Full Width 
Shortcuts 
Sign Up 
Log In 
Articles 
Reading List More from pcloadletter
Generative AI will probably make blogs better. Have you ever searched for something on Google and found the first one, two, or three blog posts to be utter nonsense? That's because these blog posts have been optimized not for human consumption, but rather to entertain the search engine ranking algorithms. People have figured out the right buzzwords to include in headings, how to game backlinks, and research keywords to write up blog posts about things they know nothing about. Pleasing these bots means raking in the views—and ad revenue (or product referrals, sales leads, etc.). Search Engine Optimization (SEO) may have been the single worst thing that happened to the web. Every year it seems like search results get worse than the previous. The streets of the internet are littered with SEO junk. But now, we may have an escape from this SEO hellscape: generative AI! Think about it: if AI-generated search results (or even direct use of AI chat interfaces) subsumes web search as a primary way to look up information, there will be no more motivation to crank out SEO-driven content. These kinds of articles will fade into obscurity as the only purpose for their existence (monetization) is gone. Perhaps we will be left with the blogosphere of old with webrings and RSS (not that these things went away but they're certainly not mainstream anymore). This, anyways, is my hope. No more blogging to entertain the robots. Just writing stuff you want to write and share with other like-minded folks online.
I write this blog because I enjoy writing. Some people enjoy reading what I write, which makes me feel really great! Recently, I took down a post and stopped writing for a few months because I didn't love the reaction I was getting on social media sites like Reddit and Hacker News. On these social networks, there seems to be an epidemic of "gotcha" commenters, contrarians, and know-it-alls. No matter what you post, you can be sure that folks will come with their sharpest pitchforks to try to skewer you. I'm not sure exactly what it is about those two websites in particular. I suspect it's the gamification of the comment system (more upvotes = more points = dopamine hit). Unfortunately, it seems the easiest way to win points on these sites is to tear down the original content. At any rate, I really don't enjoy bad faith Internet comments and I have a decent-enough following outside of these social networks that I don't really have to endure them. Some might argue I need thicker skin. I don't think that's really true: your experience on the Internet is what you make of it. You don't have to participate in parts of it if you don't want. Also, I know many of you reading this post (likely RSS subscribers at this point) came from Reddit or Hacker News in the first place. I don't mean to insult you or suggest by any means that everyone, or even the majority of users, on these sites are acting in bad faith. Still, I have taken a page from Tom MacWright's playbook and decided to add a bit of javascript to my website that helpfully redirects users from these two sites elsewhere: try { const bannedReferrers = [/news\.ycombinator\.com/i, /reddit\.com/i]; if (document.referrer) { const ref = new URL(document.referrer); if (bannedReferrers.some((r) => r.test(ref.host))) { window.location.href = "https://google.com/"; } } } catch (e) {} After implementing this redirect, I feel a lot more energized to write! I'm no longer worried about having to endlessly caveat my work for fear of getting bludgeoned on social media. I'm writing what I want to write and, if for those of you here to join me, I say thank you!
Here we go again: I'm so tired of crypto web3 LLMs. I'm positive there are wonderful applications for LLMs. The ChatGPT web UI seems great for summarizing information from various online sources (as long as you're willing to verify the things that you learn). But a lot fo the "AI businesses" coming out right now are just lightweight wrappers around ChatGPT. It's lazy and unhelpful. Probably the worst offenders are in the content marketing space. We didn't know how lucky we were back in the "This one weird trick for saving money" days. Now, rather than a human writing that junk, we have every article sounding like the writing voice equivalent of the dad from Cocomelon. Here's an approximate technical diagram of how these businesses work: Part 1 is what I like to call the "bilking process." Basically, you put up a flashy landing page promising content generation in exchange for a monthly subscription fee (or discounted annual fee, of course!). No more paying pesky writers! Once the husk of a company has secured the bag, part 2, the "bullshit process," kicks in. Customers provide their niches and the service happily passes queries over to the ChatGPT (or similar) API. Customers are rewarded with stinky garbage articles that sound like they're being narrated by HAL on Prozac in return. Success! I suppose we should have expected as much. With every new tech trend comes a deluge of tech investors trying to find the next great thing. And when this happens, it's a gold rush every time. I will say I'm more optimistic about "AI" (aka machine learning, aka statistics). There are going to be some pretty cool applications of this tech eventually—but your ChatGPT wrapper ain't it.
I have noticed a trend in a handful of products I've worked on at big tech companies. I have friends at other big tech companies that have noticed a similar trend: The products are kind of crummy. Here are some experiences that I have often encountered: the UI is flakey and/or unintuitive there is a lot of cruft in the codebase that has never been cleaned up bugs that have "acceptable" workarounds that never get fixed packages/dependencies are badly out of date the developer experience is crummy (bad build times, easily breakable processes) One of the reasons I have found for these issues is that we simply aren't investing enough time to increase product quality: we have poorly or nonexistent quality metrics, invest minimally in testing infrastructure (and actually writing tests), and don't invest in improving the inner loop. But why is this? My experience has been that quality is simply a hard sell in bigh tech. Let's first talk about something that's an easy sell right now: AI everything. Why is this an easy sell? Well, Microsoft could announce they put ChatGPT in a toaster and their stock price would jump $5/share. The sad truth is that big tech is hyper-focused on doing the things that make their stock prices go up in the short-term. It's hard to make this connection with quality initiatives. If your software is slightly less shitty, the stock price won't jump next week. So instead of being able to sell the obvious benefit of shiny new features, you need to have an Engineering Manager willing to risk having lower impact for the sake of having a better product. Even if there is broad consensus in your team, group, org that these quality improvements are necessary, there's a point up the corporate hierarchy where it simply doesn't matter to them. Certainly not as much as shipping some feature to great fanfare. Part of a bigger strategy? # Cory Doctorow has said some interesting things about enshittification in big tech: "enshittification is a three-stage process: first, surpluses are allocated to users until they are locked in. Then they are withdrawn and given to business-customers until they are locked in. Then all the value is harvested for the company's shareholders, leaving just enough residual value in the service to keep both end-users and business-customers glued to the platform." At a macro level, it's possible this is the strategy: hook users initially, make them dependent on your product, and then cram in superficial features that make the stock go up but don't offer real value, and keep the customers simply because they really have no choice but to use your product (an enterprise Office 365 customer probably isn't switching anytime soon). This does seem to have been a good strategy in the short-term: look at Microsoft's stock ever since they started cranking out AI everything. But how can the quality corner-cutting work long-term? I hope the hubris will backfire # Something will have to give. Big tech products can't just keep getting shittier—can they? I'd like to think some smaller competitors will come eat their lunch, but I'm not sure. Hopefully we're not all too entrenched in the big tech ecosystem for this to happen.
More in science
Image generators are designed to mimic their training data, so where does their apparent creativity come from? A recent study suggests that it’s an inevitable by-product of their architecture. The post Researchers Uncover Hidden Ingredients Behind AI Creativity first appeared on Quanta Magazine
I participated in a program about 15 years ago that looked at science and technology challenges faced by a subset of the US government. I came away thinking that such problems fall into three broad categories. Actual science and engineering challenges, which require foundational research and creativity to solve. Technology that may be fervently desired but is incompatible with the laws of nature, economic reality, or both. Alleged science and engineering problems that are really human/sociology issues. Part of science and engineering education and training is giving people the skills to recognize which problems belong to which categories. Confusing these can strongly shape the perception of whether science and engineering research is making progress. There has been a lot of discussion in the last few years about whether scientific progress (however that is measured) has slowed down or stagnated. For example, see here: https://www.theatlantic.com/science/archive/2018/11/diminishing-returns-science/575665/ https://news.uchicago.edu/scientific-progress-slowing-james-evans https://www.forbes.com/sites/roberthart/2023/01/04/where-are-all-the-scientific-breakthroughs-forget-ai-nuclear-fusion-and-mrna-vaccines-advances-in-science-and-tech-have-slowed-major-study-says/ https://theweek.com/science/world-losing-scientific-innovation-research A lot of the recent talk is prompted by this 2023 study, which argues that despite the world having many more researchers than ever before (behold population growth) and more global investment in research, somehow "disruptive" innovations are coming less often, or are fewer and farther between these days. (Whether this is an accurate assessment is not a simple matter to resolve; more on this below.) There is a whole tech bro culture that buys into this, however. For example, see this interview from last week in the New York Times with Peter Thiel, which points out that Thiel has been complaining about this for a decade and a half. On some level, I get it emotionally. The unbounded future spun in a lot of science fiction seems very far away. Where is my flying car? Where is my jet pack? Where is my moon base? Where are my fusion power plants, my antigravity machine, my tractor beams, my faster-than-light drive? Why does the world today somehow not seem that different than the world of 1985, while the world of 1985 seems very different than that of 1945? Some of the folks that buy into this think that science is deeply broken somehow - that we've screwed something up, because we are not getting the future they think we were "promised". Some of these people have this as an internal justification underpinning the dismantling of the NSF, the NIH, basically a huge swath of the research ecosystem in the US. These same people would likely say that I am part of the problem, and that I can't be objective about this because the whole research ecosystem as it currently exists is a groupthink self-reinforcing spiral of mediocrity. Science and engineering are inherently human ventures, and I think a lot of these concerns have an emotional component. My take at the moment is this: Genuinely transformational breakthroughs are rare. They often require a combination of novel insights, previously unavailable technological capabilities, and luck. They don't come on a schedule. There is no hard and fast rule that guarantees continuous exponential technological progress. Indeed, in real life, exponential growth regimes never last. The 19th and 20th centuries were special. If we think of research as a quest for understanding, it's inherently hierarchal. Civilizational collapses aside, you can only discover how electricity works once. You can only discover the germ theory of disease, the nature of the immune system, and vaccination once (though in the US we appear to be trying really hard to test that by forgetting everything). You can only discover quantum mechanics once, and doing so doesn't imply that there will be an ongoing (infinite?) chain of discoveries of similar magnitude. People are bad at accurately perceiving rare events and their consequences, just like people have a serious problem evaluating risk or telling the difference between correlation and causation. We can't always recognize breakthroughs when they happen. Sure, I don't have a flying car. I do have a device in my pocket that weighs only a few ounces, gives me near-instantaneous access to the sum total of human knowledge, let's me video call people around the world, can monitor aspects of my fitness, and makes it possible for me to watch sweet videos about dogs. The argument that we don't have transformative, enormously disruptive breakthroughs as often as we used to or as often as we "should" is in my view based quite a bit on perception. Personally, I think we still have a lot more to learn about the natural world. AI tools will undoubtedly be helpful in making progress in many areas, but I think it is definitely premature to argue that the vast majority of future advances will come from artificial superintelligences and thus we can go ahead and abandon the strategies that got us the remarkable achievements of the last few decades. I think some of the loudest complainers (Thiel, for example) about perceived slowing advancement are software people. People who come from the software development world don't always appreciate that physical infrastructure and understanding are hard, and that there are not always clever or even brute-force ways to get to an end goal. Solving foundational problems in molecular biology or quantum information hardware or photonics or materials is not the same as software development. (The tech folks generally know this on an intellectual level, but I don't think all of them really understand it in their guts. That's why so many of them seem to ignore real world physical constraints when talking about AI.). Trying to apply software development inspired approaches to science and engineering research isn't bad as a component of a many-pronged strategy, but alone it may not give the desired results - as warned in part by this piece in Science this week. More frequent breakthroughs in our understanding and capabilities would be wonderful. I don't think dynamiting the US research ecosystem is the way to get us there, and hoping that we can dismantle everything because AI will somehow herald a new golden age seems premature at best.
Humans are dramatically changing the environment of the Earth in many ways. Only about 23% of the land surface (excluding Antarctica) is considered to be “wilderness”, and this is rapidly decreasing. What wilderness is left is also mostly managed conservation areas. Meanwhile, about 3% of the surface is considered urban. I could not find a […] The post Animals Adapting to Cities first appeared on NeuroLogica Blog.
The basis for much of modern electronics is a set of silicon technologies called CMOS, which stands for complementary metal oxide semiconductor devices and processes. "Complementary" means using semiconductors (typically silicon) that is locally chemically doped so that you can have both n-type (carriers are negatively charged electrons in the conduction band) and p-type (carriers are positively charged holes in the valence band) material on the same substrate. With field-effect transistors (using oxide gate dielectrics), you can make very compact, comparatively low power devices like inverters and logic gates. There are multiple different approaches to try to implement quantum information processing in solid state platforms, with the idea that the scaling lessons of microelectronics (in terms of device density and reliability) can be applied. I think that essentially all of these avenues require cryogenic operating conditions; all superconducting qubits need ultracold conditions for both superconductivity and to minimize extraneous quasiparticles and other decoherence sources. Semiconductor-based quantum dots (Intel's favorite) similarly need thermal perturbations and decoherence to be minimized. The wealth of solid state quantum computing research is the driver for the historically enormous (to me, anyway) growth of dilution refrigerator manufacturing (see my last point here). So you eventually want to have thousands of error-corrected logical qubits at sub-Kelvin temperatures, which may involve millions of physical qubits at sub-Kelvin temperatures, all of which need to be controlled. Despite the absolute experimental fearlessness of people like John Martinis, you are not going to get this to work by running a million wires from room temperature into your dil fridge. Fig. 1 from here. The alternative people in this area have converged upon is to create serious CMOS control circuitry that can work at 4 K or below, so that a lot of the wiring does not need to go from the qubits all the way to room temperature. The materials and device engineering challenges in doing this are substantial! Power dissipation really needs to be minimized, and material properties to work at cryogenic conditions are not the same as those optimized for room temperature. There have been major advances in this - examples include Google in 2019, Intel in 2021, IBM in 2024, and this week, folks at the University of New South Wales supported by Microsoft. In this most recent work, the aspect that I find most impressive is that the CMOS electronics are essentially a serious logic-based control board operating at milliKelvin temperatures right next to the chip with the qubits (in this case, spins-in-quantum-dots). I'm rather blown away that this works and with sufficiently low power dissipation that the fridge is happy. This is very impressive, and there is likely a very serious future in store for cryogenic CMOS.
The Trump administration is outwardly hostile to clean energy sourced from solar and wind. But thanks to close ties to the fossil fuel industry and new technological breakthroughs, U.S. geothermal power may survive the GOP assaults on support for renewables and even thrive. Read more on E360 →
