In Isaac Asimov’s The Caves of Steel he imagines a future in which most of humanity lives in gigantic cities, extending many levels underground. This leaves the majority of the Earth’s surface for industrial farming, necessary to feed all those densely populated cities. If you take a similar strategy, however, and keep the population at sustainable numbers, this could maximize land for natural ecosystems and also minimize the environmental footprint of the average person. So if we are going to plan our civilization around environmental sustainability we would not necessarily need Asimovian megacities, but we could concentrate the population in cities and in densely developed areas around cities, and leave large stretches of land in between undeveloped. This is far better than endless suburban sprawl.

But of course, we are not starting from scratch, and our current layout was not planned by some central committee but evolved organically. Pragmatically, the question we need to ask is – where do we go from here. Cities are growing dynamic things, so we can use city development to move in a certain desirable direction, even if we can’t tear it all down and start anew. This means it is important to study what the best city planning and development would be going forward.

When most people think of city planning to reduce the carbon emissions of transportation, the first thing that comes to mind is planning city centers so that they are walkable/bikeable and to provide public transportation, in order to minimize reliance on fossil-fuel burning cars. This also has the advantage of reducing city traffic, which can be a nightmare. However, a recent study suggests that, while important, this may be of secondary concern with respect to impact on CO2 emissions. For many cities, especially those with a single concentrated hub (as opposed to multicentric cities, like LA), the most impactful strategy might be the densification of a ring surrounding the city center. The range of this ring depends on the city, but is something like 10-20 km for a typical large city, but can extend to 40 km. Increased density can be accomplished with infill development, as many such zones are only moderately developed leaving lots of room for densification.

The idea is that the workplaces will be concentrated in the city center, and the workers will live in the ring around the city center, minimizing their commute distance. This could have a significant impact on the average commute distance that people have and therefore their transportation carbon footprint. This approach would work better for some cities than others, depending on geography. This plan could also maximize the impact of public transport, like buses and trains, dedicated to bringing people back and forth from this densified ring to the city center.

Continue Reading »

It is old news that all the tech we now live with is constantly gathering data about us. It is important, however, not to become complacent about this or to assume the situation cannot or is not getting worse. Pretty much every piece of digital technology that we interact with likely is gathering some personal information about you which is used to target advertising and to sell to third parties. Regulations in most countries are inadequate and fail to keep up with technological changes.

One of the latest venues to soak up information about you may be surprising – your car. Cars are increasingly computerized, and they typically collect driving behavior data – how fast your drive, how hard you break, and how tight you make turns. But also, some vehicles have cameras facing the driver which means they can detect your behavior visually. Sometimes this is sold as a safety feature, to tell if you are too sleepy or inebriated to drive. Sometimes this is part of a system to get your insurance company to reduce your rates if you think you are a safe driver. But often it is done without disclosure. Recent GM was found guilty of collecting and selling such data without the permission of the user, and was banned for doing so for five years. But many other car manufacturers also do this.

All they really have to do is bury some disclosure deep in the user agreement, which functionally nobody reads, and they are covered. You may have the ability to opt-out of such data selling. Of course, putting the burden on the end user to find and read any such disclosures and then go through the steps necessary to opt out of data selling is a huge problem. In fact insurance companies will buy data from car companies and then use that data to increase your insurance premiums, without you opting into any of it.

The basic fact is that collecting data from users, packaging that data and then selling it to third parties is a huge industry. It is estimated that globally this is a $240 billion industry. When that kind of money is on the line, companies are going to do everything they can to capitalize on it, while avoiding legal issues by either flying under the radar or hiding behind legal fig leaves (like the buried consumer disclosures). They will also use that money to lobby the government to let them continue to do so, or even to mandate certain things that will help this industry. For example, some car monitoring technology is sold as a safety feature, and it can legitimately be used for this purpose. Others are convenience features, like GPS. But once all the sensors and cameras are in place, they will soak up all the data they can – because that data is worth billions.

Continue Reading »

Richard Dawkins is a public intellectual of some renown, although not without his controversies. So it is noteworthy when he writes an article claiming that the chatbot Claude is likely conscious. I found the article fascinating, not because I agree with his core claim or feel that he has contributed anything significant to the conversation, but because it seems to represent a scholar and deep thinker writing about a topic in which he lacks specific expertise. I also see no evidence in the article that he engaged meaningfully, or at least adequately, with a topic expert. As a result he makes some thoughtful and instructive errors.

He begins with a discussion of the Turing test, which has long been discussed as an early thought experiment about how we might determine if an AI is actually conscious. Dawkins essentially accepts the Turing test and write:

“It was one thing to grant consciousness to a hypothetical machine that — just imagine! — could one day succeed at the Imitation Game. But now that LLMs can actually pass the Turing Test? “Well, er, perhaps, um… Look here, I didn’t really mean it when, back then, I accepted Turing’s operational definition of a conscious being…””

He feels saying that LLMs have passed the Turing test but still not accepting them as conscious is moving the goalpost. However, the Turing test was never generally accepted by AI experts or philosophers as a true test of consciousness. Rather, it was understood that such a test really is only a measure of a machine’s ability to imitate human speech. I wrote about it in 2008, writing: “Ever since Alan Turing proposed his test it has provoked two still relevant questions: what does it mean to be intelligent, and what is the Turing test actually testing.” I went on to write:

“But I can imagine a day in the not-too-distant future when such AI can pass a Turing test. The algorithms will have to become much more complex, allow for varying answers to the same question, and make what seem to be abstract connections which take the conversation is new and unanticipated directions. You can liken computer AI simulating conversation to computer graphics (CG) simulating people. At first they appeared cartoonish, but in the last 20 years we have seen steady progress. Movement is now more natural, textures more subtle and complex. One of the last layers of realism to be added was imperfection. CG characters still seem CG when they are perfect, and so adding imperfections adds to the sense of reality. Similarly, an AI conversation might want to sprinkle some random quirkiness into the responses.

The questions is – will sophisticated-enough algorithms running on powerful-enough computers ever be conscious? What Loebner is saying, and I agree, is that the answer is no. Something more is needed.”

Basically, the limitation of the Turing test is that it is looking only at output, and therefore there is no way to distinguish the output of true consciousness from a really good simulation. This is not a new idea, and no one is moving the goalpost. We need to know something about how a computer is working to conclude whether or not it is conscious. What LLM experts will tell you is that these chatbots are just really good autocompletes – they are mimicking language, and since language is how we communicate thoughts, this creates the powerful illusion that they are mimicking thought, but they aren’t. They do not think, they do not truly understand.

Continue Reading »

I came across a few news items that I could possibly write about today and couldn’t decide which to cover, so I will write about all of them, since they all relate to renewable energy. The first is a new study comparing direct air capture (DAC) to installing new wind and solar. This is a direct comparison between these two options, to see which provides the most bang for the buck.

DAC involves taking CO2 directly out of the atmosphere in order to mitigate carbon release through burning fossil fuels. If this technology were sufficiently efficient it could be hugely useful in reducing future climate change. This is the only approach that can potentially have a negative carbon footprint, actually reducing the amount of CO2 in the atmosphere. Other technologies simply reduce the amount released. This negative carbon factor is highly attractive since it could theoretically zero out our carbon release and even take us back in time to an atmosphere with less CO2. Right now, it should be noted, we are not only continuing to release massive amounts of CO2 into the atmosphere, the amount continues to increase. In 2025 the world emitted 38.1 billion tonnes, of carbon, a 1.1% increase over 2024.

But there are problems with DAC – it is currently not very efficient and is not scalable enough to have enough of an impact. Also, the efficiency of DAC depends heavily on how you power it – if you connect it to the grid and there is some fossil fuel energy on that grid, you may actually increase CO2 rather than decreasing it. Ideally DAC would be powered entirely by low carbon energy sources. This is why critics of DAC argue that it simply makes no sense to deploy this technology before we have decarbonized the energy sector, which we should do first.

Continue Reading »

I thought everyone needed one more thing to worry about, so here you go: evolving AI. When I hear this phrase I think of two things. The first are AI systems designed to simulate organic evolution. The second are artificially intelligent systems that are capable of evolving themselves. That latter one is the type you need to worry about.

Systems that simulate evolution already exist – Avida, Biogenesis,  Grovolve, Tierra, Framsticks: and others. They basically have some code that competes for some resource or to complete some task and the code randomly mutates and reproduces. That’s it, all you need for an evolution simulation. Code can compete for computer resources, or be a physics simulator with digital creature trying to move quickly across terrain. These are sometime gamified for entertainment, but are also used for serious research, to study patterns within evolutionary systems. I would love to see these kinds of systems get more and more sophisticated, even to the point of reasonably simulating living systems. Such systems could be used to test hypotheses about evolution – and would also disprove a lot of silly creationist talking points.

But now we are talking about evolvable AI – AI systems that are capable of developing themselves through evolutionary processes. A new paper in PNAS discusses the potential power and risks of such systems. They echo they kinds of issues that have been explored in science fiction for decades. The authors write: “Evolvable AI (eAI), i.e., AI systems whose components, learning rules, and deployment conditions can themselves undergo Darwinian evolution, may soon emerge from current trends in generative, agentic, and embodied AI.” The results, they argue, have not been adequately addressed when discussing the potential risks of rapidly developing AI ability.

The authors distinguish two types of evolving AI – breeder systems and ecological systems. In breeder scenarios the programmers are in control of the process, selecting which code to “breed” and evaluating the outcome. This process is like a digital version of domestication, and has the potential, if done wisely, to maintain control. In fact, systems can be bred to have greater predictability and control. There are still risks here. So far humanity has not bred an animal to be more intelligent than humans. This could theoretically happen with AI, resulting in emergent behavior not specifically selected for that could get out of the control of human programmers.

Continue Reading »

The recent rapid advance in the capabilities of artificial intelligence (AI) applications I think qualifies as a disruptive technology. The term “disruptive technology” was popularized in 1997 by Clayton M. Christensen. To summarize, a disruptive technology is “an innovation that fundamentally alters the way industries operate, businesses function, or consumers behave, often rendering existing technologies, products, or services obsolete.” AI is potentially so powerful, and changing so quickly, that it is challenging to optimally regulate it. We are caught in a classic dilemma – we do not want to hamper our own competitiveness in a critical new technology, but we also don’t want to unwittingly create new vulnerabilities or unintended negative consequences. For now we seem to be erring on the side of not hampering competitiveness, which basically places us at the tender mercies of tech bros.

Which is partly why I found the conflict between Anthropic and the Department of Defense (still the legal name) so fascinating. In short, Anthropic’s powerful AI application, Claude, has at least two significant internal “red lines” or guardrails – it cannot be used for massive domestic surveillance, and it cannot be used for final military targeting, without a human in the loop. Anthropic CEO Dario Amodei has not backed down on this – he says that the first restriction on domestic surveillance is simply a matter of ethics. The second restriction, however, is mainly a matter of quality control – their system is still vulnerable to hallucinations and is not reliable enough to count on for final targeting decisions. Hegseth has criticized his concerns as “woke” and a critical vulnerability for the US military. More charitably, he say essentially that the US military is using the application lawfully, and should not be restricted in any lawful use of the software. Others have also stated that in an emergency they have to know the software will do whatever they ask it.

This conflict has many deep implications, and is beyond what I intend for this blog post. What I want to focus on is the fact that an AI application is creating this ethical dilemma, and forcing us to ask – who should control such awesome power, the CEO of a tech company or the Federal government? It seems that we are facing or about to face many similar questions provoked by the disruptive nature of recent AI applications.

Continue Reading »

Last week I wrote about the possibilities of genetically engineering humans. The quickie version is this – we are already using genetic engineering (CRISPR) for somatic changes to treat diseases, and other applications are likely to follow. Engineering germline cells, which would get into the human gene pool, are legally and ethically fraught, but it’s hard to predict how this will play out. I have also written often about genetically engineering food. I think this is a great technology with many powerful applications, but it should be, and largely is, highly regulated to make sure that anything that gets into the human food chain is safe.

I haven’t written as much about genetically engineering pets, and this is likely to be the lowest hanging fruit. That is because pets are neither food nor are they a human medical intervention. But that does not mean they are not regulated – they are regulated in the US under the FDA and USDA. Genetic engineering is treated as an animal drug, and must be deemed safe to the animals being engineered. The USDA also can regulate engineered plants and animals to make sure they do not pose any risk to the environment, humans, or livestock. This makes sense. We would not want, for example, to allow a company to release a genetically engineered bee, pest, or predator into the environment without proper oversight.

Pets, as a category, are domesticated, are not intended to be used as food, nor are they intended to be released into the wild. I say “intended” because pets can become food for predators, and they can escape or be released into the wild, and even become feral. But these contingencies are much easier to prevent than with food or wild plants or animals. For example, if you get a rescue pet, it has likely already automatically been spade or neutered. One easy way to reduce risk would be to make any GE pet sterile, which is likely what the company would want to do anyway to prevent violation of their patents through breeding. In short, it seems that reasonable regulatory hurdles should not be a major problem for any effort to commercialize GE pets.

Continue Reading »

Are we getting close to the time when parents would have the option of genetically engineering their children at the embryo stage? If so, is this a good thing, a bad thing, or both? In order for this to happen such engineering would need to be technically, legally, and commercially viable. Let’s take these in order, and then discuss the potential implications.

The main reason this is even a topic for discussion is because genetically engineering is technically feasible. Obviously we do it to plants and animals all the time. We also have increasingly powerful and affordable technology for doing so, such as CRISPR. This is already powerful and practical enough for small startups to perform CRISPR as a service, if it were legal. We already have FDA-approved CRISPR treatments, and have performed personalized CRISPR therapy. CRISPR is fast and affordable enough to have made its way into the clinic. But there is a crucial difference between these treatments and genetic modification – these treatments affect somatic cells, not germ-line cells. This means that whatever change is made will stay confined to that one individual, and cannot get into the human gene pool. What we are talking about now is genetically modifying an embryo at an early enough stage that it will affect all cells, including germ cells. This means that these changed can be passed down to the next generation, and effectively enter the human gene pool.

This difference is precisely why there is regulation dealing with such procedures in many countries, including the US. In the US the situation is a little complex. It is not explicitly illegal to perform germ line gene editing on humans. However, there is a ban on federal funding for any such research.  This does allow for private funding of such research, but any resulting treatment would still need FDA approval, which is highly unlikely in the current environment. Despite this, there is discussion among several startups to start exploring this idea. Why this is happening all at once is not clear, but it seems like we have crossed some threshold and startups have noticed. With current regulation, where does that leave us regarding our three criteria?

Technically a CRISPR-based germ-line treatment for humans is possible. We do have the technology. What needs to be worked out is specific changes and their results. This would require clinical trials, and that is the main stumbling block in the US and some other countries. It seems unlikely the FDA would approve such trials, and therefore there would be no way to even work towards FDA approval. A company could theoretically do privately funded studies that are not part of FDA approval, but they would still need ethical approval (IRB approval) for such studies, which may prove difficult (although not necessarily impossible). Such research could be carried out in countries with more lax regulations, however. Over 70 nations have such regulations, which means many do not. So technically we are theoretically close to having marketable treatments designed to change actual human genetic inheritance.

Continue Reading »

Many teachers are panicking over AI (artificial intelligence), and for good reason. This goes beyond students using AI to cheat on their homework or write their essays for them. If you have AI essentially think for you, then you will not learn to think. On the other hand optimists point out that AI can be a powerful tool to aid in learning. It all comes down to how we use, regulate, and manage our AI tools.

The cautionary approach was captured well, I think, by Mark Crislip in this SBM commentary, in which worries about the effects of AI on doctor education. How will a new generation of physicians learn how to think like expert clinicians if they can have AIs do all their clinical thinking for them? My question is – is AI fundamentally different than all the other technological advances that have come before. Did calculators take away our ability to do math? The answer appears to be no. Students still gain basic math skills at the same rate with or without access to calculators. But there are lots of confounding factors here, and so some teachers still warn of allowing kids access to calculators too soon. Others point out that access to calculators has simply shifted our math abilities, away from basic operations toward more modeling, problem solving, and complex concepts. It seems we are in the middle of the same exact conversation about AI.

We can also think about things like GPS. My ability to navigate from point A to point B without GPS, or to navigate with maps, has definitely declined. But using GPS has also made my navigating to unfamiliar locations easier and more efficient. I would not want to go back to a world without it.

Continue Reading »

As we anticipate the Artemis II launch, now slated for early April with plans to take four astronauts on a trip around the Moon and back to Earth, NASA has been unveiling some significant changes to its plans for returning to the Moon and beyond. If you have fallen behind these announcements, here is a summary of the important bits.

Artemis II will continue as planned, marking the first crewed deep space mission since 1972 (Apollos 17). The original plan was for Artemis III to land on the Moon in 2027, but this mission has been pushed to an Artemis IV mission in 2028. A new Artemis III mission has been inserted – this will go only to low Earth orbit (LEO) and will test the integration of all the systems necessary to land on the Moon. This will include docking with one or both of the two landers, one being built by SpaceX and one by Blue Origin. This sounds like a really good idea, and it did seem unusual that they were planning on going straight to the Moon without ever test docking with the lander.

Even though landing on the Moon will be delayed by at least a year, NASA says this will set them up to have at least annual landings on the Moon after that, with a goal of a landing every six months. The reason for this frequent pace is the the more recent announcement by NASA last week – that they are putting on pause plans for a Lunar Gateway in lunar orbit and instead are going to focus on building a permanent Moon base near the lunar south pole.

In order to make this possible, and to support the future Moon base (no word yet on whether this will be called Moon Base Alpha, as it should) NASA plans about 30 uncrewed robotic landings on the Moon every year. They will be scoping out the location for the base and delivering equipment and supplies.

Continue Reading »