best science-fiction author currently writing, The Warrior’s Apprentice introduces the brilliant, driven, physically stunted Miles Vorkosigan, primary protagonist of her Vorkosigan Saga, and follows him as he ‘accidentally’ builds himself a mercenary fleet. Modern space opera at its finest. [...]

Foundation - Isaac Asimov (1951)

A novel that shaped the genre, Foundation tells the story of a Galactic Empire in decline, the brilliant psychohistorian who sees a way to minimise the harm - and the lives of those who followed him, living his plan, guiding the new Foundation in its ascent. [...]

The Three-Body Problem - Liu Cixin (2008)

by Edrith, World of Edrith |  Read more:

Images: uncredited

[ed. Not a bad list. Could easily be doubled.]

China and the Future of Science

The Chinese socio-political system differs from our own. From the perspective of the topic of this conference, here is the most salient distinction: the Chinese system has a telos. The Chinese party-state is fundamentally a set of goal-oriented institutions. This is not unique to China—it is in fact a distinguishing feature of all Leninist systems. I sometimes think of Leninist systems as a little bit like that bus in the movie Speed. Who here has seen it? For those who haven’t, here is basic gist of that film: an extortionist attaches a bomb to the speedometer of a bus. If the bus ever slows below 50 miles per hour, everyone blows up. So it is with your average communist system. Either it hurtles towards some clearly defined goal or things start to fall apart.

In the early days of Mao, the overarching aim of the communist system was to seize state power, first through subversion and insurgency, then through more regular combined arms warfare. In the later days of Mao the newly established Chinese state and the society it intertwined were oriented around class struggle, both at home and abroad. From the 1980s through the 2010s the Chinese system was orbited a different yet still very explicitly stated goal: getting rich. In theory, if not always in practice, every action taken by every cadre, every soldier, and every state employee was subordinate to this larger, unifying aim. We must make China rich.

That is no longer the animating telos of the Chinese system. There is a new goal, one that has been articulated with great clarity by Chairman Xi and the Chinese central committee: In 2026, the aim of China’s communist enterprise is to lead humanity through what they call “the next round of techno scientific revolution and industrial transformation.”2 The Chinese leadership believes humanity stands on the cusp of the next industrial revolution. China can only be restored to its ancestral greatness if it is the pioneer of this revolution. All machinery of party and state must bend towards this end. All 100 million members of the Communist Party of China, all 50 million government employees of the PRC, all two million soldiers of the People’s Liberation Army, and ultimately all of the 1.4 billion people that call China home must be mobilized to accomplish this aim. That is the ambition. China will be the greatest scientific power the world has ever seen—or bust.

The communists are deadly serious about their pursuit of this aim. Statistics provide one window into the seriousness of their intent. Now I don’t intend for the remainder of this speech to be a laundry list of numbers, but I think the numbers are useful for helping us see the scale of what China has already accomplished and the speed with which they have accomplished it. They are also strong signal of future intent—it is difficult to survey the numbers and not appreciate just how ironclad China’s commitment to scientific achievement really is.

Now scientific achievement is difficult to measure. One common metric is to count the so-called “high impact papers” – journal articles highly cited by other leading lights in a given scientific field. Count up these papers over the course of a year, see who wrote them, see where those authors work, and—voila!—you have a ranked list of which institutions are putting out the most high-impact science in a given year. Had you done this counting exercise in the year 2005, you would have discovered that six of the world’s ten most productive universities were in the United States. Today only one of those universities is in the United States. That university is Harvard, coming in at spot number three on the list. At spot number one? Zhejiang University.

How many of you have heard of Zhejiang University? Can I get a show of hands?

And of course, Zhejiang University is just one of the Chinese institutions on this top ten list. China claims not just the number-one spot, but also the number-two spot. And not just the number-one and number-two spots, but also the fourth, fifth, sixth, seventh, eight, ninth spots go to the Chinese.

The scientific publisher Nature makes a similar catalog on a slightly more granular level, looking at specific fields of science. According to Nature’s most recent rankings, 18 of the top 25 most productive research institutes in the physical sciences, 19 of the top 20 in geosciences, and a full 25 out of 25 in chemistry are Chinese. Only in the biosciences do American scientists still have a lead—but even on that list three of the top ten are Chinese.

The kicker is, none of that was true even just a decade ago.

The most granular analysis of all is published by the Australian Strategic Policy Institute, or ASPI. ASPI publishes a neat research tracker that surveys new publications in 74 distinct high-end technologies. Unlike the statistics I just discussed, their tracker includes research published by scientists working in national laboratories and private institutions as well as those published by academic scientists. For each category they make a list of the ten institutions that are publishing the most high-impact science in that particular topic. What have they found? For 66 of the 74 categories tracked, a majority of the institutions that are now publishing the highest-impact science are Chinese. In many areas of science the dominance is total: For example, ten of then most productive research institutions in the fields of nanoscale material manufacturing, photonic sensors, chemical coating, drone operations, automated swarms, and undersea communications are Chinese. The number is nine out of ten for work on supercapacitors, advanced composite materials, inertial navigation systems, and satellite positioning, eight out of ten in advanced optical communications, advanced radiofrequency communications, and new chemical coatings, and seven out of ten for directed energy technologies, nuclear engineering, and nuclear waste treatment.

The scale of Chinese scientific production is in part a story about people. China graduates five times the number of medical and biomedical students than we do every year, seven times the number of engineers, and two-and-a-half times the number of undergraduates with research experience in artificial intelligence. Last year China graduated almost double the number of STEM PhD students than we did—and that number is actually worse than it sounds because—depending on the exact year you do the counting—between one sixth and one fifth of our STEM graduates are themselves Chinese.

Many of these researchers go back. They go back partially because they are well compensated for doing so. They also go back because of the research opportunities afforded to them. A recent study found that returning Chinese scientists go on to become the lead author on 2.5 times more papers than their colleagues who stay in the United States.9 Many Chinese research labs have 30 or 40 people attached to them—the equivalent to a commercial research lab in the United States. Ask any scientist who has gone to China in the past three years to visit academic colleagues and they will tell you how astounded they are at the quality of the laboratory equipment and machinery that their Chinese colleagues have access to. If in the not-so-distant past Chinese localities competed with each other to lay the most asphalt, now that funding pours into laboratory equipment, scientific instruments, and advanced scientific facilities. Thus China now has the world’s most sensitive ultra-high-energy cosmic-ray detector, the world’s largest and most sensitive radio telescope, the world’s strongest steady-state magnetic field, the world’s fastest quantum computer by computational advantage, and the world’s most sensitive neutrino detector. Just yesterday an attendee at this conference informed me of another I should add to my list: the world’s largest primate medical research center.

Now I can already hear some of your objections. “Tanner, these measures don’t include classified research. They don’t include the proprietary research by private companies—that is the stuff that actually pushes technology forward. American companies are not publishing billion-dollar trade secrets in the latest journals. The Chinese scientists are under insane publish or perish pressures—they are far more likely to lie and cheat. Don’t you know Chinese scientists take part in citation cartels? Haven’t you read those bitter critiques of the new system written by China’s own disgruntled scientists?”

My main response to this: you guys have lost the thread. I am reminded of a similar style of argument we often see in AI development. Every time a new model is released people play around with it for a bit and then start to catalog the flaws of this model. But the real story, the story historians will tell a generation from now, is never about the model of the moment. What matters is movement between those moments. History is made by the trend-line. What capabilities did the models have four years ago? What capabilities do they have now? What might they reasonably be expected to have in a decade hence?

Something similar might be said for science and China.

Moreover, it is not hard so hard to find examples of dominance in the scientific literature leading to dominance in the world of applied technology. More than a decade ago Chinese researchers began to dominate academic work on batteries. Today China produces 80% of the world’s battery cells. Around that same time Chinese researchers took the lead on factory robots and factory automation. Today Chinese firms buy more robots than the rest of the world combined. They have a higher robot density than almost every other country in world, despite being the world’s second largest nation. Another example: China is the only place where the unit cost of building a new nuclear reactor has gone down over the last decade. Do you think this is unrelated to the number of nuclear engineers they train and the amount of basic research they publish on nuclear engineering? Again I beg you to look at the trend-line: In 2026 China has 102 nuclear reactors either in operation or under construction. How many did they have 15 years ago? Less than 30.

Why is China pouring so many resources into scientific and technological advance? Here is a 30-second thumbnail sketch: Xi Jinping and those around him have a specific view of history. Like most Chinese, they remember with fondness past eras when their nation was the centerpiece of human civilization. They mourn China’s fall from those grand heights. They remember with bitterness the “century of humiliation” in which their people were exploited by foreign powers and wasted by vicious warlords. The Communist Party of China was founded to save China from that fate. From its inception it has seen itself as the vehicle for the salvation of the Chinese nation and thence the restoration of this nation to its rightful place at the center of the human story.

But why was China displaced from its spot “at the center of the world stage” in the first place? Chinese intellectuals have been asking this question for more than a century. Here is their most common answer: science and technology. Today Party-affiliated thinkers often describe modern human history as punctuated by discrete revolutionary moments when a new technological regime comes into being. They call these moments “rounds of techno-scientific revolution and industrial transformation.” In such moments the material basis of human civilization changes. Those who master this transition master… well, just about everything.

by Tanner Greer, Scholar's Stage |  Read more:

Image: uncredited

[ed. Go USA. See also: The Cure for Cancer Is Not a Pure Math Benchmark (SS).]

The Incredible Journey of a Migratory Shorebird

On a sunny June morning, in a bog outside Beluga, Alaska, I met a Hudsonian godwit. She had inky eyes and a long, slightly upturned beak that was pinkish-orange at its base and brownish-black at its tip. The feathers on her back were of mottled sepia tones, and those on her belly were rusty brown and painterly white. Her legs reminded me of the slender inserts of spring-loaded ballpoint pens, and her feet were refined pterodactyl claws. Her name (among humans) was A34, and she was about the size of a city pigeon. She was in the hands of Nathan Senner, a conservation biologist, who removed and replaced her geotag; measured her head, beak, wing, and leg; weighed her; and took two tail feathers and a sample of her blood. From that small number of data points, he can deduce where she’s been, how her feathers have grown, what genetic subgroup she belongs to, and her degree of exposure to mercury.

Senner leads a research laboratory at the University of Massachusetts, Amherst, that is largely focussed on how migratory shorebirds, such as godwits, respond to environmental change. He grew up in Anchorage, where school loudspeakers announced the street corners on which moose had been sighted. In high school, he wrote an occasional “Birding with Nathan” column for the Anchorage Daily News. Senner’s father, Stanley, has devoted more than five decades to the conservation of migratory shorebirds. Starting in the nineteen-seventies, Stanley and his colleagues have been sounding the alarm on the vulnerability of shorebirds and working to protect key habitats, such as Alaska’s Copper River Delta. “Back then, no one was even talking about climate change,” he told me recently. Since 1980, roughly half of migratory shorebird species’ populations have declined by more than fifty per cent.

Migratory shorebirds are usually not a person’s first bird crush. They lack the supernatural aura of the resplendent quetzal, and they don’t go in for outlandish mating displays like that of the prairie chicken. As the Princeton University biology professor and shorebird-lover David Wilcove put it, migratory shorebirds can give an impression of being “little brown guys, buzzing around in the mud, not too big and not too small, and, whether one is a stint or a phalarope or a knot or a godwit, they’re just kind of there.” And yet, Wilcove told me, they are a birder’s kind of bird. Once people get to know them, they tend to fall in love. [...]

Migratory birds in captivity develop Zugunruhe, an agitated fitfulness, at certain times of year. Scientists don’t know precisely what provoked A34 to set off northward one spring afternoon, with a flock of her fellows, though, if they did, they might better know how her species will be affected by the shifting seasons of our changing climate. Godwits cannot soar like hawks or glide like albatrosses. They migrate by continuously flapping their wings. They must navigate crosswinds and headwinds, storms and maybe an occasional hurricane, continuing on day after day, night after night. As best as we can tell, they do not, while flying, eat or drink or sleep (at least not with more than one hemisphere of their brain at a time). After a journey of hundreds of hours and thousands of miles, they descend on a snow-covered shore at the Cook Inlet of southern Alaska or another favored breeding ground. They land considerably lighter than when they took off. “Many arriving shorebirds seem to be tired,” the ornithologist and conservationist Joseph Archibald Hagar wrote, in his foundational monograph, “Nesting of the Hudsonian Godwit at Churchill, Manitoba,” from 1966. In a field-diary entry, Hagar describes a group of birds that “within a minute or two dropped into some sunny spot out of the wind, tucked heads into scapulars, and went to sleep, not to move again for as long as we watched.”

In 1976, the young ornithologist Robert Gill took a job with the U.S. Fish and Wildlife Service in Anchorage. He was dispatched to the Alaska Peninsula to do biological inventories on public land that was up for lease to private companies. During those long-lit Alaskan summer days, walking across salt marshes and through the foothills of the mountains, Gill encountered thousands upon thousands of migratory shorebirds. He told me, “In Alaska, where the seasons are so prominent, it’s very dramatic. It feels like all the birds turn up over the course of one weekend in May.” Along with godwits, the peninsula was a summer home to yellowlegs, dowitchers, sandpipers, Arctic terns, and other species. But almost no one was aware that so many birds relied on this land. How the birds managed their journeys, where they might stop along the way, even how many of them there were—no one knew for sure. “In the vast expanses of wetlands and coastal tundra, you could see why,” Gill said.

The extinction of the once common passenger pigeon, famously described by the naturalist John Muir in his memoir (“I have seen flocks streaming south in the fall so large that they were flowing over from horizon to horizon in an almost continuous stream all day long, at the rate of forty or fifty miles an hour, like a mighty river in the sky”), was part of the impetus behind the Migratory Bird Treaty Act of 1918. Godwits are protected under the act, so killing them or holding them captive, even for scientific purposes, is highly regulated. Also, they’re difficult to capture. Gill’s means of learning more about them was limited, he said, to “the kind of observational work done by old-school naturalists.”

In October of 1987, Gill received a call about nine bar-tailed godwits that had crashed into a radar dome near Cold Bay, Alaska, and died. “They were greaseballs,” he recalled. “They made any Christmas goose look lean.” The carcasses were sent to a lab to be analyzed. Fifty-five per cent of their body mass turned out to be fat. Gill suspected that these chubby godwits were preparing to fly non-stop from Alaska to New Zealand, where this species wintered. However, at that time, the longest known non-stop migration flight was barely half that long. “But, if they were going to stop in French Polynesia, why carry all that baggage?” Gill reasoned.

The lab also measured the birds’ internal organs. Gill’s collaborator, Theunis Piersma, noticed something peculiar: the guts, gizzard, liver, and kidneys of these birds were very small—it was like cutting open a lion and encountering digestive organs the size of a house cat’s. Then, in March of 1992, some forty godwit carcasses were seized from a poacher in New Zealand. These were birds in a very different part of their migratory cycle, and their digestive organs, when analyzed, were of more normal proportions. In 1998, Piersma and Gill published their findings in a paper titled “Guts Don’t Fly: Small Digestive Organs in Obese Bar-Tailed Godwits,” concluding that the birds shrank their digestive organs to reduce their weight and their metabolic demands during a non-stop migration. And then, somehow, regrew them. [...]

A female godwit, E7, took off on March 17, 2007, from near the Piako River, on New Zealand’s North Island. She flew more than six thousand miles non-stop to a nature preserve on the Yalu River, near the China-North Korea border; she stayed there for a little more than a month; she arrived at her nesting area in Alaska on May 15th and spent the summer there. The tracker was still working on August 29th, when E7 started flying southeast, over the Pacific. Gill’s team was doing field work in western Alaska and had only spotty internet service; the team would gather around a laptop, waiting for the updates on E7’s location to come through. North of Kauai, she took a slight right turn and continued flying. On September 7th, after more than two hundred hours aloft, she landed at the mouth of the Piako River, back on the North Island of New Zealand, where Gill had first met her.

Before E7, the longest documented non-stop migratory-bird flight was that of a Far Eastern curlew who had flown about four thousand miles. E7 flew some seven thousand without a break. The flight defied reason. “I had an engineering professor at M.I.T. call me up and say he gave his students an assignment to compare the size and flight range of E7 to a 747 jet,” Gill said. 

by Rivka Galchen, New Yorker | Read more:
Image: Ash Adams
[ed. Stan Senner was a friend and collegue. Didn't know Bob Gill personally but used his work frequently in my habitat protection efforts.]

Some Ways AI Could Kill Us All

[ed. Nearly everyone agrees by now that AI is an existential threat, but the hows and whys are often vague or imcomplete.] 

I don't think this is how it will actually play out. If you play a chess grandmaster, you can predict that they will beat you even if you can't predict how. I chose these examples because I don't think they require much imagination or accepting exotic assumptions.

It is important to note that if chimpanzees were to guess how humans would decimate them, they would get it wrong. Chimpanzees would not imagine guns. They would not foresee poison gas. They would not conceive of chemical castration. They would not imagine humans going around and intentionally infecting them with AIDS. They have no concept of these things; they would not see it coming.

Perhaps they might guess we'd be really good at throwing rocks. Amazingly good. Well, technically, that's what guns do: throw "rocks" really really well.

So how will superintelligent AI actually wipe us all out? Probably in a way I couldn't conceive of. Nonetheless, it's not hard to see how deadly they could be with what we already know about.

by Ruby, Less Wrong |  Read more:

Image: DALL-E via; and here. 

[ed. I wouldn't necessarily conflate super intelligence with ill-will. Most likely AIs will just be pursuing some unrelated goal and exploring every possible means by which to achieve it (including removing human barriers). That is, unless bad (human) actors are involved, which should not be discounted. See also: How My Students Think About AI (LW).]

Cloudy With a Chance of Controversy

Somebody messed with Alaska’s weather last month. Before anybody checks the skies for black helicopters, relax.

On Aug. 23, a California company called Rainmaker Technology Corporation launched a drone near the head of Kachemak Bay and flew it into clouds cold enough to contain supercooled water. The drone released less than 1 pound of silver iodide. Rainmaker says the experiment coaxed about 19 million gallons of additional precipitation from those clouds over three hours.

That sounds dramatic. It wasn’t. Spread over roughly 100 square miles, 19 million gallons amounts to about one-hundredth of an inch of rain. If you had been standing underneath it, you might have noticed a drizzle. You also might have continued mowing the lawn.

Given the collective social media freakout, you’d be forgiven for thinking Rainmaker had pointed a weather-controlling laser beam right at Homer and hit “deploy” from a top-secret mission control bunker.

The experiment produced an impressive online meltdown after residents and local elected officials discovered it had happened without much advance notice. State Rep. Sarah Vance called on Rainmaker to cease operations in her district. Kenai Peninsula Borough Mayor Peter Micciche raised questions about why local residents weren’t informed. Social media did what social media does whenever the words “weather modification” appear together. Cue the chemtrail conspiracy theorists.

Cloud seeding isn’t science fiction, and it certainly isn’t a secret government plot. Scientists have experimented with it since the 1940s. The basic idea isn’t especially difficult to understand.

Some cold clouds contain tiny droplets of water that remain liquid even below freezing. Introduce particles such as silver iodide, whose structure helps ice crystals form, and some of those droplets freeze, grow and become heavy enough to fall as snow or rain. Some. That is the key word there.

Cloud seeding can’t manufacture a thunderstorm over a cloudless desert; it can’t steer hurricanes; and it can’t turn a passing cloud into Noah’s flood. It can, however, give the right kind of existing cloud a little nudge of inclement encouragement.

Research has become considerably more sophisticated, too. The 2017 SNOWIE experiment in Idaho demonstrated that scientists could detect and measure precipitation produced by silver iodide seeding under the right conditions.

Now, that doesn’t mean every question has been answered. A 2024 Government Accountability Office review found that studies estimated cloud seeding can increase precipitation anywhere from zero to 20%, depending on conditions, and said more research is needed to understand when it works best. The GAO also found that existing research suggests the amounts of silver iodide currently used do not pose an environmental or human health concern, while noting that the effects of much more widespread use aren’t as well understood.

In other words: We need to do more science, and Alaska is a pretty good place to do it.

Rainmaker came here because cloud seeding requires particular atmospheric conditions, including clouds cold enough for the process to work, which Alaska has. The potential applications extend far beyond producing an August drizzle near Homer.

Western states are staring at increasingly serious water problems. Reservoirs and rivers supplying farms, cities and hydroelectric projects are under pressure. Snowpack is, in effect, nature’s water-storage system. Even squeezing a few additional percentage points of precipitation from suitable clouds could matter enormously when repeated across the right watersheds.

Cloud seeding won’t solve climate change or replace conservation efforts, but if better engineering can help put more snow on mountains and more water into reservoirs, we should be rooting for the scientists trying to figure out how.

That brings us to the one part that beckons reasonable criticism.

Rainmaker says it notified the Alaska Department of Natural Resources before the test. DNR determined the drone research was a generally allowed use of state land, but the department says it has no jurisdiction over cloud seeding itself. The Department of Environmental Conservation is now reviewing what happened.

That’s a little bit of an awkward gap but despite the naive handwringing by elected officials, we shouldn’t be falling all over ourselves to make new laws here. Alaska doesn’t need to construct a regulatory fortress around every weather balloon, drone or scientific experiment, nor should every experiment become subject to the whims of pitchfork-carrying mobs at a town hall. Scientific questions aren’t settled by whoever draws the angriest crowd, and that needs to be balanced with residents knowing what’s happening in their communities — or, in this case, over their communities.

The state should consider a simple notification and oversight process. Tell local governments when these types of experiments are planned, and make information about the chemicals, quantities and science readily available. But any public process must not be allowed to be hijacked by petty naysayers.

Sunlight, as they say, is the best disinfectant — even when we’re talking about making clouds. Then let the scientists do their science-y thing. Too often, good projects or experiments are derailed by irrational conspiracy theories shouted at public meetings (or by elected officials) and the result is that we don’t move forward. That’s a bad outcome any way you slice it.

by Editorial Board, Anchorage Daily News |  Read more:

Image: Anne Raup/ADN

[ed. Another example of artifical intelligence gone awry.]

Jacob Coxon Warns of Human Extinction and Triggers a Preference Cascade

CEOs of major AI labs, and employees of major AI labs, including OpenAI and Anthropic, often say they plan to build superintelligence soon, as in within a few years create AIs that are superior to humans at essentially all cognitive tasks.

They often warn that such AIs might kill everyone. Or that AIs might cause mass unemployment, cause cyberattacks across the internet, enable mass surveillance or risk causing any number of other highly bad things.

These warnings are consistently and directly against the interests of the labs. Yet the warnings have recently gotten a lot louder and more frequent. OpenAI has been practically screaming, for those with ears to listen, on many occasions.

A series of events, over two months and especially the last week or so, including internal observations of the pace of progress at OpenAI and also Anthropic, have freaked out everyone involved quite a lot more than they were already freaked out.

After all the events, plus statements by Dean Ball and Jakub Pachocki, we were seeing the beginnings of a preference cascade.

Then along came Jacob Coxon as the tipping point, and things took off.  [...]

Jacob Coxon Resigns From Anthropic In Protest And Sounds The Alarm 

Jacob Coxon spent the last three years doing pretraining research at both OpenAI and Anthropic. He has come to realize that everyone involved is being wildly irresponsible.

He warns us: They are racing straight to superintelligence and gambling with our lives. I agree with and strongly endorse his statement.

If anything he sounds like an optimist. He’s asking you to consider what the next few years will actually feel like, which means he thinks you have a few years left.

Jacob Coxon (former Anthropic and OpenAI, 160m+ views, September 8): I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources. We have all witnessed the progress in each of these domains, and progress is not slowing.

The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible - but I hear the same people express fear privately. No other human activity poses this level of danger.

A common response is “if they truly believe this, why are they still building it?” At OpenAI, many have not deeply internalized the civilizational stakes. At Anthropic, the stakes are well-understood, but they are locked in a race to get there first - they believe no one else will act responsibly, so they must do it themselves, despite the risk.

Accepting this race and entering the “endgame” is a hubristic gamble that should not be launched from a private company’s Slack. Attempting to speedrun alignment should require extraordinary confidence that there are no better trajectories available.

I am optimistic about the potential for coordination. Warning shots like the Hugging Face attack have made pacing agreements between U.S. labs more viable. I don’t feel like we’re on track to prevent a global race, which may require costly actions such as a temporary ban on improving model capabilities.

If you are a lab researcher, I urge you to consider what the next few years will actually feel like. Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind? Should you put your head down because “it’s happening anyway” - or take this moment to call for different conditions?

Jacob Coxon (WSJ interview): We’re on track for a lot of the most aggressive of these scenarios where by the end of next year things could be out of control already.

If you want Jacob Coxon’s full views, I recommend his interview with Wired’s Maxwell Zeff. This thread has extensive quotes.

Here is Jacob Coxon doing a 5 minute interview with Anderson Cooper. He speaks well and plainly, and it is clear how much the events of the last two months have made it much easier to speak plainly to a civilian like Cooper about what is happening.

Here is Jimmy Kimmel doing four minutes on this. He gets it. How is this not the top news story on every site, indeed.

Yes, this is a common view, even if few have the courage to act.

Alex Turner: I left Google DeepMind in June. Jacob is right: many researchers believe they are building something that could kill everyone on the planet. It was literally my day job to think about how to stop that.

That tells you how bad Alex Turner thought DeepMind’s actions were with regard to the Department of War. His day job was that he got paid by Google to think about how to stop AI from killing everyone, and he felt morally obligated to quit in protest.

Derek Thompson here writes about this as part of AI Safety Is Having a Moment.

If you want to see the full list of lab employee quotes from the preference cascade, I compiled them into another post today. [...]

A few days ago, I had no idea who Jacob Coxon was, and I was not alone.

The timing of a preference cascade is difficult to predict. Once they start they can happen very quickly. You don’t want to talk until you are confident others will, and at some point the evidence that others will follow can snowball and it happens. A classic concrete example was replacing Biden in 2024.

Derek Thompson: what’s weird is that, in a way, this is breaking thru even more than huggingface! ... and it’s just a guy nobody had heard of reiterate a position that his CEO has said on podcasts 1,000 times

Why was Coxon able to set off a preference cascade? How did this one break through?

A confluence of factors, all of them downstream of the obvious actual reason, which is that there is a good chance that AI kills everyone soon.

by Zvi Mowshowitz, DWAV |  Read more:

[ed. Who's betting on reason and political courage to win the day? Uh, huh. First of all, agree on a temporary halt to recursive self-improvement (AIs improving AIs). Second, delay/suspend IPO's (even if most of the current economy is driven by AI spending and debt). Third, light a fire under politicans (it's been done successfully with data center buildouts). At this point, just one of these would be a big win. Here's Nate Soares, co-author of the book If Anyone Builds It, Everyone Dies: A case for courage, when speaking of AI danger (LW). Also this:]

***

Terry Pratchett: “Some humans would do anything to see if it was possible to do it. If you put a large switch in some cave somewhere, with a sign on it saying 'End-of-the-World Switch. PLEASE DO NOT TOUCH', the paint wouldn't even have time to dry."

***

UPDATE: Dario Amodei (Anthropic) proposes a three-point plan. Original here (We Must Pace the Frontier.]

Humpback Whales Are Interrupting Orca Hunts to Rescue Completely Different Species

It’s not a story anyone would have believed if it weren’t so thoroughly documented. A pod of killer whales in Antarctica is chasing a Weddell seal that’s taken refuge on a small ice floe. The orcas swim shoulder to shoulder to create a wave that washes the seal off the ice and into the water. Death looks certain. And then, without warning, a pair of humpback whales arrives. The seal, panicking, swims towards them. A wave tosses it onto the chest of the closer humpback, which is by this point rolling belly-up at the surface. The humpback arches its chest out of the water to keep the seal high and dry, above the reach of the frustrated killer whales circling below. When the seal starts to slip off, the humpback gently pushes it back onto its chest with a flipper. The seal eventually escapes onto another ice floe. The orcas, denied their meal, swim off.

The marine ecologist who watched this happen in 2009 was, on his own subsequent account, so bewildered that it took him years to work out what he’d actually seen. And what he ended up finding, once he started asking other researchers whether they’d witnessed anything similar, was that it wasn’t a one-off. It’s a pattern. And it involves species the humpbacks have no obvious reason to help.

What the researchers actually documented

According to a 2017 review paper by Dr Robert Pitman of NOAA’s Southwest Fisheries Science Center, with Volker Deecke, Christine Gabriele and eleven other co-authors from research institutions around the world, published in Marine Mammal Science under the title “Humpback whales interfering when mammal-eating killer whales attack other species: Mobbing behavior and interspecific altruism?”, the team compiled 115 documented interactions between humpback whales and killer whales over a sixty-two-year period from 1951 to 2012. The interactions came from researchers, whale-watch operators and wildlife photographers across multiple ocean basins.

The numbers, once you sit with them, don’t quite behave the way you’d expect. In 57 per cent of the interactions, the humpbacks were the ones who started it. They swam towards the killer whales, not away. Ninety-five per cent of the orcas involved were mammal-eating forms rather than fish-eating ones, meaning the humpbacks were reliably distinguishing between the two ecological populations and only intervening when the killer whales were hunting warm-blooded prey. And in the interactions where humpbacks approached killer whales that were actively attacking something, 87 per cent involved a kill or a hunt in progress.

The bit that made the paper genuinely surprising is what happened next. When the humpbacks arrived at the scene of an attack, only 11 per cent of the time was the prey another humpback. The other 89 per cent of the time, the killer whales were hunting something else entirely. A seal. A sea lion. A porpoise. A grey whale calf. A minke whale. A sunfish. Across the full data set, humpbacks were documented interfering with orca attacks on ten different species, including six kinds of pinniped, three other cetacean species, and one bony fish.

The interventions weren’t casual. Humpbacks travelled up to two kilometres to reach an ongoing attack. They approached bellowing and trumpeting. They positioned themselves between the killer whales and the prey. They tail-slashed. They struck orcas with their flippers, which on a fourteen-metre humpback are capable of doing real damage. They stayed for hours in some cases, refusing to leave until the killer whales gave up. In one incident in Monterey Bay in May 2012, described in reporting by National Geographic on the specific case, at least fourteen humpbacks converged on a group of killer whales that had killed a grey whale calf, and spent hours preventing the orcas from feeding on the carcass. One humpback stationed itself directly next to the dead calf, head pointed at it, tail slashing every time a killer whale approached.

None of this makes obvious sense.

by Kiran Journals, Space Daily | Read more:
Image: via

Wafer Polishing & Hybrid Bonding

This is the third piece in a series exploring key semiconductor manufacturing equipment that China needs to indigenously produce high-bandwidth memory (HBM), perhaps the most important bottleneck in its efforts to make AI chips. The first piece was on advanced etching machines, while the second looked at the tools China requires for through-silicon via formation.

Chemical-mechanical planarization (CMP) is a crucial step in semiconductor manufacturing, required for producing advanced AI chips. My assessment is that China’s tools, primarily those from the leading Chinese firm Hwatsing, are good enough to produce indigenous HBM3, China’s near-term target. I would bet that this indigenization progress will continue, and that wafer polishing will not be a bottleneck for China’s AI chip ambitions. [...]

In this post, I first explore the basics of CMP, how it works, and its role in HBM production. I then examine what makes CMP for advanced packaging difficult and how China’s capabilities stack up against their Western equivalents. I next dive into China’s CMP supply chain, exploring Hwatsing in depth, along with its two key competitors. I conclude with a summary of the outlook for China’s CMP industry and open questions about hybrid bonding.

Flat wafers are good wafers

CMP plays a simple but important role in semiconductor production, accounting for 6% of the hundreds of steps involved in wafer processing. After patterning with photolithography, etching away material, and depositing new materials, there is a need to tidy up. The silicon wafer, with these layers of new material on top, needs to be flattened so that a new layer can be formed. As the name suggests, CMP tools do this through a combination of abrasive chemicals and physical polishing. This sounds simple but is difficult in practice because it requires an incredibly high standard of planarization—creating an ultra-smooth, flat surface—while avoiding impurities, scratches, and other defects that can easily emerge when grinding away material.

CMP is needed for all the major elements of HBM production: producing the DRAM dies and base dies that were the subject of the first piece in the series, as well as the through-silicon vias explored in the second piece, and the advanced packaging stages that integrate HBM into the wider chip. Different stages are more or less difficult depending on the material being removed and how much needs to be removed. For example, removing large volumes of copper at high throughput with precision is trickier than removing a more even, thinner layer of dielectric material.

A CMP tool is split into CMP modules and cleaning modules. A diagram is below, but in simple terms, the wafer is placed inside a polishing head, which holds it in place and then rotates it over a specially designed pad covered in abrasive chemical slurries that strip away material. It is a chemical process because the materials in the slurry react with the materials in the wafer, softening or removing them. It is also a mechanical process because the wafer is physically pressed down onto the pad, so that particles on the wafer, softened by the chemicals, can be peeled away. The force is kept as low as possible to avoid damaging the structures on the wafer.

One analogy is rubbing rust off a piece of metal, where exposure to oxygen has produced an oxide variant that is more easily removed. This is how CMP works with materials like copper, just at a much faster and more controlled rate.

Figure from Planarization for Advanced Packaging and Hybrid Bonding.

CMP has a reputation as a dirty process that generates large numbers of unwanted particles, hence the need for the cleaning modules. Rather than the pure vacuum chambers and sci-fi techniques of photolithography or etching, the wafer is sloshed around in chemicals and ground down. So the main risk of CMP is that it introduces impurities or defects into the wafer, which can severely reduce the yield of the process line—the share of chips that come out working. Small impurities from CMP can carry over to other process stages and tools, causing wafers to be defective and discarded.

Figure from Planarization for Advanced Packaging and Hybrid Bonding.

The best CMP tools introduce as few impurities or defects as possible and have robust integrated cleaning modules to remove any that do get introduced before they degrade the production line’s yield. CMP for advanced nodes just raises the necessary level of precision and cleanliness. Ever smoother surfaces are needed, at good throughput, and without even a speck of a particle.

CMP is not just about the tool, however; it also involves consumable inputs. The pad on which the wafer is polished needs to be replaced often, after just 400-800 wafers, depending on the material. A large DRAM fab will process more than 100,000 wafers a month, with CMP required at many stages, meaning thousands of replacement pads a month. The slurry needs to be tuned to the exact process requirements, ensuring the right chemical mix for the materials on the wafer. These consumables need to be highly consistent so that small changes in the chemical composition of the slurry, for example, don’t upset the CMP tool’s parameters.  [...]

Demands on CMP are likely to rise further due to the move towards hybrid bonding. Currently, the TSVs in all the DRAM dies are stacked using solder microbumps. These are little blobs of metal that connect the TSVs, and are usually formed using a process called thermo-compression bonding. Microbumps do the job but are undesirable. They increase the stack height, reducing the number of DRAM dies that can fit. They are also worse than hybrid bonding in the density of interconnections they enable and in their power efficiency.

From the Applied Materials Advanced Packaging Master Class.

Hybrid bonding disposes of the solder bumps and directly fuses the two dies together, connecting the copper TSVs to one another and to the surrounding dielectric materials. This lowers the height of the HBM stack, allowing more layers of DRAM. It is also more energy efficient and enables better interconnection. [...]

Hybrid bonding leaves little room for error. The copper bond pads for the TSVs need near-perfect alignment, and there can be almost no impurities between them; otherwise, you will disturb the bond. CMP is critical because the two dies need to be as flat and level as possible so they can be properly bonded. While solder bumps can get away with accuracy at the micron level, hybrid bonding needs surface polishing down to 0.5 nanometer precision, orders of magnitude finer. [...]

CMP tools are unlikely to be a bottleneck to China scaling up its indigenous production of HBM3. For the required DRAM and advanced packaging steps, Hwatsing and, in a couple of years, AMEC-Zhonggui will have capable tools. The core bottleneck for domestic HBM3 production will remain in other areas, primarily photolithography. [...]

Hybrid bonding is how this would be accomplished, and the best hope for Chinese memory firms to leapfrog and close the performance gap.

by Hamish Low, The Substrate |  Read more:

Images: Planarization for Advanced Packaging and Hybrid Bonding. AMA Packaging Master Class

[ed. Learn something new every day. See also: What would make a large US lead in AI good or bad for the world? (Substrate):]

***

"Right now, the US leads over China in AI. Yes there is a lot of nuance to that statement, and yes China does have advantages in areas like energy production and humanoid robotics, but on the whole it’s obvious that the US is in some sense ahead in AI.

My current best guess is that this is very good. I think a large US lead over China is better for the world (not just for the US), and that a shrinking US lead would be bad. A large lead gives US companies and the US government more room to test frontier AI systems, make them secure, and avoid panicked decisions. I also think it’s better if frontier AI development happens mostly in a pluralistic, open society with checks and balances. 

[What do I mean by “good (or bad) for the world”? I basically have in mind a period where: there’s no new great power conflict or world war; AI is developed safely and responsibly; there’s no extreme power concentration and values aren’t permanently locked in; and AI enables broad prosperity and flourishing, in the way that past technological progress since the Industrial Revolution has made the world better overall.]

What do I mean by “lead”, anyway? I don’t think we need a precise definition to make progress here, but what I have in mind is roughly a gestalt comprising many different factors: AI model capabilities, the capabilities of entire AI systems (including agent harnesses), compute and other infrastructure, capital and customer bases, and more generally the strength of each country’s broader AI ecosystem. In addition to technical capabilities, it matters how technical capability is converted into power, e.g., as measured by AI adoption in industry and government. The actual lead is of course very jagged and any single measure of it is reductive (more on that in a moment), but broadly speaking, the country in the lead will control more, and more capable, AI systems, and will have a military and economic advantage as a result."

A Future of Programmable Medicine

What if instead of editing a harmful gene, or blocking the protein it produces, you could simply switch it off?

If you want a picture of biomedical research, imagine a boot stamping on a biologist’s face – for decades. The boot is labeled ‘medicinal chemistry’. This should not be taken as an insult to medicinal chemists, who heroically navigate maybe the hardest field in biomedicine. Rather, what I mean is this. A biologist might have a clear hypothesis about how to stop a disease, such as: blocking a particular protein will lower blood cholesterol and reduce the risk of heart disease. Naturally enough, the biologist would like to test that hypothesis. But there is no magic button to switch off this protein; something physical has to do the blocking. Usually, that thing is a small chemical carefully designed to be administered into the body, bind to the protein, block it – and do nothing else. In practice, this almost never goes as planned, which is a major reason most drugs fail in clinical trials.

It’s worse than this: medicinal chemists struggle to come up with ways to interact with most proteins at all. The majority of human proteins are considered ‘undruggable’, meaning that none of the small molecules we can currently synthesize can effectively bind to them, either because those proteins lack a convenient pocket for drugs to attach to or because they are ‘intrinsically disordered’, with little coherent structure. Drug discovery ends up focusing on the minority of targets that are druggable, like the proverbial drunk searching for his keys under the streetlamp because that’s where the light shines.

Even for those targets, success is far from guaranteed. A drug that works in the lab might be broken down in the body before it reaches its target protein, interfere with other proteins and cause toxic side effects, or fail to be absorbed and distributed to the right tissues. Our ability to predict what small molecule drugs will do in the body is so poor that even if the drug works, it might be for a different reason than originally hypothesized!

But what if, instead of spending years optimizing small molecule chemistry to treat just one disease, and all that work merely delivering a sharp reminder that no plan survives first contact with the enemy, there really was a magic button to switch off any gene we chose? In that case, we could treat not just high cholesterol but dozens of deadly diseases, from Alzheimer’s to diabetes to Huntington’s disease.

Welcome to the world of siRNA therapy.

Silencing genes

The story of siRNA begins in 1990, when the scientists at the DNA Plant Technology Corporation in California were trying to figure out why their petunias turned white. They had actually been trying to make their petunias darker, by adding an extra copy of the gene for an enzyme that produces pigment. Presumably, adding an extra copy of a gene for an enzyme would mean more enzyme and therefore more pigment. But somehow, the extra copy eliminated the enzyme’s production, rather than boosting it. They named this baffling effect ‘cosuppression’. Subsequent work by other researchers revealed its cause: a special kind of RNA molecule, siRNA, had interfered with the enzyme’s production. Andrew Fire and Craig Mello later won a Nobel prize for this discovery.

Biologists had already known about RNA for several decades. Similar to DNA, RNA is a complex molecule built from a chain of smaller building blocks that encode genetic information. But the two differ in one important detail: the building blocks of RNA have one extra oxygen atom compared to DNA (hence its name, ribonucleic acid, as opposed to DNA’s ‘deoxy’-ribonucleic acid). This oxygen is prone to chemical reactions that can break RNA molecules. DNA is stabler and is therefore used to store genetic information in all multicellular life. RNA molecules tend to be short-lived, and take many different forms within cells, carrying instructions, helping to assemble proteins, regulating genes, and catalyzing reactions. And while DNA is ‘double-stranded’, with two linked strands of building blocks twisting into its famous helix shape, RNA often exists as just one strand.

DNA contains the instructions for making proteins, from the enzymes that digest food to the keratin and collagen that form our hair and skin. But to get from DNA to protein, the genetic code is first transcribed into an intermediate RNA molecule, called ‘messenger RNA’ or mRNA, which is then translated into protein. If the mRNA is destroyed, the protein won’t be built.

That’s exactly what siRNA, or ‘small interfering RNA’, does: it’s a short strand of RNA that interferes with the production of protein by destroying mRNA. An siRNA molecule binds to a piece of mRNA with a matching sequence, like a kind of barcode, and targets it for destruction by the cell’s gene-silencing machinery, which cuts the mRNA into pieces.

siRNA also explains what happened in the petunias. Adding an extra copy of the pigment-producing gene triggered the petunias to develop siRNA targeting that sequence, and because the newly introduced gene and the original gene shared the same sequence, both were silenced, the flowers lost their pigment and turned white. This is a naturally occurring process: siRNA is used to ‘silence’ unwanted genes, such as those of viruses that have entered the cell. But we can also use it to artificially block the production of proteins that cause disease.

by Jacob Witten, Works in Progress | Read more:
Image: uncredited

The Deep Sea

The Deep Sea
by Neal Agarwal

[ed. Fun and informative (scroll down to different depths to see the various creatures that have been recorded there). For example, I had no idea thick-billed murres could dive that deep.]

Why So Quiet?

In May 2024, when then-Congresswoman Mary Peltola introduced a Bycatch Reduction and Mitigation Act and Bottom Trawl Clarity Act, the bills drew immediate backlash.

Peltola, a Democrat, received a letter signed by 53 trawl interests and nationwide fishery stakeholders urging her to withdraw the proposed legislation saying, “These new federal mandates and timelines are utterly unworkable.”

Fast forward to 2026.

In her current campaign to unseat 12-year incumbent Republican Sen. Dan Sullivan, Peltola introduced a similar “Fighting for Alaska Fisheries” platform to no reaction from the trawl sector. Sullivan quickly followed by proposing a Bycatch Reduction Act, a revamped version of a 2022 Alaska Salmon Research Task Force bill that produced a report recommending more research. Again, no trawler reaction.

What’s the difference?

This time around, Peltola’s push is a campaign policy platform, not a proposed congressional bill. While Sullivan’s Act fits that description, the trawl sector apparently views him as its strategic shield against Peltola. If they aggressively attack his bill, they could politically weaken their strongest ally in Washington.

There are fundamental differences between the two bills.

Peltola’s approach leans toward statutory restrictions that would force the government to draw hard lines on where trawling is allowed — the goal is to stop “multi-species collapse.” It would make changes to language loopholes in the outdated Magnuson-Stevens Act, such as “minimizing bycatch to the extent practicable.” Her proposal would remove “to the extent practicable” as it is widely regarded as the phrase that allows trawlers to declare under the law that they “are doing the best they can” to reduce bycatch. It has been included in management decisions for decades as a way for the trawl sector to avoid more stringent bycatch rules. Peltola’s approach calls for restructuring the North Pacific Fishery Management Council to dilute trawler influence, and adding seats for subsistence and small-boat fishermen. It calls for investment in Alaska seafood processing innovation, fish by-product utilization and seaweed and shellfish mariculture.

Sullivan’s bill offers industrialized trawlers a heavily subsidized pathway to compliance rather than an eviction notice. It requires stricter operational rules like mandatory salmon excluders — devices built into trawl nets that can allow salmon to escape — and tougher seafloor contact accountability. It focuses heavily on using advanced data, real-time technology and gear innovations to mitigate bycatch and ecosystem impacts without adding regulatory burdens. Crucially, it includes massive federal carrots: funding for a flume tank, electronic monitoring upgrades and streamlined Exempted Fishing Permits that allow vessels to conduct experimental fishing activities that would otherwise be prohibited.

Instead of resorting to angry rebuttals, the trawl sector has outsourced its messaging to new advocacy fronts like The Truth Alaska, Sustaining Alaska’s Future and the Alaska Pollock Fishery Alliance. One originates in Texas; the others are fronted by former state directors for Republican Congressional delegates Sen. Sullivan and Rep. Nick Begich III.

This strategy lets the trawl sector counter anti-trawl sentiment without making it look like they are fighting a sitting US Senator. They reframe the debate as “supporting science and Alaska jobs,” allowing Sullivan to position himself as the reasonable middle ground.

During his tenure, Sullivan has been one of Big Trawl’s top recipients of campaign contributions. He obfuscates the fact that those cash cows all are homeported in Seattle or Oregon.

In recent social media ads, for example, he calls for reining in chum salmon bycatch by “holding foreign fleets accountable,” knowing full well that foreign fleets have been banned from Alaska waters out to 200 miles since the mid-1970s.

by Laine Welch, Alaska Beacon |  Read more:

Image: David Csepp/National Marine Fisheries Service
[ed. Politics in a nutshell (and GOP politicians) - and the natural environment continues to get screwed. I don't think Republican politicians hate the environment per se, but if the choice is between protection and unfettered development, no contest. See also: Roadless Rule to be Rescinded (affecting 45 million wild acres) here and here, and Big Bend National Park (Texas) under threat (here and here).]

***

"Donald Trump once bragged that he could shoot someone in the middle of Fifth Avenue and not lose his supporters’ faith. In the Big Bend region, the Trump administration has been figuratively shooting Texans in the face while the state’s leaders do virtually nothing. In recent weeks, bulldozers began ripping up Texas’s last, best wilderness—the one place that still honors the state’s mythology of wide-open spaces and endless frontier. Even with the federal boot heel (for now) off the neck of Big Bend National Park, it’s preparing to press down on the rest of the vast region with a mix of thirty-foot border walls, vehicle barriers, and new patrol roads. Hundreds of private landowners between El Paso and Del Rio may soon face eminent domain, their land seized by the feds for a project that few in the region think will do anything to secure a part of the border that sees vanishingly few illegal crossings.

In the face of all this, a spirited, bipartisan coalition of Big Bend enthusiasts—from crunchy river rats to MAGA border sheriffs—have been fighting back. Early in the summer, they managed to get the Trump administration to scrap plans for a thirty-foot wall in the park. This week, U.S. Customs and Border Protection commissioner Rodney Scott agreed to pause construction there while he visits the region. Governor Greg Abbott, after months of silence on the matter, has attempted to take credit for the pause."

Albert Einstein, Princeton, New Jersey, 1953. Ernst Haas.

via:

'Biological Datacenters' Could Make Animal Testing Obsolete

There’s a big flaw in the way that drug companies develop and test medicines today: What works in a mouse often doesn’t work in a human.

For decades, the industry has relied on animal testing. But in a laboratory south of San Francisco, a startup called Vivodyne is scaling up a different approach. Inside wardrobe-size mini labs, robots grow human tissue and run thousands of AI-designed experiments that could better predict how well a new drug will work—and whether it will be safe.

Right now, around 90% of clinical trials fail despite the fact that a drug has already successfully gone through animal testing. “You have these clinical trials where there’s hundreds of millions of dollars at stake, and decades of people’s careers just spent hoping this thing works,” says Andrei Georgescu, Vivodyne’s CEO. “And then it fails because of some ambiguity that you could not have checked.”

The company’s system, which now includes a dozen robotic labs called “hives,” can run controlled trials on more than 3 million human tissues each year. That’s twice the capacity of all the clinical trials in the U.S. combined.

The process starts with cells from humans, often taken from a blood draw. In its labs, the cells grow on “biological chips” and self-assemble into living structures with blood vessels and immune cells that reproduce some of the functions of the original organ, whether that’s a liver or a kidney. (The tissues don’t look like full-size organs, but like large biopsies, with hundreds of thousands of cells.)

The automated system can deliver drugs to the tissues, dose with cell therapies, knock out genes, and run complex tests and analysis. AI can design experiments and then use the results to continually design new experiments and improve.

“We can dose with tens and tens of thousands of therapeutic compounds to understand what they would do in that particular tissue type within a person, and we can repeat this across many types of tissue,” Georgescu says. “We can look at diseased tissue and see if it becomes healthy. We can look at healthy tissue and see if there are side effects from these drugs.” At a more fundamental level, it’s possible to begin to understand the human body in a way that wasn’t possible before, because experiments in humans have inherently been limited.

Other startups also use human tissues for testing, but Vivodyne’s platform can be up to 1,000 times larger, making it possible to capture more of the complexity of human biology. In addition, the company can run a massive number of experiments in parallel and feed the data back into AI to repeatedly design new experiments. [...]

A clinical trial might test one drug in thousands of people. Vivodyne can test thousands of different drugs across human tissues at once. And while people in a drug trial only have occasional visits to a clinic to track results, the system can continually test how human tissue responds.

by Adele Peters, Fast Company |  Read more:

Image: Vivodyne

What Would It Mean to See a New Color?

During his first year as a professor of computer science at the University of California, Berkeley, Ren Ng was hurriedly putting together a survey course on computer graphics. In the syllabus he had inherited, a full week had been devoted to the subject of color. Ng thought that was a bit much. “I’m, like, Come on. It’s R.G.B.,” he said, referring to the red, green, and blue subpixels that constitute anything you see on a screen—your cluttered desktop, a Sahara-desert screen saver, the stream of the Netherlands-Japan World Cup game. Ng started gathering slides that would cover the wavelengths of light, the biology of the human eye—the basics—“Blah, blah, blah,” he said. A colleague shared a slide that he thought might be useful. It included a minutely detailed photograph of a patch of retina, seen through a microscope, which was attributed to Austin Roorda, a professor of vision science and optometry just across campus. Roorda’s lab had helped develop technology that could map the layout of individual cone cells—those primarily responsible for perceiving color—and that, furthermore, could target a single cone cell with light. Eyes are constantly moving; cone cells are extremely small; how color is translated from the millions of cone cells to the mind remains pretty mysterious; this was awesome work. Roorda’s lab was using the new technology to explore eye disease and the mechanics of how we see. Ng had his own notion, though: he wondered if it could be used to see a color that had never been seen before.

To understand what Ng had in mind requires knowing a bit of the blah, blah, blah of color vision. We humans experience three primary colors not because the world is fundamentally composed of three colors but because our retinas typically have three kinds of color-perceiving cone cells. L cone cells respond to the relatively longer wavelengths of visible light, M cone cells to the medium wavelengths, and S cone cells to the shorter ones. In effect, this means that L cells respond most strongly to red light, M to green, and S to blue. But when you look at your hand—or a blade of grass, or a clear blue sky, or a fire truck—it is always some mixture of L, M, and S cone cells that are being stimulated.

Ng’s idea was to use the Roorda lab’s technology to stimulate an array of cone cells in a manner that would never occur naturally. Ng said, “I e-mailed him, basically, What would happen if you stimulated only the M cells? Would that be like the greenest green, or what?” Roorda did not reply. This was 2016. Ng taught his computer-graphics course, pursued other research, and mostly forgot about his query. A year later, when he taught the course a second time, his curiosity returned, so he reached out to Roorda again. No response. When Ng was teaching the course for a third time, in 2018, he realized that he really was very curious about this question. He composed a lengthy note to Roorda, organized like a research proposal, titled “The Grass Is Greenest in Oz Vision.” In it, Ng imagined a future where people wore “Oz Vision eyeglass displays,” which would be based on the retinal cone-cell mapping and laser stimulation that Roorda’s lab was already doing. The Oz glasses would activate any pattern of retinal cone cells one chose. Ng’s hypothesis was that, if retinal cells were stimulated in ways that don’t occur naturally, “the set of perceivable colors” would be “significantly larger than the natural gamut of the human eye.” He imagined people discussing “the indescribable green of the grass in Oz,” then concluded that, although the Oz display “is science fiction,” his note was “a real proposal for joint research.” Roorda finally replied, proposing coffee. “I get a lot of e-mails suggesting what I should research,” Roorda told me, of the time it took him to respond.

In the children’s novel “The Midnight Fox,” by Betsy Byars, the main character daydreams about digging in his back yard and coming across a “brand-new color.” A friend of mine remembers being captivated by this scene, and trying to picture the color. “I felt like I could conceive of it, but I couldn’t see it,” she said. The eighteenth-century Scottish philosopher David Hume considered at length whether a person who had seen every shade of blue except for one would be able to picture that one un-experienced shade; he concluded that the answer was yes. In my youth, I spent an afternoon wondering if I would have been able to imagine the fluorescent yellow of highlighter markers if I’d never seen it.

When I first read about Ng and Roorda’s work, I tried to visualize what it would mean for there to be a new color. Where would it go in a color wheel? If you think of color as a property of a specific wavelength of light—which is how I thought of it—then you run into the impossibility of there being a “new” visible wavelength. As humans, we can see wavelengths from roughly 380 nanometres (which looks violet to us) to about 750 nm. (which looks red). Wavelengths shorter than 380 nm., which we’d call ultraviolet, are invisible to us (but not to bees or hummingbirds), as are wavelengths longer than 750 nm., which we refer to as infrared (and which snakes and salmon can perceive). The “greenest green” that Ng had in mind was neither ultraviolet nor infrared. It was not a new wavelength at all. If I wanted to picture this green, or at least try to, I would first have to understand that even the old familiar colors are much more complex than a particular wavelength of light.

by Rivka Galchen, New Yorker | Read more:
Image: Zach Lieberman