Nehaveigur

Category: Technology

  • Big in 2025

    Here is a list of scientific and engineering news of 2025, ranked by potential impact. I like the idea of considering both the probability that a finding holds up and how important it’d be. If it turns out to be true. My biggest criticism is that some of the findings aren’t…

  • Can Pharma Scale?

    Pharma and tech are different industries. For example, tech benefits from network effects (if everyone uses LinkedIn it makes more sense to join), pharma doesn’t. In other words, tech scales. Could the massive success of GLP-1 drugs change this? Is the recent valuation of Eli Lilly at of $1 trillion…

  • Overfitting Towards Blandness

    Our culture has become bland, as evidenced by fashion, building, cars, book covers, and household objects all looking the same. Even people seem to be less weird than they once were. It’s an observation that has been made by me and others, on this blog and in many other places,…

  • The Hundred-Light-Year Diary

    Thinking about forecasting and AI, I sometimes remember this story by Greg Egan. It was published as part of his collection Axiomatic. Here is my review of Egan’s novel Distress.

  • Data Archival

    You don’t have a lot of options if you want to preserve sure your data (your photos for example) for many years without any maintenance. Data stored on hard drives and flash memory like SD cards will degrade within a decade or two. Cloud storage like Google Drive requires the…

  • Child of Freedom, Parent of Prosperity

    How much should the government spend on science? One view is that it should spend a lot, since every dollar pays back many times over. The other view is that governmental research spending is wasteful and that many of its benefits could be better realized by industry-funded research. I know…

  • Death by AI

    The most likely cause of death today is AI. It’s a reasonable statement. The most common cause of death right now is ischemic heart disease, which kills around 13% of people worldwide. However, if we assume that the chance of superintelligent AI turns against humans is higher than 13%, and…

  • Life Without Clocks or Mirrors

    I’ve always been immoderatly clock-oriented. But that was part of what seemed wrong with my infrequent periods of actual labor: the deadly predictability of jobs everyone sighs about, a glut of clocks and my thin neck twisting to their perfect circles, around and around and around. This is from Wolf…

  • Getting Started With AI

    If you haven’t yet used AI, or if you’re thinking about paying for a premium AI, this is a good guide. I find paying for the more powerful AI features worthwhile.

  • Free Energy

    The 1990s were the golden age of free energy: Technologies that, through new or underappreciated physics, generated abundant and clean electricity. It wasn’t about boring old ideas like wind or solar, but about exciting new inventions like zero-point energy and cold fusion. There were books that described how Nikola Tesla…

  • Effective Mess

    Chaos always defeats order because it is better organized Terry Pratchett A few weeks ago, I shared a pointer to a podcast about the internal messiness of AIs. They don’t represent the world neatly, but in a way that has been described as Fractured Entanglement. Interestingly, organizations suffer from the…

  • In Tech America, AI Fact-Check You

    This happens to scientists fairly often: You remember some finding you came across a few weeks ago. You don’t remember where you read it, but if you want to include it in a paper you’re writing, you’ll have to cite the source. It’s frustrating. LLMs like ChatGPT are helpful in…

  • Fractured Entanglement

    The most interesting perspectives on AI can be encountered on the Jim Rutt Show, which I have previously referenced with regards to AI risk. In a more recent episode, Rutt interviews Ken Stanley about his Fractured Entanglement Representation hypothesis. A preprint describing this hypothesis is available here. Here is how…

  • Electric Network Frequency Analysis

    Given an audio or video file, it’s possible to determine where and when it was recorded based on the electrical power grid’s hum in the background. Small fluctuations in the power supply to devices and appliances that are close enough during the recording, while not perceptible to our ears, can…

  • Not People

    With ChatGPT-5 just having come out, Adam Mastroianni has posted a timely reminder on Experimental History: Trying to understand LLMs by using the rules of human psychology is like trying to understand a game of Scrabble by using the rules of Pictionary. These things don’t act like people because they…

  • Spinning Sun-Powered Space Catapult

    For years now, I’ve been following what’s happening in the field of interstellar travel. Not closely, but close enough to know what kinds of technology are out there. Paul Gilster’s blog Centauri Dreams is a good resource. Because of the large distances involved, and because of the energy needed to…

  • The 1,000 Year House

    There is a series of blog posts by Brian Potter on how one would build a house that would last for one thousand years. It’s a fun question to think about, not only for houses but also for other buildings we may want to last a long time, like Star…

  • Talking of Children and AI

    We talk about our children and AI the same way. We say, “Did you notice what they can do now?” and “Can you believe that cute thing they’re coming up with?” and “I don’t know where they got that from, it definitely doesn’t come from me.“ We’re also similarly annoyed…

  • Computational Irreducibility

    Simple rules can lead to complex outcomes. If those outcomes aren’t predictable in any other way than executing the rules, this is called computational irreducibility. The concept was first proposed by mathematician Stephen Wolfram in his book, A New Kind of Science, which I read early on during my PhD…

  • Bigger Brains

    To a first approximation, bigger brains = more neurons = smarter. Dig deeper, and it turns out to be more complicated than that. Honeybees have ten times fewer neurons than zebrasfish, yet by some measures are just as smart. Even discounting species-specific differences, the relationship between neuron count and intelligence…

  • AI Rationalizations

    AIs like ChatGPT’s o3 take time to think before they answer. While doing so, o3 provides some commentary on its thinking process. For example, it mentions the websites it consults to find an answer. The strange thing is that much of the commentary may not at all be related to…

  • Questions about Intelligence

    Do we understand intelligence enough to formalize it in mathematical or computer science terms? We don’t, because otherwise there’d be no need for AI benchmarking. In fact, the concept of intelligence is remarkably fuzzy. More questions about: Cave art | Domestication | Appearance

  • Primitive Technology

    Primitive Technology is a popular YouTube channel about making things from scratch without any modern tools or materials. It’s not just what it’s about that I like, but, to quote Roger Ebert, how it’s about what it’s about. There is no talking. Context is provided in terse subtitles that remind…

  • o3 Can’t Stay Silent

    I’ve recently speculated about how it may be possible to use AI to monitor the news and to alert me only when anything noteworthy happens. The advantage would be that there’d be no need to check the news, since I know that I get an alert if something actually important…

  • Stainless Steel

    The oldest piece of kitchenware I own is a stainless steel teapot. The previous occupants had left it behind in a flat in England I rented with my girlfriend when we first moved in together 15 years ago. We have been married for eleven years, moved to California, and had…

  • LMMs as Information Retrieval

    Are we close to developing large language models (LMMs) that have artificial general intelligence (AGI) soon? Some think we’re already there, but according to this paper, assuming that LMMs are on the cusp of AGI is based on a misunderstanding. Cosma Shalizi, one of the co-authors, has more on his…

  • AI Benchmarking

    A month ago, I observed that out of three big magazines dedicated to literature, none had a recent discussion of AI and what it means for writers. Since then, Paul Taylor has published a piece on DeepSeek in the London Review of Books. He mentions how the performance of AIs…

  • Commodification

    A friend sends me Christmas cards every year that feature cartoon versions of him and his wive. They hire an artist to create the images from photos that they have taken during the year. Now that it has become easy to generate such drawings using AI, I wonder if he…

  • When Being Smart is Not Enough

    What kinds of problems can be solved with more intelligence, and for which is intelligence not sufficient? Dynomight speculates that a superintelligent AI could solve most problems in philosophy and maths, only improve on forecasting a little, and not solve physics or cure cancer. This is in line with my…

  • Why are More Neurons Better?

    This is an excellent question whose answer is only obvious at first glance, asked by Scott Alexander on Astral Codex Ten. The correlation between the number of neurons and intelligence holds for biological brains and for AI, if we take the number of parameters to be equivalent to the number…

  • The Humanities are Avoiding AI

    Few people working in the humanities have extensively tested the latest large language models, and most people base their opinions on what they have heard or read others say about them. This is surprising, as those who write and read for a living are well positioned to judge an algorithm…

  • AI in Biology

    AI will soon design effective and safe drugs for any ailment. At least that’s commonly assumed, and on the surface it’s a reasonable prediction. After all, AI can already predict protein structures and sometimes even the effects of genetic variants. Having worked in drug development for a decade and having…

  • AI Overreliance

    I use AI at work and for fun, but I’m worried what it may do to us individually and as a society if we’re not careful. Erik Hoel discusses one particular concern: We may become too reliant on AI doing our thinking for us. There isn’t convincing evidence either way…

  • Do Composers Die Young?

    When Mozart was my age, he was dead for five years already. He had gone from composing to decomposing. He wasn’t the only composer to die young. Chopin died at 39, Gershwin at 38, Felix Mendelssohn at 38. This made me wonder if there’s a pattern. Are composers splendid torches…

  • Paperclip Apocalypse or Profit Apocalypse?

    Let’s hope AI doesn’t turn the universe into paperclips and remains aligned with its masters. Even so, this could lead to suboptimal outcomes. Imagine AIs running corporations, replacing its employees, either from the bottom up or from the CEO down. If those AIs are well aligned with the company’s missison,…

  • Implementing AI

    The genetics research team I manage has been experimenting with AI agents for a few months now. Within my biotech company, we’re well suited to drive the implementation of AI since we have both the ability and the motivation. We know how to code and we are good at dealing…

  • Less is More

    I have previously written about how sometimes, knowing less can be an advantage. One example of this was that it’s easier to detect if someone is lying when only listening to them compared to a situation where we also watch them talk. Most of the information that predicts if someone…

  • The Cost of Transportation

    From Austin Vernon’s blog post on reducing human and freight transportation costs: It might be difficult to fathom that a mature sector like ground transportation could be 10x cheaper and have 10x greater throughput on existing right-of-way. But that is what the combination of electrification and driving automation software can…

  • Moon Diagrams

    Bartosz Ciechanowski has created a wonderful page explaining a lot of what there is to know about the moon using animated diagrams. Also look at his past work, including on sound, watches, the internal combustion engine, curves and surfaces and the earth and sun system. Here‘s more on the Apollo…

  • Resist Summary

    This is from Simon Sarris’ blog, The Map is Mostly Water: It is an interesting feature of stories and fiction that they resist summary. You cannot read a summary of Anna Karenina and somehow stockpile its pleasures and charms. Narrative resists compression. I think avoiding summary is even more important…

  • Thresholds

    From L. M. Sacasas’ blog, The Convivial Society, in which he asks if and when technology is beneficial to the individual. One way view this is through Ivan Illich’s concept of thresholds. Illich invited us to evaluate technologies and institutions by identifying relevant thresholds, which, when crossed, rendered the technology or…

  • Boring News

    I like my news like I like my visits to the doctor: Unemotional and infrequent. I’ve recently written about one way to achieve this. Packy McCormick introduces another way on Not Boring: News stories inspired by a prediction market and selected by an AI, which he thinks will reduce bias…

  • Total Recall

    Around the time the iPhone came out 17 years ago, Charles Stross wrote an article in which he imagined a future in which all of human experience is recorded: With your phone converting all the speech it hears to text […], and indexing it by time and location it becomes…

  • Minimizing Engagement

    In recent years, I’ve tried to limit my news consumption. I’d rather spend my time and attention on things where I can make a difference. I couldn’t put it any better than Dynomight does here: One of my strongest beliefs is that way too many people allow politics to play…

  • Railguns

    In Jules Verne’s 1865 science fiction novel From the Earth to the Moon, three men travel to the moon aboard a projectile launched from a huge cannon. In reality, launching crews this way is impossible because of the acceleration required to reach Earth’s escape velocity of 11.2 km/s. Even using…

  • Wikibibliography

    I wish there were a website that curated the best resources to learn about any topic. For example, if I want to learn about the history of Kabul, Wikipedia offers an introduction, but there’s no list of links to the best documentaries, online courses, websites, papers, videos or X accounts…

  • The Precautionary Principle

    Or: Should you really never change a working system? Given an innovation whose future positive and negative impact are uncertain, which position should be taken regarding its adoption? The precautionary principle states that proponents of the innovations should have to prove its harmlessness before it is introduced. That’s because we…

  • AI Tutoring

    I’m increasingly encountering the belief that capable leaders have become rarer or have even disappeared. While most often observed in the political realm, some argue that there are also fewer scientific leaders. Erik Hoel makes the point that today there are fewer geniuses of any description. I don’t think it’s…

  • Bicycle Conversion

    Leaving a hardware store, I saw an e-bike unlike any I had seen before chained to a lamppost. It was a standard road bike with two motors added to the frame. They were connected to the rear wheel hub with separate bike chains. The battery was housed in a hard…

  • Power Table

    The only sort of power that ultimately matters is of course physical power, defined as energy per unit time. I’m not used to thinking about power and lack any concept of what a watt or kilowatt is. To remedy this, I made the table below. A more comprehensive but also…

  • The Maniac

    Biographies are a waste of time. There’s little to be gleaned from the lives of those we admire. The details of someone’s childhood or their private lives rarely hold any explanatory power for their achievements. If there is any generalizable insight, it can be summarized in a few paragraphs. The…

  • God Hates Singletons

    Wolf Tivy on Palladium explores the risk of a single artificial intelligence, through recursive self-improvement, taking over the world. This hypothetical event has been termed the Singularity and has been discussed in depth by Nick Bostrom and others. For a current perspective on AI risk, I recommend Jim Rutt’s podcast.…

  • Can AI Solve Science?

    To this ultimate question we’re going to see that the answer is inevitably and firmly no (Stephen Wolfram).