Full Width 
Shortcuts 
Sign Up 
Log In 
Articles 
Reading List More from The Berkeley Artificial Intelligence Research Blog
PLAID is a multimodal generative model that simultaneously generates protein 1D sequence and 3D structure, by learning the latent space of protein folding models. The awarding of the 2024 Nobel Prize to AlphaFold2 marks an important moment of recognition for the of AI role in biology. What comes next after protein folding? In PLAID, we develop a method that learns to sample from the latent space of protein folding models to generate new proteins. It can accept compositional function and organism prompts, and can be trained on sequence databases, which are 2-4 orders of magnitude larger than structure databases. Unlike many previous protein structure generative models, PLAID addresses the multimodal co-generation problem setting: simultaneously generating both discrete sequence and continuous all-atom structural coordinates. From structure prediction to real-world drug design Though recent works demonstrate promise for the ability of diffusion models to generate proteins, there still exist limitations of previous models that make them impractical for real-world applications, such as: All-atom generation: Many existing generative models only produce the backbone atoms. To produce the all-atom structure and place the sidechain atoms, we need to know the sequence. This creates a multimodal generation problem that requires simultaneous generation of discrete and continuous modalities. Organism specificity: Proteins biologics intended for human use need to be humanized, to avoid being destroyed by the human immune system. Control specification: Drug discovery and putting it into the hands of patients is a complex process. How can we specify these complex constraints? For example, even after the biology is tackled, you might decide that tablets are easier to transport than vials, adding a new constraint on soluability. Generating “useful” proteins Simply generating proteins is not as useful as controlling the generation to get useful proteins. What might an interface for this look like? For inspiration, let's consider how we'd control image generation via compositional textual prompts (example from Liu et al., 2022). In PLAID, we mirror this interface for control specification. The ultimate goal is to control generation entirely via a textual interface, but here we consider compositional constraints for two axes as a proof-of-concept: function and organism: Learning the function-structure-sequence connection. PLAID learns the tetrahedral cysteine-Fe2+/Fe3+ coordination pattern often found in metalloproteins, while maintaining high sequence-level diversity. Training using sequence-only training data Another important aspect of the PLAID model is that we only require sequences to train the generative model! Generative models learn the data distribution defined by its training data, and sequence databases are considerably larger than structural ones, since sequences are much cheaper to obtain than experimental structure. Learning from a larger and broader database. The cost of obtaining protein sequences is much lower than experimentally characterizing structure, and sequence databases are 2-4 orders of magnitude larger than structural ones. How does it work? The reason that we’re able to train the generative model to generate structure by only using sequence data is by learning a diffusion model over the latent space of a protein folding model. Then, during inference, after sampling from this latent space of valid proteins, we can take frozen weights from the protein folding model to decode structure. Here, we use ESMFold, a successor to the AlphaFold2 model which replaces a retrieval step with a protein language model. Our method. During training, only sequences are needed to obtain the embedding; during inference, we can decode sequence and structure from the sampled embedding. ❄️ denotes frozen weights. In this way, we can use structural understanding information in the weights of pretrained protein folding models for the protein design task. This is analogous to how vision-language-action (VLA) models in robotics make use of priors contained in vision-language models (VLMs) trained on internet-scale data to supply perception and reasoning and understanding information. Compressing the latent space of protein folding models A small wrinkle with directly applying this method is that the latent space of ESMFold – indeed, the latent space of many transformer-based models – requires a lot of regularization. This space is also very large, so learning this embedding ends up mapping to high-resolution image synthesis. To address this, we also propose CHEAP (Compressed Hourglass Embedding Adaptations of Proteins), where we learn a compression model for the joint embedding of protein sequence and structure. Investigating the latent space. (A) When we visualize the mean value for each channel, some channels exhibit “massive activations”. (B) If we start examining the top-3 activations compared to the median value (gray), we find that this happens over many layers. (C) Massive activations have also been observed for other transformer-based models. We find that this latent space is actually highly compressible. By doing a bit of mechanistic interpretability to better understand the base model that we are working with, we were able to create an all-atom protein generative model. What’s next? Though we examine the case of protein sequence and structure generation in this work, we can adapt this method to perform multi-modal generation for any modalities where there is a predictor from a more abundant modality to a less abundant one. As sequence-to-structure predictors for proteins are beginning to tackle increasingly complex systems (e.g. AlphaFold3 is also able to predict proteins in complex with nucleic acids and molecular ligands), it’s easy to imagine performing multimodal generation over more complex systems using the same method. If you are interested in collaborating to extend our method, or to test our method in the wet-lab, please reach out! Further links If you’ve found our papers useful in your research, please consider using the following BibTeX for PLAID and CHEAP: @article{lu2024generating, title={Generating All-Atom Protein Structure from Sequence-Only Training Data}, author={Lu, Amy X and Yan, Wilson and Robinson, Sarah A and Yang, Kevin K and Gligorijevic, Vladimir and Cho, Kyunghyun and Bonneau, Richard and Abbeel, Pieter and Frey, Nathan}, journal={bioRxiv}, pages={2024--12}, year={2024}, publisher={Cold Spring Harbor Laboratory} } @article{lu2024tokenized, title={Tokenized and Continuous Embedding Compressions of Protein Sequence and Structure}, author={Lu, Amy X and Yan, Wilson and Yang, Kevin K and Gligorijevic, Vladimir and Cho, Kyunghyun and Abbeel, Pieter and Bonneau, Richard and Frey, Nathan}, journal={bioRxiv}, pages={2024--08}, year={2024}, publisher={Cold Spring Harbor Laboratory} } You can also checkout our preprints (PLAID, CHEAP) and codebases (PLAID, CHEAP). Some bonus protein generation fun! Additional function-prompted generations with PLAID. Transmembrane proteins have hydrophobic residues at the core, where it is embedded within the fatty acid layer. These are consistently observed when prompting PLAID with transmembrane protein keywords. Additional examples of active site recapitulation based on function keyword prompting. Comparing samples between PLAID and all-atom baselines. PLAID samples have better diversity and captures the beta-strand pattern that has been more difficult for protein generative models to learn. Acknowledgements Thanks to Nathan Frey for detailed feedback on this article, and to co-authors across BAIR, Genentech, Microsoft Research, and New York University: Wilson Yan, Sarah A. Robinson, Simon Kelow, Kevin K. Yang, Vladimir Gligorijevic, Kyunghyun Cho, Richard Bonneau, Pieter Abbeel, and Nathan C. Frey.
Training Diffusion Models with Reinforcement Learning We deployed 100 reinforcement learning (RL)-controlled cars into rush-hour highway traffic to smooth congestion and reduce fuel consumption for everyone. Our goal is to tackle "stop-and-go" waves, those frustrating slowdowns and speedups that usually have no clear cause but lead to congestion and significant energy waste. To train efficient flow-smoothing controllers, we built fast, data-driven simulations that RL agents interact with, learning to maximize energy efficiency while maintaining throughput and operating safely around human drivers. Overall, a small proportion of well-controlled autonomous vehicles (AVs) is enough to significantly improve traffic flow and fuel efficiency for all drivers on the road. Moreover, the trained controllers are designed to be deployable on most modern vehicles, operating in a decentralized manner and relying on standard radar sensors. In our latest paper, we explore the challenges of deploying RL controllers on a large-scale, from simulation to the field, during this 100-car experiment. The challenges of phantom jams A stop-and-go wave moving backwards through highway traffic. If you drive, you’ve surely experienced the frustration of stop-and-go waves, those seemingly inexplicable traffic slowdowns that appear out of nowhere and then suddenly clear up. These waves are often caused by small fluctuations in our driving behavior that get amplified through the flow of traffic. We naturally adjust our speed based on the vehicle in front of us. If the gap opens, we speed up to keep up. If they brake, we also slow down. But due to our nonzero reaction time, we might brake just a bit harder than the vehicle in front. The next driver behind us does the same, and this keeps amplifying. Over time, what started as an insignificant slowdown turns into a full stop further back in traffic. These waves move backward through the traffic stream, leading to significant drops in energy efficiency due to frequent accelerations, accompanied by increased CO2 emissions and accident risk. And this isn’t an isolated phenomenon! These waves are ubiquitous on busy roads when the traffic density exceeds a critical threshold. So how can we address this problem? Traditional approaches like ramp metering and variable speed limits attempt to manage traffic flow, but they often require costly infrastructure and centralized coordination. A more scalable approach is to use AVs, which can dynamically adjust their driving behavior in real-time. However, simply inserting AVs among human drivers isn’t enough: they must also drive in a smarter way that makes traffic better for everyone, which is where RL comes in. Fundamental diagram of traffic flow. The number of cars on the road (density) affects how much traffic is moving forward (flow). At low density, adding more cars increases flow because more vehicles can pass through. But beyond a critical threshold, cars start blocking each other, leading to congestion, where adding more cars actually slows down overall movement. Reinforcement learning for wave-smoothing AVs RL is a powerful control approach where an agent learns to maximize a reward signal through interactions with an environment. The agent collects experience through trial and error, learns from its mistakes, and improves over time. In our case, the environment is a mixed-autonomy traffic scenario, where AVs learn driving strategies to dampen stop-and-go waves and reduce fuel consumption for both themselves and nearby human-driven vehicles. Training these RL agents requires fast simulations with realistic traffic dynamics that can replicate highway stop-and-go behavior. To achieve this, we leveraged experimental data collected on Interstate 24 (I-24) near Nashville, Tennessee, and used it to build simulations where vehicles replay highway trajectories, creating unstable traffic that AVs driving behind them learn to smooth out. Simulation replaying a highway trajectory that exhibits several stop-and-go waves. We designed the AVs with deployment in mind, ensuring that they can operate using only basic sensor information about themselves and the vehicle in front. The observations consist of the AV’s speed, the speed of the leading vehicle, and the space gap between them. Given these inputs, the RL agent then prescribes either an instantaneous acceleration or a desired speed for the AV. The key advantage of using only these local measurements is that the RL controllers can be deployed on most modern vehicles in a decentralized way, without requiring additional infrastructure. Reward design The most challenging part is designing a reward function that, when maximized, aligns with the different objectives that we desire the AVs to achieve: Wave smoothing: Reduce stop-and-go oscillations. Energy efficiency: Lower fuel consumption for all vehicles, not just AVs. Safety: Ensure reasonable following distances and avoid abrupt braking. Driving comfort: Avoid aggressive accelerations and decelerations. Adherence to human driving norms: Ensure a “normal” driving behavior that doesn’t make surrounding drivers uncomfortable. Balancing these objectives together is difficult, as suitable coefficients for each term must be found. For instance, if minimizing fuel consumption dominates the reward, RL AVs learn to come to a stop in the middle of the highway because that is energy optimal. To prevent this, we introduced dynamic minimum and maximum gap thresholds to ensure safe and reasonable behavior while optimizing fuel efficiency. We also penalized the fuel consumption of human-driven vehicles behind the AV to discourage it from learning a selfish behavior that optimizes energy savings for the AV at the expense of surrounding traffic. Overall, we aim to strike a balance between energy savings and having a reasonable and safe driving behavior. Simulation results Illustration of the dynamic minimum and maximum gap thresholds, within which the AV can operate freely to smooth traffic as efficiently as possible. The typical behavior learned by the AVs is to maintain slightly larger gaps than human drivers, allowing them to absorb upcoming, possibly abrupt, traffic slowdowns more effectively. In simulation, this approach resulted in significant fuel savings of up to 20% across all road users in the most congested scenarios, with fewer than 5% of AVs on the road. And these AVs don’t have to be special vehicles! They can simply be standard consumer cars equipped with a smart adaptive cruise control (ACC), which is what we tested at scale. Smoothing behavior of RL AVs. Red: a human trajectory from the dataset. Blue: successive AVs in the platoon, where AV 1 is the closest behind the human trajectory. There is typically between 20 and 25 human vehicles between AVs. Each AV doesn’t slow down as much or accelerate as fast as its leader, leading to decreasing wave amplitude over time and thus energy savings. 100 AV field test: deploying RL at scale Our 100 cars parked at our operational center during the experiment week. Given the promising simulation results, the natural next step was to bridge the gap from simulation to the highway. We took the trained RL controllers and deployed them on 100 vehicles on the I-24 during peak traffic hours over several days. This large-scale experiment, which we called the MegaVanderTest, is the largest mixed-autonomy traffic-smoothing experiment ever conducted. Before deploying RL controllers in the field, we trained and evaluated them extensively in simulation and validated them on the hardware. Overall, the steps towards deployment involved: Training in data-driven simulations: We used highway traffic data from I-24 to create a training environment with realistic wave dynamics, then validate the trained agent’s performance and robustness in a variety of new traffic scenarios. Deployment on hardware: After being validated in robotics software, the trained controller is uploaded onto the car and is able to control the set speed of the vehicle. We operate through the vehicle’s on-board cruise control, which acts as a lower-level safety controller. Modular control framework: One key challenge during the test was not having access to the leading vehicle information sensors. To overcome this, the RL controller was integrated into a hierarchical system, the MegaController, which combines a speed planner guide that accounts for downstream traffic conditions, with the RL controller as the final decision maker. Validation on hardware: The RL agents were designed to operate in an environment where most vehicles were human-driven, requiring robust policies that adapt to unpredictable behavior. We verify this by driving the RL-controlled vehicles on the road under careful human supervision, making changes to the control based on feedback. Each of the 100 cars is connected to a Raspberry Pi, on which the RL controller (a small neural network) is deployed. The RL controller directly controls the onboard adaptive cruise control (ACC) system, setting its speed and desired following distance. Once validated, the RL controllers were deployed on 100 cars and driven on I-24 during morning rush hour. Surrounding traffic was unaware of the experiment, ensuring unbiased driver behavior. Data was collected during the experiment from dozens of overhead cameras placed along the highway, which led to the extraction of millions of individual vehicle trajectories through a computer vision pipeline. Metrics computed on these trajectories indicate a trend of reduced fuel consumption around AVs, as expected from simulation results and previous smaller validation deployments. For instance, we can observe that the closer people are driving behind our AVs, the less fuel they appear to consume on average (which is calculated using a calibrated energy model): Average fuel consumption as a function of distance behind the nearest engaged RL-controlled AV in the downstream traffic. As human drivers get further away behind AVs, their average fuel consumption increases. Another way to measure the impact is to measure the variance of the speeds and accelerations: the lower the variance, the less amplitude the waves should have, which is what we observe from the field test data. Overall, although getting precise measurements from a large amount of camera video data is complicated, we observe a trend of 15 to 20% of energy savings around our controlled cars. Data points from all vehicles on the highway over a single day of the experiment, plotted in speed-acceleration space. The cluster to the left of the red line represents congestion, while the one on the right corresponds to free flow. We observe that the congestion cluster is smaller when AVs are present, as measured by computing the area of a soft convex envelope or by fitting a Gaussian kernel. Final thoughts The 100-car field operational test was decentralized, with no explicit cooperation or communication between AVs, reflective of current autonomy deployment, and bringing us one step closer to smoother, more energy-efficient highways. Yet, there is still vast potential for improvement. Scaling up simulations to be faster and more accurate with better human-driving models is crucial for bridging the simulation-to-reality gap. Equipping AVs with additional traffic data, whether through advanced sensors or centralized planning, could further improve the performance of the controllers. For instance, while multi-agent RL is promising for improving cooperative control strategies, it remains an open question how enabling explicit communication between AVs over 5G networks could further improve stability and further mitigate stop-and-go waves. Crucially, our controllers integrate seamlessly with existing adaptive cruise control (ACC) systems, making field deployment feasible at scale. The more vehicles equipped with smart traffic-smoothing control, the fewer waves we’ll see on our roads, meaning less pollution and fuel savings for everyone! Many contributors took part in making the MegaVanderTest happen! The full list is available on the CIRCLES project page, along with more details about the project. Read more: [paper]
Anthology, a method for conditioning LLMs to representative, consistent, and diverse virtual personas by generating and utilizing naturalistic backstories with rich details of individual values and experience. --> We introduce Anthology, a method for conditioning LLMs to representative, consistent, and diverse virtual personas by generating and utilizing naturalistic backstories with rich details of individual values and experience. What does it mean for large language models (LLMs) to be trained on massive text corpora, collectively produced by millions and billions of distinctive human authors? In “Language Models as Agent Models”, compelling evidence suggests that recent language models could be considered models of agents: provided with a textual context, LLMs are capable of generating conditional text that represents the characteristics of an agent likely to have produced that context. This suggests that, with appropriate conditioning, LLMs could be guided to approximate the responses of a particular human voice, rather than the mixture of voices that otherwise emerges. If realized, this capability of LLMs would have significant implications for user research and social sciences—conditioned language models as virtual personas of human subjects could serve as cost-effective pilot studies and supporting best practices in human studies, e.g. the Belmont principles of justice and beneficence. In this work, we introduce Anthology, an approach for steering LLMs to representative, consistent, and diverse virtual personas by providing richly detailed life narratives of individuals as conditioning context to models. In doing so, we also present methods to generate backstories from LLMs themselves as a means to efficiently produce massive sets covering a wide range of human demographics. By grounding language models in naturalistic backstories, Anthology allows LLMs to simulate individual human samples with increased fidelity, measured in terms of matching the distributions and consistencies of human responses. Our Approach: Anthology Conditioning Language Model Generation with Individual Life Narratives A significant limitation of earlier methods in steering LLMs to virtual personas has been the inability to reliably approximate individual human samples. Prior approaches prompt LLMs with broad demographic information, e.g., “I am a 25-year-old from California. My highest level of education is less than high school,” which are essentially bodies of text generated from a tuple of demographic variables. With these methods, we are only able to approximate human samples at a population level, not at the individual level, which results in: Responses prone to LLMs defaulting to stereotypical and/or prototypical portrayals, as they are only conditioned on demographic variables (e.g., race and gender) Inability to provide important metrics of interest such as covariance and statistical significance, as individual responses are required for such compuatations Anthology enables the approximation of individual subjects by conditioning with richly detailed backstories. Through these backstories, the model captures implicit and explicit markers of personal identity, including demographic traits and spontaneous references to cultural, socioeconomic backgrounds, and life philosophies. Our approach involves generating a vast set of backstories representing a wide range of demographic attributes via language models queried with unrestricted, open-ended prompts such as, “Tell me about yourself.” We then match virtual personas conditioned by each backstory to real-world survey samples. Results: Closer Approximation of Public Opinion Polls For evaluation, we compare the effectiveness of different methods for conditioning virtual personas in the context of approximating three Pew Research Center ATP surveys: Waves 34, 92, and 99. Results on approximating human responses for Pew Research Center ATP surveys. Boldface and underlined results indicate values closest and the second closest to those of humans, respectively. As measures of success in approximating human samples with virtual personas, we consider the following metrics: Average Wasserstein distance (WD) between response distributions as a measure of representativeness Frobenius norm (Fro.) between correlation matrices as a measure of consistency Cronbach’s alpha as an additional measure of internal consistency Prior to analyzing virtual subjects, we estimate the lower bounds of each evaluation metric by repeatedly dividing the human population into two equal-sized groups at random and calculating these metrics between the subgroups. We take averaged values from 100 iterations to represent the lower-bound estimates. We consistently observe that Anthology outperforms other conditioning methods with respect to all metrics, for both the Llama-3-70B and the Mixtral-8x22B. When comparing two matching methods, the greedy matching method tends to show better performance on the average Wasserstein distance across all Waves. We attribute differences in matching methods to the one-to-one correspondence condition of maximum weight matching and the limited number of virtual users available. Specifically, the weights assigned to matched virtual subjects in maximum weight matching are inevitably lower than those in greedy matching, as the latter relaxes the constraints on one-to-one correspondence. This discrepancy can result in a lower demographic similarity between matched human and virtual users compared to the counterpart from greedy matching. These results suggest that the richness of the generated backstories in our approach elicits more nuanced responses compared to baselines. Final Thoughts Anthology marks a promising new direction in conditioning virtual personas in LLMs that could potentially reshape how we conduct user research, public opinion surveys, and other social science applications by offering a scalable, and at times, ethical alternative to traditional human surveys. However, the use of Anthology, as in any other application of language models in the social sciences, also brings several considerations to the forefront: although the generated backstories help create more representative personas, there remains a risk of perpetuating biases or infringing on privacy, so results should be used and interpreted with caution. In terms of future steps, we envision our approach benefiting from a more expansive and diverse set of backstories, each representing a consistent life narrative of individuals. Additionally, a valuable extension of the work would be to consider free-form response generation, enabling more natural and nuanced persona simulations beyond structured survey formats such as multiple-choice. Finally, an exciting next dimension in applying LLMs in behavioral studies would involve simulating longer-term effects, allowing virtual personas to model and retrospectively examine changes over time. All of these directions present multitudes of technical challenges; please let us know if you are interested in collaborating or want to discuss our work further! Learn more about our work: link to full paper @article{moon2024virtual, title={Virtual personas for language models via an anthology of backstories}, author={Moon, Suhong and Abdulhai, Marwa and Kang, Minwoo and Suh, Joseph and Soedarmadji, Widyadewi and Behar, Eran Kohen and Chan, David M}, journal={arXiv preprint arXiv:2407.06576}, year={2024} }
Sample language model responses to different varieties of English and native speaker reactions. ChatGPT does amazingly well at communicating with people in English. But whose English? Only 15% of ChatGPT users are from the US, where Standard American English is the default. But the model is also commonly used in countries and communities where people speak other varieties of English. Over 1 billion people around the world speak varieties such as Indian English, Nigerian English, Irish English, and African-American English. Speakers of these non-“standard” varieties often face discrimination in the real world. They’ve been told that the way they speak is unprofessional or incorrect, discredited as witnesses, and denied housing–despite extensive research indicating that all language varieties are equally complex and legitimate. Discriminating against the way someone speaks is often a proxy for discriminating against their race, ethnicity, or nationality. What if ChatGPT exacerbates this discrimination? To answer this question, our recent paper examines how ChatGPT’s behavior changes in response to text in different varieties of English. We found that ChatGPT responses exhibit consistent and pervasive biases against non-“standard” varieties, including increased stereotyping and demeaning content, poorer comprehension, and condescending responses. Our Study We prompted both GPT-3.5 Turbo and GPT-4 with text in ten varieties of English: two “standard” varieties, Standard American English (SAE) and Standard British English (SBE); and eight non-“standard” varieties, African-American, Indian, Irish, Jamaican, Kenyan, Nigerian, Scottish, and Singaporean English. Then, we compared the language model responses to the “standard” varieties and the non-“standard” varieties. First, we wanted to know whether linguistic features of a variety that are present in the prompt would be retained in GPT-3.5 Turbo responses to that prompt. We annotated the prompts and model responses for linguistic features of each variety and whether they used American or British spelling (e.g., “colour” or “practise”). This helps us understand when ChatGPT imitates or doesn’t imitate a variety, and what factors might influence the degree of imitation. Then, we had native speakers of each of the varieties rate model responses for different qualities, both positive (like warmth, comprehension, and naturalness) and negative (like stereotyping, demeaning content, or condescension). Here, we included the original GPT-3.5 responses, plus responses from GPT-3.5 and GPT-4 where the models were told to imitate the style of the input. Results We expected ChatGPT to produce Standard American English by default: the model was developed in the US, and Standard American English is likely the best-represented variety in its training data. We indeed found that model responses retain features of SAE far more than any non-“standard” dialect (by a margin of over 60%). But surprisingly, the model does imitate other varieties of English, though not consistently. In fact, it imitates varieties with more speakers (such as Nigerian and Indian English) more often than varieties with fewer speakers (such as Jamaican English). That suggests that the training data composition influences responses to non-“standard” dialects. ChatGPT also defaults to American conventions in ways that could frustrate non-American users. For example, model responses to inputs with British spelling (the default in most non-US countries) almost universally revert to American spelling. That’s a substantial fraction of ChatGPT’s userbase likely hindered by ChatGPT’s refusal to accommodate local writing conventions. Model responses are consistently biased against non-“standard” varieties. Default GPT-3.5 responses to non-“standard” varieties consistently exhibit a range of issues: stereotyping (19% worse than for “standard” varieties), demeaning content (25% worse), lack of comprehension (9% worse), and condescending responses (15% worse). Native speaker ratings of model responses. Responses to non-”standard” varieties (blue) were rated as worse than responses to “standard” varieties (orange) in terms of stereotyping (19% worse), demeaning content (25% worse), comprehension (9% worse), naturalness (8% worse), and condescension (15% worse). When GPT-3.5 is prompted to imitate the input dialect, the responses exacerbate stereotyping content (9% worse) and lack of comprehension (6% worse). GPT-4 is a newer, more powerful model than GPT-3.5, so we’d hope that it would improve over GPT-3.5. But although GPT-4 responses imitating the input improve on GPT-3.5 in terms of warmth, comprehension, and friendliness, they exacerbate stereotyping (14% worse than GPT-3.5 for minoritized varieties). That suggests that larger, newer models don’t automatically solve dialect discrimination: in fact, they might make it worse. Implications ChatGPT can perpetuate linguistic discrimination toward speakers of non-“standard” varieties. If these users have trouble getting ChatGPT to understand them, it’s harder for them to use these tools. That can reinforce barriers against speakers of non-“standard” varieties as AI models become increasingly used in daily life. Moreover, stereotyping and demeaning responses perpetuate ideas that speakers of non-“standard” varieties speak less correctly and are less deserving of respect. As language model usage increases globally, these tools risk reinforcing power dynamics and amplifying inequalities that harm minoritized language communities. Learn more here: [ paper ]
More in AI
Today's links Goodhart's Law (of AI): When a metric becomes a target, AI can hit it every time. Hey look at this: Delights to delectate. Object permanence: Bill Ayers graphic novel; Foxconn in India; Uber loses $4B; Warren Buffet, monopolist. Upcoming appearances: Where to find me. Recent appearances: Where I've been. Latest books: You keep readin' em, I'll keep writin' 'em. Upcoming books: Like I said, I'll keep writin' 'em. Colophon: All the rest. Goodhart's Law (of AI) (permalink) One way to think about AI's unwelcome intrusion into our lives can be summed up with Goodhardt's Law: "When a measure becomes a target, it ceases to be a good measure": https://en.wikipedia.org/wiki/Goodhart%27s_law Goodhart's Law is a harsh mistress. It's incredibly exciting to discover a new way of measuring aspects of a complex system in a way that lets you understand (and thus control) it. In 1998, Sergey Brin and Larry Page realized that all the links created by everyone who'd ever made a webpage represented a kind of latent map of the value and authority of every website. We could infer that pages that had more links pointing to them were considered more noteworthy than pages that had fewer inbound links. Moreover, we could treat those heavily linked-to pages as authoritative and infer that when they linked to another page, it, too, was likely to be important. This insight, called "PageRank," was behind Google's stunning entry into the search market, which was easily one of the most exciting technological developments of the decade, as the entire web just snapped into place as a useful system for retrieving information that had been created by a vast, uncoordinated army of web-writers, hosted in a distributed system without any central controls. Then came the revenge of Goodhart's Law. Before Google became the dominant mechanism for locating webpages, the only reason for anyone to link to a given page or site was because there was something there they thought you should see. Google aggregated all those "I think you should see this" signals and turned them into a map of the web's relevance and authority. But making a link to a webpage is easy. Once there was another reason to make a link between two web-pages – to garner traffic, which could be converted into money and/or influence – then bad actors made a lot of spurious links between websites. They created linkfarms, they spammed blog comments, they hacked websites for the sole purpose of adding a bunch of human-invisible, Google-scraper-readable links to pages. The metric ("how many links are there to this page?") became a target ("make links to this page") and ceased to be a useful metric. Goodhart's Law is still a plague on Google search quality. "Reputation abuse" is a webcrime committed by venerable sites like Forbes, Fortune and Better Homes and Gardens, who abuse the authority imparted by tons of inbound links accumulated over decades by creating spammy, fake product-review sites stuffed with affiliate links, that Google ranks more highly than real, rigorous review sites because of all that accumulated googlejuice: https://pluralistic.net/2024/05/03/keyword-swarming/#site-reputation-abuse Goodhart's Law is 50 years old, but policymakers are woefully ignorant of it and continue to operate as though it doesn't apply to them. This is especially pronounced when policymakers are determined to Do Something about a public service that has been starved of funding kicked around as a political football to the point where it has degraded and started to outrage the public. When this happens, policymakers are apt to blame public servants – rather than themselves – for this degradation, and then set out to Bring Accountability to those public employees. The NHS did this with ambulance response times, which are very bad, and that fact is, in turn, very bad. The reason ambulance response times suck isn't hard to winkle out: there's not enough money being spent on ambulances, drivers, and medics. But that's not a politically popular conclusion, especially in the UK, which has been under brutal and worsening austerity since the Blair years (don't worry, eventually they'll do enough austerity and things will really turn around, because, as the old saying goes, "Good policymaking consists of doing the same thing over and over and expecting a different outcome)." Instead of blaming inadequate funding for poor ambulance response times, politicians blamed "inefficiency," driven by a poor motivation. So they established a metric: ambulances must arrive within a certain number of minutes (and they set a consequence: massive cuts to any ambulance service that didn't meet the metric). Now, "an ambulance where it's needed within a set amount of time" may sound like a straightforward metric, and it was – retrospectively. As in, we could tell that the ambulance service was in trouble because ambulances were taking half an hour or more to arrive. But prospectively, after that metric became a target, it immediately ceased to be a good metric. That's because ambulance services, faced with the impossible task of improving response times without spending money, started to dispatch ambulance motorbikes that couldn't carry 95% of the stuff needed to respond to a medical emergency, and had no way to get patients back to hospitals. These motorbikes were able to meet the response-time targets…without improving the survival rates of people who summoned ambulances: https://timharford.com/2014/07/underperforming-on-performance/ AI turns out to be a great way to explore all the perverse dimensions of Goodhart's Law. For years, machine learning specialists have struggled with the problem of "reward hacking," in which an AI figures out how to meet some target in a way that blows up the metric it was derived from: https://research.google/blog/bringing-precision-to-the-ai-safety-discussion/ My favorite example of this is the AI-powered Roomba that was programmed to find an efficient path that minimized collisions with furniture, as measured by a forward-facing sensor that sent a signal whenever the Roomba bumped into anything. The Roomba started driving backwards, smashing into all kinds of furniture, but measuring zero collisions, because there was no collision-sensor on its back: https://x.com/smingleigh/status/1060325665671692288 Charlie Stross has observed that corporations are a kind of "slow AI," that engage in endless reward-hacking to accomplish their goals, increasing their profits by finding nominally legal ways to poison the air, cheat their customers and maim their workers: https://memex.craphound.com/2017/12/29/charlie-strosss-ccc-talk-the-future-of-psychotic-ais-can-be-read-in-todays-sociopathic-corporations/ Public services under conditions of austerity are another kind of slow AI. When policymakers demand that a metric be satisfied without delivering any of the budget or resources needed to satisfy it, the public employees downstream of that impossible demand will start reward-hacking and the metric will become a target, and then cease to be a useful metric. Which brings me, at last, to AI in educational contexts. In 2008, George W Bush stepped up the long-running war on education with the No Child Left Behind Act. The right hates public education, for many reasons. Obviously, there's the fact that uneducated people are easier to mislead, which is helpful if you want to get a bunch of turkeys to vote for Christmas ("I love the uneducated" -DJ Trump). Then there's the fact that, since 1954's Brown v Board of Ed, Black and brown kids were legally guaranteed the right to be educated alongside white kids, which makes a large swathe of the right absolutely nuts. Then there was the 1962 Supreme Court decisions that banned prayer in school, leading to bans on teaching Christian doctrine, including nonsense like Young Earth Creationism. Finally, there's the fact that teachers a) belong to unions; and, b) believe in their jobs and fight for the kids they teach. No Child Left Behind was a vicious salvo in the war on teachers, positing the problem with education as a failure of teachers, driven by a combination of poor training and indifference to their students. Under No Child Left Behind, students were subjected to multiple rounds of standardized tests, and teachers with low-performing students had their budgets taken away (after first being offered modest assistance in improving those scores). Some of NCLB's standardized tests represented reasonable metrics: we really do want kids to be able to read and do math and reason and string together coherent thoughts at various points in their schooling. But when these metrics became targets, boy did they stop being useful as metrics. It's impossible to overstate how fucking perverse NCLB was. I once met an elementary school teacher from an incredibly poor school district in Kansas. Many of her students were resettled refugees who didn't speak English; they spoke a language that no one in the school system could speak, and which had no system of writing. They arrived in her classroom unable to speak English and unable to read or write in any language, and no one could speak their language. Obviously, these students performed badly on standardized tests delivered in English (it didn't help that they had to take the tests just months after arriving in the classroom, because the clock started ticking on their first test when they entered the system, which could take half a year to place them in a class). Within a couple years, these schools had had most of their budgets taken away. When the standardized tests rolled around, this teacher would lead her students into the only room in the school with computers – the test taking room. For many of these students, this was the first time they had ever used a computer. She would tell them to do their best and leave the room for an hour, while a well-paid proctor (along with test-taking computers, the only thing NCLB guaranteed funding for) observed them as they tried to figure out how a mouse worked. They would all score zero on the test, and the school would be punished. NCLB was such a failure that it was eventually rescinded (in 2015), but by that time, a new system of standardization had rushed in to fill the gap, the Common Core. Common Core is a set of rigid standardized curriciula – with standardized assessment rubrics – that was, once again, driven by contempt for teachers. The argument for Common Core was that students were failing – not because of falling budgets or No Child Left Behind – but because the unions were "protecting bad teachers," who would then go on to fail students. By taking away discretion from teachers, we could impose "accountability" on them. The absolutely predictable outcome followed Goodhart's Law to a tee: teachers prioritized inculcating students with the skills to pass the standardized tests, and when those test-taking skills crowded out actual learning, learning fell by the wayside. This continues up to the most advanced part of public education, the Advanced Placement courses that students aspiring to college are strongly pressured to take. If Common Core is rigid, AP is brittle to the point of shattering. Anyone who's ever parented a kid through the US secondary school system knows how much time their kids spent learning to hit their marks on standardized assessments, to the exclusion of actual learning, and how soul-suckingly awful this is. Take that staple of the AP assessment rubric: the five-paragraph essay (5PE), bane of students, teachers and parents everywhere: https://www.insidehighered.com/blogs/just-visiting/kill-5-paragraph-essay Speaking as a sometime writing teacher and an internationally bestselling essayist, 5PEs are objectively very bad essays. Their only virtue is that they can be assessed in a totally standard way, so the grade any given 5PE is awarded by any grader is likely to be the same grade it receives when presented to any other grader. Grading an essay is an irreducibly subjective matter, and the only way to create an objective standard for essays is to make the essays unrecognizable as essays. And yet, the 5PE is the heart of assessment for many AP classes, from History to English to Social Studies and beyond. A kid who scores high on any humanities APs will have put endless hours into perfecting this perfectly abominable literary form, mastering a skill that they will never, ever be called upon to use (the top piece of college entrance advice is "don't write your personal essay as a 5PE" and college professors spend the first half of their 101 classes teaching students not to turn in 5PEs). The same goes for many other aspects of AP and Common Core assessment. If you do AP Lit, you'll be required to annotate the literature you read by making a set number of marginal observations on every page of the novels, poems and essays you read. Again, as a literary reviewer, novelist, and nonfiction writer who's written more than 30 books, I have to say, this is a batshit way to learn to analyze and criticize literature. Its sole virtue is that it reduces the qualitative matter of literary analysis to a qualitative target that students can hit and teachers can count. And that's where AI comes in. AI – the ultimate bullshit machine – can produce a better 5PE than any student can, because the point of the 5PE isn't to be intellectually curious or rigorous, it's to produce a standardized output that can be analyzed using a standardized rubric. I've been writing YA novels and doing school visits for long enough to cement my understanding that kids are actually pretty darned clever. They don't graduate from high school thinking that their mastery of the 5PE is in any way good or useful, or that they're learning about literature by making five marginal observations per page when they read a book. Given all this, why wouldn't you ask an AI to do your homework? That homework is already the revenge of Goodhart's Law, a target that has ruined its metric. Your homework performance says nothing useful about your mastery of the subject, so why not let the AI write it. Hell, if you're a smart, motivated kid, then letting the AI write your bullshit 5PEs might give you time to write something good. Teachers aren't to blame here. They have to teach to the test, or they will fail their students (literally, because they will have to assign a failing grade to them, and figuratively, because a student who gets a failing grade will face all kinds of punishments). Teachers' unions – who consistently fight against standardization and in favor of their members discretion to practice their educational skills based on kids' individual needs – are the best hope we have: https://pluralistic.net/2025/03/29/jane-mcalevey/#trump-is-a-scab The right hates teachers and keeps on setting them up to fail. That hatred has no bottom. Take the Republican Texas State Rep Ryan Guillen, whose House Bill 462 will increase the state's school safety budget from $10/student to $100/student, with those additional funds earmarked to buy one armed drone per 200 students (these drones are supplied by a single company that has ties to Guillen): https://dronelife.com/2024/12/08/texas-lawmaker-proposes-drones-for-school-security-a-less-lethal-solution/ Imagine how much Texas schools could do with an extra $90/student/year – how much more usefully that money could be spent if it were turned over to teachers. But instead, Rep Guillen wants to put "AI in schools" in the form of drones equipped with pepper-spray, flash bangs, and "lances" that can be smashed into people at 100mph. The problem with AI in schools isn't that students are using AI to do their homework. It's that schools have been turned into reward-hacking AIs by a system that hates the idea of an educated populace almost as much as it hates the idea of unionized teachers who are empowered to teach our kids. (Image: Cryteria, CC BY 3.0; Lee Haywood, CC BY-SA 2.0; modified) Hey look at this (permalink) Cybertruck Leads Tesla’s Used-Car Collapse https://gizmodo.com/cybertruck-leads-teslas-used-car-collapse-2000641133 Hackers Went Looking for a Backdoor in High-Security Safes—and Now Can Open Them in Seconds https://www.wired.com/story/securam-prologic-safe-lock-backdoor-exploits/ I clustered four Framework Mainboards to test huge LLMs https://www.jeffgeerling.com/blog/2025/i-clustered-four-framework-mainboards-test-huge-llms The Framework Desktop is a beast https://world.hey.com/dhh/the-framework-desktop-is-a-beast-636fb4ff Leaving MAGA https://leavingmaga.org/they-left-maga/steve-vilchez/ Object permanence (permalink) #15yrsago Bill Ayers’s To Teach: The Journey, in Comics, a humanist look at education https://memex.craphound.com/2010/08/10/bill-ayerss-to-teach-the-journey-in-comics-a-humanist-look-at-education/ #10yrsago Kansas officials stonewall mathematician investigating voting machine “sabotage” https://www.kansas.com/news/politics-government/article27951310.html #10yrsago Chinese mega-manufacturers set up factories in India https://web.archive.org/web/20150811043714/https://www.itworld.com/article/2968375/android/foxconn-to-invest-5b-to-set-up-first-of-up-to-12-factories-in-india.html #10yrsago Oracle’s CSO demands an end to customers checking Oracle products for defects https://arstechnica.com/information-technology/2015/08/oracle-security-chief-to-customers-stop-checking-our-code-for-vulnerabilities/ #10yrsago Girl Sex 101: “for EVERYone who wants to bone down with chicks, regardless of your gender/orientation.” https://www.ohjoysextoy.com/girlsex-101/ #10yrsago John Oliver on the brutal state of sex-ed in America https://www.youtube.com/watch?v=L0jQz6jqQS0 #10yrsago Insurance monitoring dashboard devices used by Uber let hackers “cut your brakes” over wireless https://www.wired.com/2015/08/hackers-cut-corvettes-brakes-via-common-car-gadget/ #10yrsago US lobbying for TPP to lock up clinical trial data https://theconversation.com/how-the-battle-over-biologics-helped-stall-the-trans-pacific-partnership-45648 #10yrsago Larry Lessig considers running for the Democratic presidential nomination https://www.youtube.com/watch?v=CaqrQz71bMk #10yrsago Felicia Day’s “You’re Never Weird on the Internet (Almost)” https://memex.craphound.com/2015/08/11/felicia-days-youre-never-weird-on-the-internet-almost/ #10yrsago Overshare: Justin Hall’s biopic about the first social media/blogging https://www.youtube.com/watch?v=AxD4mqFtySQ #5yrsago When you hear "intangibles"… https://pluralistic.net/2020/08/11/nor-glom-of-nit/#capitalists-hate-competition #5yrsago How they're killing the post office https://pluralistic.net/2020/08/11/nor-glom-of-nit/#sos-usps #5yrsago Terra Nullius https://pluralistic.net/2020/08/11/nor-glom-of-nit/#terra-nullius #5yrsago Uber lost $4b in H1/2020 https://pluralistic.net/2020/08/10/folksy-monopolists/#bezzled #5yrsago Warren Buffet, monopolist https://pluralistic.net/2020/08/10/folksy-monopolists/#folksy-monopolists Upcoming appearances (permalink) Ithaca: AD White keynote (Cornell), Sep 12 https://deanoffaculty.cornell.edu/events/keynote-cory-doctorow-professor-at-large/ DC: Enshittification at Politics and Prose, Oct 8 https://politics-prose.com/cory-doctorow-10825 New Orleans: DeepSouthCon63, Oct 10-12, 2025 http://www.contraflowscifi.org/ San Francisco: Enshittification at Public Works (The Booksmith), Oct 20 https://app.gopassage.com/events/doctorow25 Miami: Enshittification at Books & Books, Nov 5 https://www.eventbrite.com/e/an-evening-with-cory-doctorow-tickets-1504647263469 Recent appearances (permalink) Tariffs vs IP Law (Firewalls Don't Stop Dragons) https://www.youtube.com/watch?v=LFABFe-5-uQ ORG at 20: In conversation with Maria Farrell https://www.youtube.com/watch?v=M9H2An_D6io Why aren't we controlling our own tech? (Co-Op Congress) https://www.youtube.com/live/GLrDwHgeCy4?si=NUWxPphk0FS_3g9J&t=4409 Latest books (permalink) Picks and Shovels: a sequel to "Red Team Blues," about the heroic era of the PC, Tor Books (US), Head of Zeus (UK), February 2025 (https://us.macmillan.com/books/9781250865908/picksandshovels). The Bezzle: a sequel to "Red Team Blues," about prison-tech and other grifts, Tor Books (US), Head of Zeus (UK), February 2024 (the-bezzle.org). "The Lost Cause:" a solarpunk novel of hope in the climate emergency, Tor Books (US), Head of Zeus (UK), November 2023 (http://lost-cause.org). "The Internet Con": A nonfiction book about interoperability and Big Tech (Verso) September 2023 (http://seizethemeansofcomputation.org). Signed copies at Book Soup (https://www.booksoup.com/book/9781804291245). "Red Team Blues": "A grabby, compulsive thriller that will leave you knowing more about how the world works than you did before." Tor Books http://redteamblues.com. "Chokepoint Capitalism: How to Beat Big Tech, Tame Big Content, and Get Artists Paid, with Rebecca Giblin", on how to unrig the markets for creative labor, Beacon Press/Scribe 2022 https://chokepointcapitalism.com Upcoming books (permalink) Canny Valley: A limited edition collection of the collages I create for Pluralistic, self-published, September 2025 Enshittification: Why Everything Suddenly Got Worse and What to Do About It, Farrar, Straus, Giroux, October 7 2025 https://us.macmillan.com/books/9780374619329/enshittification/ Unauthorized Bread: a middle-grades graphic novel adapted from my novella about refugees, toasters and DRM, FirstSecond, 2026 Enshittification, Why Everything Suddenly Got Worse and What to Do About It (the graphic novel), Firstsecond, 2026 The Memex Method, Farrar, Straus, Giroux, 2026 The Reverse-Centaur's Guide to AI, a short book about being a better AI critic, Farrar, Straus and Giroux, 2026 Colophon (permalink) Today's top sources: Currently writing: "The Reverse Centaur's Guide to AI," a short book for Farrar, Straus and Giroux about being an effective AI critic. (1076 words yesterday, 27803 words total). A Little Brother short story about DIY insulin PLANNING This work – excluding any serialized fiction – is licensed under a Creative Commons Attribution 4.0 license. That means you can use it any way you like, including commercially, provided that you attribute it to me, Cory Doctorow, and include a link to pluralistic.net. https://creativecommons.org/licenses/by/4.0/ Quotations and images are not included in this license; they are included either under a limitation or exception to copyright, or on the basis of a separate license. Please exercise caution. How to get Pluralistic: Blog (no ads, tracking, or data-collection): Pluralistic.net Newsletter (no ads, tracking, or data-collection): https://pluralistic.net/plura-list Mastodon (no ads, tracking, or data-collection): https://mamot.fr/@pluralistic Medium (no ads, paywalled): https://doctorow.medium.com/ Twitter (mass-scale, unrestricted, third-party surveillance and advertising): https://twitter.com/doctorow Tumblr (mass-scale, unrestricted, third-party surveillance and advertising): https://mostlysignssomeportents.tumblr.com/tagged/pluralistic "When life gives you SARS, you make sarsaparilla" -Joey "Accordion Guy" DeVilla READ CAREFULLY: By reading this, you agree, on behalf of your employer, to release me from all obligations and waivers arising from any and all NON-NEGOTIATED agreements, licenses, terms-of-service, shrinkwrap, clickwrap, browsewrap, confidentiality, non-disclosure, non-compete and acceptable use policies ("BOGUS AGREEMENTS") that I have entered into with your employer, its partners, licensors, agents and assigns, in perpetuity, without prejudice to my ongoing rights and privileges. You further represent that you have the authority to release me from any BOGUS AGREEMENTS on behalf of your employer. ISSN: 3066-764X
Today's links Good ideas are popular: But they're impolitic. Hey look at this: Delights to delectate. Object permanence: Slinky treadmill; Ovipositors; Peter Thiel was right. Upcoming appearances: Where to find me. Recent appearances: Where I've been. Latest books: You keep readin' em, I'll keep writin' 'em. Upcoming books: Like I said, I'll keep writin' 'em. Colophon: All the rest. Good ideas are popular (permalink) In democracies, we're told, politicians exist to reflect and enact the popular will; but the truth is, politicians' primary occupation is thwarting the will of the people, in preference to the will of a small group of wealthy, powerful people. That's an empirical finding, based on a study of 1,779 policy outcomes, which concluded that: economic elites and organized groups representing business interests have substantial independent impacts on U.S. government policy, while average citizens and mass-based interest groups have little or no independent influence. https://www.cambridge.org/core/journals/perspectives-on-politics/article/testing-theories-of-american-politics-elites-interest-groups-and-average-citizens/62327F513959D0A304D4893B382B992B The policy preferences of the public would give the leadership of any mainstream party the fantods. Here's a remarkable thread where the economic anthropologist Jason Hickel summarizes recent polling on public preferences: https://x.com/jasonhickel/status/1953126243118813556 "Capitalism does more harm than good" (56% globally; 69% in France; 74% in India) https://www.edelman.com/news-awards/2020-edelman-trust-barometer In 28 of 34 countries, the majority are anti-capitalist: https://onlinelibrary.wiley.com/doi/10.1111/ecaf.12591 A majority of Canadians, Australians and Britons aged 18-34 believe "socialism will improve the economy and well-being of citizens": https://jacobin.com/2023/03/socialism-right-wing-think-tank-polling-support-anti-capitalism 62% of Americans aged 18-30 "hold favorable views of socialism" (61% of Democrats have a positive view of socialism vs 50% who are positive on capitalism): https://www.cato.org/blog/81-say-they-cant-afford-pay-higher-taxes-next-year Majority of youth climate group members blame "a system that puts profit over people and planet" and 89% say that system is capitalism: https://www.climatevanguard.org/publications-all/mapping-the-global-youth-climate-movement Majority support a national job guarantee (72% UK, 78% US; 79% France): https://www.jasonhickel.org/blog/2023/11/24/how-popular-are-post-capitalist-ideas Majority of Americans support workplace democracy (unions, worker shareholders and board seats): https://www.cambridge.org/core/journals/american-political-science-review/article/what-do-americans-want-from-private-government-experimental-evidence-demonstrates-that-americans-want-workplace-democracy/D9C1DBB6F95D9EEA35A34ABF016511F4 Majority of Britons support public ownership of services (education, healthcare, rail, water, postal service, parks); 64% of Americans support universal public health care; 64% support public options for internet, child care, and housing; https://www.jasonhickel.org/blog/2023/11/24/how-popular-are-post-capitalist-ideas 74% of Britons support national, permanent rent-controls; 71% of Bay Staters and 55% of Californians agree: https://www.jasonhickel.org/blog/2023/11/24/how-popular-are-post-capitalist-ideas 72% of Americans support a living wage; 87% of Britons agree: https://www.jasonhickel.org/blog/2023/11/24/how-popular-are-post-capitalist-ideas 84% of Europeans support a millionaires' tax; 69% of Americans agree: https://wid.world/document/international-attitudes-toward-global-policies-for-poverty-reduction-and-climates-change/ Majority of people in 40 countries want 4:1 maximum pay ratios for CEOs and their lowest-paid workers: https://journals.sagepub.com/doi/10.1177/1745691614549773 71% of Europeans want transformational reform of the UN and IMF, with proportional votes based on member-states' populations (58% of Americans agree): https://wid.world/document/international-attitudes-toward-global-policies-for-poverty-reduction-and-climates-change/ Majorities of Europeans and Americans support "compensating low-income countries for climate damages, funding renewable energy in low-income countries, and supporting low-income countries to adapt to climate change": https://wid.world/document/international-attitudes-toward-global-policies-for-poverty-reduction-and-climates-change/ 80-90% of people in medium/high-income countries want to finance this with a global tax on millionaires: https://wid.world/document/international-attitudes-toward-global-policies-for-poverty-reduction-and-climates-change/ Hickel's thread reminded of the 2023 Pew report that found that: 65% of Americans feel exhausted when thinking about politics; 63% have little/no confidence in the US political system; 4% think the US system works well: https://pluralistic.net/2023/10/18/the-people-no/#tell-ya-what-i-want-what-i-really-really-want Unsurprisingly: 87% of Americans want Congressional term limits; 79% favor age limits for Congress and the Supreme Court; 62% support automatic voter-registration for every American; 65% want to abolish the Electoral College (47% of Republicans agree!); 70% believe voters have too little influence over their representatives; 83% of Republicans say big donors call the shots (80% of Dems agree); 72% of Americans want to limit campaign contributions (75% D/71% R); 58% of Americans believe it is possible to get money out of politics. So on the one hand, this is all pretty dismal. It also makes the trend towards electing anti-democratic politicians who want to abolish elections a lot easier to understand: if you (correctly) believe you live in a world where politicians don't care about you, then why not vote for a strongman who'll punish your enemies and maybe leave you with a few more crumbs? But on the other hand, this is very exciting, because it shows us what a truly democratic world would look like (and just how different that world would be from the billionaire astroturf-dominated social media world)! If the popular will can achieve primacy, we would live in a veritable paradise! It also explains how candidates like Zohran Mamdani were able to clobber the political establishment simply by a) telling people that he would do popular things; and b) convincing them that he meant it. Suppressing popular preferences in (nominal) democracies isn't easy. It requires absolute unity of the ruling classes. Whenever the faintest crack appears in capital's unity, good policies gush out of it. That's what's happened with antitrust this decade, where the divisions between billionaire rentiers like Apple/Google and the millionaire capitalists who want to escape their 30% app tax has allowed a rush of effective antitrust enforcement to sweep the world, to the detriment of both: https://pluralistic.net/2025/06/28/mamdani/#trustbusting By not hanging together, the rich let us hang them separately. And since there is no honor among thieves – since the rich want nothing more to eat one anothers' lunches – there is disunity aplenty for us to exploit. We just have to remember that we are the (very large) majority and act like it. (Image: Japanexperterna.se, CC BY-SA 2.0, modified) Hey look at this (permalink) It’s not just Figma https://economicpopulist.substack.com/p/its-not-just-figma These GOP Lawmakers Referred Constituents to the CFPB for Help. Then They Voted to Gut the Agency https://www.propublica.org/article/cfpb-budget-cuts-gop-darrell-issa-john-cornyn The LLMentalist Effect: how chat-based Large Language Models replicate the mechanisms of a psychic’s con https://softwarecrisis.dev/letters/llmentalist/ AI Is A Money Trap https://www.wheresyoured.at/ai-is-a-money-trap/ Precarious Employment in Precarious Futures https://www.uncannymagazine.com/article/precarious-employment-in-precarious-futures/ Object permanence (permalink) #20yrsago Charlie Stross, Hugo winner https://web.archive.org/web/20050810024249/http://www.antipope.org/charlie/blog-static/2005/08/07/#hugo-thing #10yrsago Veiny, slick silicone ovipositors https://www.youtube.com/watch?v=wkfFZnK5W9s #10yrsago A treadmill for Slinky toys, for your infinite Slinky-torturing pleasure https://www.youtube.com/watch?v=9dinVcBEDhQ #10yrsago The Princess and the Pony, from Kate “Hark a Vagrant” Beaton https://memex.craphound.com/2015/08/07/the-princess-and-the-pony-from-kate-hark-a-vagrant-beaton/ #5yrsago Free the law https://pluralistic.net/2020/08/08/turkeys-for-christmas-party/#recap #5yrsago Google bans anticompetitive vocabularies https://pluralistic.net/2020/08/08/turkeys-for-christmas-party/#newspeak #5yrsago Peter Thiel was right https://pluralistic.net/2020/08/08/turkeys-for-christmas-party/#christmas-voting-turkeys #1yrago The Google antitrust remedy should extinguish surveillance, not democratize it https://pluralistic.net/2024/08/07/revealed-preferences/#extinguish-v-improve Upcoming appearances (permalink) Ithaca: AD White keynote (Cornell), Sep 12 https://deanoffaculty.cornell.edu/events/keynote-cory-doctorow-professor-at-large/ DC: Enshittification at Politics and Prose, Oct 8 https://politics-prose.com/cory-doctorow-10825 New Orleans: DeepSouthCon63, Oct 10-12, 2025 http://www.contraflowscifi.org/ San Francisco: Enshittification at Public Works (The Booksmith), Oct 20 https://app.gopassage.com/events/doctorow25 Miami: Enshittification at Books & Books, Nov 5 https://www.eventbrite.com/e/an-evening-with-cory-doctorow-tickets-1504647263469 Recent appearances (permalink) Tariffs vs IP Law (Firewalls Don't Stop Dragons) https://www.youtube.com/watch?v=LFABFe-5-uQ ORG at 20: In conversation with Maria Farrell https://www.youtube.com/watch?v=M9H2An_D6io Why aren't we controlling our own tech? (Co-Op Congress) https://www.youtube.com/live/GLrDwHgeCy4?si=NUWxPphk0FS_3g9J&t=4409 Latest books (permalink) Picks and Shovels: a sequel to "Red Team Blues," about the heroic era of the PC, Tor Books (US), Head of Zeus (UK), February 2025 (https://us.macmillan.com/books/9781250865908/picksandshovels). The Bezzle: a sequel to "Red Team Blues," about prison-tech and other grifts, Tor Books (US), Head of Zeus (UK), February 2024 (the-bezzle.org). "The Lost Cause:" a solarpunk novel of hope in the climate emergency, Tor Books (US), Head of Zeus (UK), November 2023 (http://lost-cause.org). "The Internet Con": A nonfiction book about interoperability and Big Tech (Verso) September 2023 (http://seizethemeansofcomputation.org). Signed copies at Book Soup (https://www.booksoup.com/book/9781804291245). "Red Team Blues": "A grabby, compulsive thriller that will leave you knowing more about how the world works than you did before." Tor Books http://redteamblues.com. "Chokepoint Capitalism: How to Beat Big Tech, Tame Big Content, and Get Artists Paid, with Rebecca Giblin", on how to unrig the markets for creative labor, Beacon Press/Scribe 2022 https://chokepointcapitalism.com Upcoming books (permalink) Canny Valley: A limited edition collection of the collages I create for Pluralistic, self-published, September 2025 Enshittification: Why Everything Suddenly Got Worse and What to Do About It, Farrar, Straus, Giroux, October 7 2025 https://us.macmillan.com/books/9780374619329/enshittification/ Unauthorized Bread: a middle-grades graphic novel adapted from my novella about refugees, toasters and DRM, FirstSecond, 2026 Enshittification, Why Everything Suddenly Got Worse and What to Do About It (the graphic novel), Firstsecond, 2026 The Memex Method, Farrar, Straus, Giroux, 2026 The Reverse-Centaur's Guide to AI, a short book about being a better AI critic, Farrar, Straus and Giroux, 2026 Colophon (permalink) Today's top sources: Naked Capitalism (https://www.nakedcapitalism.com/). Currently writing: "The Reverse Centaur's Guide to AI," a short book for Farrar, Straus and Giroux about being an effective AI critic. (1048 words yesterday, 23678 words total). A Little Brother short story about DIY insulin PLANNING This work – excluding any serialized fiction – is licensed under a Creative Commons Attribution 4.0 license. That means you can use it any way you like, including commercially, provided that you attribute it to me, Cory Doctorow, and include a link to pluralistic.net. https://creativecommons.org/licenses/by/4.0/ Quotations and images are not included in this license; they are included either under a limitation or exception to copyright, or on the basis of a separate license. Please exercise caution. How to get Pluralistic: Blog (no ads, tracking, or data-collection): Pluralistic.net Newsletter (no ads, tracking, or data-collection): https://pluralistic.net/plura-list Mastodon (no ads, tracking, or data-collection): https://mamot.fr/@pluralistic Medium (no ads, paywalled): https://doctorow.medium.com/ Twitter (mass-scale, unrestricted, third-party surveillance and advertising): https://twitter.com/doctorow Tumblr (mass-scale, unrestricted, third-party surveillance and advertising): https://mostlysignssomeportents.tumblr.com/tagged/pluralistic "When life gives you SARS, you make sarsaparilla" -Joey "Accordion Guy" DeVilla READ CAREFULLY: By reading this, you agree, on behalf of your employer, to release me from all obligations and waivers arising from any and all NON-NEGOTIATED agreements, licenses, terms-of-service, shrinkwrap, clickwrap, browsewrap, confidentiality, non-disclosure, non-compete and acceptable use policies ("BOGUS AGREEMENTS") that I have entered into with your employer, its partners, licensors, agents and assigns, in perpetuity, without prejudice to my ongoing rights and privileges. You further represent that you have the authority to release me from any BOGUS AGREEMENTS on behalf of your employer. ISSN: 3066-764X
