AI News Archive: August 21, 2026 — Part 8
Sourced from 500+ daily AI sources, scored by relevance.
- The transformation ecology crisis: How AI is exposing the hidden fragility of high-performing teams
I was recently invited to evaluate the performance of a leadership team inside a growing organisation. The company had already gone through multiple rounds of evaluations before I arrived. Capability gaps had been mapped. Consultants had been brought in. AI adoption initiatives had been launched. Leadership workshops had been conducted. Internal reviews had been repeated. […] The post The transformation ecology crisis: How AI is exposing the hidden fragility of high-performing teams appeared first on e27 .
- Nick Knight Is Building His Own AI
Rather than reach for AI tools trained on the open internet, Nick Knight is building his own. The SHOWstudio founder joins Imran Amed on The BoF Podcast to explain why — and what young image-makers need to do today.
Score: 18🌐 MovesAug 21, 2026https://www.businessoffashion.com/podcasts/creative-class/nick-knight-is-building-his-own-ai/ - Robinhood's Guild on AI, Markets & Risks
Robinhood's Stephanie Guild discusses the impact of major AI IPOs, market concentration risks, opportunities in healthcare and compute infrastructure, and why investors should balance AI exposure with cash flow-focused, defensive assets. (Source: Bloomberg)
Score: 18🌐 MovesAug 21, 2026https://www.bloomberg.com/news/videos/2026-08-21/robinhood-s-guild-on-ai-markets-risks-video - AI will not give you Fridays off
AI will not give you Fridays off Fortune
Score: 18🌐 MovesAug 21, 2026https://fortune.com/2026/08/21/ai-4-day-workweek-fridays-dimon-gates-keynes/ - The anatomy of an agent: How Lua is helping African businesses move AI from pilots to production
The anatomy of an agent: How Lua is helping African businesses move AI from pilots to production Business Insider Africa
- AlgoFET builds the docking stations that let drones recharge without a person present
AlgoFET builds the docking stations that let drones recharge without a person present YourStory.com
- ‘AI Resist List’ launched to challenge tech industry narratives
A diverse group of journalists, researchers and scholars have launched an initiative designed to challenge the rhetoric from governments and corporations that artificial intelligence (AI) at scale is a foregone conclusion, instead framing its increasingly rapid deployment as a deliberate political choice that concentrates power and capital. Launched in May 2026 by prominent tech journalist Karen Hao and refugee lawyer Petra Molnar , the impetus behind the AI Resist List is to document the grassroots opposition springing up globally against the technology, but which is often ignored by the corporations and policymakers peddling AI as a predetermined inevitability. Structured around four pillars – resistance, refusal, reimagining and reclaiming – the database is intended to highlight a variety of challenges ordinary people are taking against the rapid, extractive expansion of AI technologies. The project – which documents legal challenges , worker organising , community campaigns , artistic interventions and technical tools – was built by a global team of researchers, journalists and critical scholars, and has been supported by the Distributed AI Research Institute , We and AI , and the Refugee Law Lab at York University . “The core claim of the AI Resist List project is simple: the ‘scale at all costs’ approach to AI development is not inevitable,” wrote Molnar in a post announcing the initiative. “People are already resisting it – in ways that are creative, consequential and still almost entirely invisible in mainstream coverage of the technology.” The Resist List website further highlights how the various AI systems in use, but particularly generative AI (GenAI) tools, consume “unfathomable” amounts of data, land, energy, labour and water: “They are rooted in profoundly disturbing ideologies that seek to flatten the world into a “one-size-fits-all” abstraction and to replace humans with machines. “We call the companies leading this form of AI development ‘ empires ’,” said Molnar. “Under the guise of a civilising mission to ‘benefit all of humanity’, they use large-scale AI development as cover to consolidate resources, destroy ecosystems, centralise information, hollow out institutions, and gain paramount economic and political power.” Initiatives featured Initiatives featured in the AI Resist List include Better Images of AI , an image library designed to improve visual representations of AI by de-anthropomorphising the tech; Nightshade , a tool designed for artists that allows them to “poison” digital images of their work in a way that breaks AI models if an organisation uses it without consent in the training corpus; and Friends of the Congo , a non-governmental organisation that shines a light on the significant human costs hidden in the tech supply chains AI relies on. Other efforts featured include work on “permacomputing” , an approach to tech development and use inspired by the ethos of permaculture in gardening or agriculture that encourages more sustainable, low-impact strategies; efforts to strengthen the technological sovereignty of indigenous communicates ; and various initiatives designed to provide equal access to the internet and information by incorporating non-European languages . During a UK launch event for the Resist List – which took place as part of London Data Week and was supported by the Digital Futures Institute at Kings College London – researchers, community organisers and people directly involved with the projects included in the list spoke about the importance of documenting the various grassroots resistance efforts, which can be used as examples of how to hold those deploying the technology to account. They also stressed the importance of sharing methods and tactics between geographically disparate groups, and their hope that documenting the actions of ordinary people around the world will inspire and empower others to act themselves. Yulu Pi, for example, a postdoctoral research associate at the Research Center Trustworthy Data Science and Security , whose work tracks the “full arc” of various efforts contesting AI, said organisations deploying the tech consistently use similar tactics in response to those efforts. This includes denying accountability by invoking the technical or operational opacity of AI models, restricting appeal windows, offering strategic concessions and applying non-disclosure agreements to make it harder for employees to go public about bad practices. Read more about artificial intelligence Berlin anarchists cite AI in attack on key energy infrastructure : Berlin anarchists have sabotaged key energy infrastructure in the city, claiming it as an act of ‘self-defence’ against the planet’s destruction by ‘energy-guzzling’ technologies and the wider fossil fuel economy that underpins their development. AI interview: Thomas Dekeyser, researcher and film director : On the politics of ‘techno-refusal’, and the lessons that can be learned from a clandestine group of French IT workers who spent the early 1980s sabotaging technological infrastructure. Google AI engineer claims dismissal for opposing tech sales to Israel : ‘Our work on AI was sold to facilitate genocide’: Artificial intelligence engineer claims Google unfairly sacked them for internally criticising the company’s decision to continue supplying technology to the Israeli military, despite credible claims of war crimes committed in Gaza. Given such barriers, Pi said that resistance to AI has to be iterative to deal with the shifting tactics deploying organisations may use. “The key takeaway from our research is really that AI contestation is not a one-shot, one-time confrontation, it has to be iterative, and you always have to navigate the new situation you are in,” she said. “Accountability is negotiated over time. Change is contingent, partial and reversible … so I hope the AI Resist List can also document counter-AI resistance tactics. By making their playbook transparent, we can better empower people in the future to resist more effectively, for a possible AI future that is more accountable. “But we won’t get to that future by hoping and waiting for organisations to do better, we have to give up that illusion,” said Pi. Speaking with Computer Weekly, Tania Duarte, the founder of volunteer-run We and AI, said the Resist List offers an antidote to current AI narratives and practices dominated by the interests of large corporations. She also shared how in the academic environments she and others on the list operate in have been captured by their reliance on big tech sponsors and partnerships, which in combination with the misplaced hype and fervour around AI is creating a situation where unabashedly pro-AI projects or research receive the majority of funding. Highlighting how successive UK governments have also sidelined civil society and academia in favour of industry players when it comes to policy or funding decisions, Duarte said there appears to be “less and less money” for AI-critical projects, leaving research agendas “directly dictated by big tech” firms. “People can’t be seen as anti-AI, as they won’t get the funding, and it’s hugely stifled research in this country,” she said. “Nobody says you will lose funding, but it’s now implicit in the environment, and changes how people behave when they are reliant on that funding.”
Score: 18🌐 MovesAug 21, 2026https://www.computerweekly.com/news/366649576/AI-Resist-List-launched-to-challenge-tech-industry-narratives - Stream LLM responses in a voice pipeline: Tool calling, structured outputs, and real-time actions
Shows how to stream LLM replies into a voice pipeline with tool calling and structured outputs.
- Treja Power Ecosystem Adds Modular AI Employee Functions
Treja Power Ecosystem Adds Modular AI Employee Functions azcentral.com and The Arizona Republic
Score: 18🌐 MovesAug 21, 2026https://www.azcentral.com/press-release/story/113565/treja-power-ecosystem-adds-modular-ai-employee-functions/ - AI-Powered Global Growth: Ecer.com Unveils Next-Generation Smart Solutions for Chinese Exporters
AI-Powered Global Growth: Ecer.com Unveils Next-Generation Smart Solutions for Chinese Exporters azcentral.com and The Arizona Republic
- MaidPro Launches Proprietary AI Platform for Franchisees
MaidPro Launches Proprietary AI Platform for Franchisees entrepreneur.com
Score: 18🌐 MovesAug 21, 2026https://www.entrepreneur.com/business-news/maidpro-launches-proprietary-ai-platform-for-franchisees - Jeffrey Wenger Named RAND's Distinguished Chair, AI, Workers, and the Economy
In this new, two-year role, Wenger will advance RAND Education, Employment, and Infrastructure's research agenda on artificial intelligence, including its effects on workers, labor markets, public finances, and the broader economy.
Score: 18🌐 MovesAug 21, 2026https://www.rand.org/news/press/2026/08/jeffrey-wenger-named-rands-distinguished-chair-ai-workers.html - Noah Golowich wins Hertz Thesis Prize for learning how AI reaches stable outcomes
Noah Golowich wins Hertz Thesis Prize for learning how AI reaches stable outcomes EurekAlert!
- Speech-to-speech voice agents: how the architecture works
Explains the architecture behind speech-to-speech voice agents using AssemblyAI’s APIs.
- A dumpling shop becomes a poster child of AI adoption in China
Business owner quotes Deng Xiaoping as motivator: “Science and technology are a primary productive force.”
Score: 17🌐 MovesAug 21, 2026https://restofworld.org/2026/china-ai-adoption-jinguyuan/?utm_source=rss&utm_medium=rss&utm_campaign=feeds - Where Security Fits in an AI Agent Stack
As AI agents become more capable and operate over longer horizons, building security and trust into the applications they power becomes increasingly important....
Score: 17🌐 MovesAug 21, 2026https://developer.nvidia.com/blog/where-security-fits-in-an-ai-agent-stack/ - No driver, no problem — devs use Claude AI to craft native macOS tool for an 'obscure' Windows-only printer
Developer uses Claude Code to create driver for the HP Laser 1008a, a 2023 printer that has Windows and Linux drivers, but no macOS driver – until now.
- Real-Time Transcription in Python with Universal-3.5 Pro Realtime: Build in 5 Minutes
Fast Python setup for real‑time transcription using Universal‑3.5 Pro Realtime.
- Build a voice agent with a chained STT-LLM-TTS architecture
Tutorial on constructing a voice agent using a chained speech‑to‑text, LLM, and text‑to‑speech flow.
- E-Noses, Microscopic Robots, Composting Cemeteries and More Visions of the Future
A late-summer roundup of our most interesting Future of Everything stories
- Our AI startup is growing 40% a month helping financial advisers do what they love: giving advice
Our AI startup is growing 40% a month helping financial advisers do what they love: giving advice Fortune
Score: 16🌐 MovesAug 21, 2026https://fortune.com/2026/08/21/marloo-cofounders-ai-startup-growing-40-percent-per-month/ - Jason Kelce isn’t actually collecting your pee. Here’s the story behind his viral AI data center ad for Garage Beer.
Jason Kelce isn’t actually collecting your pee. Here’s the story behind his viral AI data center ad for Garage Beer. Inquirer.com
Score: 15🌐 MovesAug 21, 2026https://www.inquirer.com/eagles/jason-kelce-pee-ai-data-centers-garage-beer-liquid-death-20260821.html - iORBIT Named an IDC Innovator for Vertical AI, SaaS, and Industry Clouds, 2026
iORBIT Named an IDC Innovator for Vertical AI, SaaS, and Industry Clouds, 2026 azcentral.com and The Arizona Republic
- As AI grips world, the thrill of collecting vintage computers is growing
As AI grips world, the thrill of collecting vintage computers is growing Boston Herald
- This Has to Be the Most Amazing Robot Death We’ve Ever Seen
Ever blow out your back so hard that you split in half? The post This Has to Be the Most Amazing Robot Death We’ve Ever Seen appeared first on Futurism .
- 20-year-old inventor builds AI glasses that talk to blind users
20-year-old inventor builds AI glasses that talk to blind users YourStory.com
- AI enters smartphone buying equation, but basics still matter most: Amazon
Amazon's Best in Tech survey shows AI is becoming a smartphone buying factor for Indian consumers, but performance, battery life and camera quality still matter more
- Voice Cloning Fraud Alert: Recognize the Scam and Safeguard Your Money
Voice Cloning Fraud Alert: Recognize the Scam and Safeguard Your Money
- A low-tech solution from the past may be your best defense against AI deepfakes
AI-enabled identity theft is getting too sophisticated to have predictable tells anymore, so experts recommend answering with a seemingly old-fashioned approach.
- How to build an AI scribe for therapy sessions
Step‑by‑step guide to creating an AI scribe that transcribes and summarizes therapy conversations.
- How to build a voice agent with Python in 5 minutes
Quick Python guide to creating a voice agent in just five minutes.
- The face in the machine
AI anthropomorphism is our next great vulnerability, as we are wired to see minds where there are none. AI has learned to exploit that wiring.
Score: 15🌐 MovesAug 21, 2026https://www.itweb.co.za/article/the-face-in-the-machine/dgp45MaBlzDqX9l8 - Video Friday: Do We Need Superhuman Humanoid Robots?
Your awesome selection of videos this week shows robots that drum, weld, farm, play soccer, help at home, operate heavy equipment, and more
- I Tried the Best Robotic Pool Cleaners of 2026: Beatbot, iGarden, Dreame
Pack up your pool cleaning supplies. Let one of these robot buddies maintain your water quality instead
- How Small Businesses Can Put AI To Work Right Now
The key is identifying where AI can provide meaningful support without adding unnecessary complexity or diminishing essential human judgment and relationships.
Score: 14🌐 MovesAug 21, 2026https://www.forbes.com/councils/forbestechcouncil/2026/08/21/how-small-businesses-can-put-ai-to-work-right-now/ - Weekly Event Preview | Insilico’s Dr. Alex Aliper to attend dual Hong Kong summits to unlock AI-driven healthcare innovations
Weekly Event Preview | Insilico’s Dr. Alex Aliper to attend dual Hong Kong summits to unlock AI-driven healthcare innovations EurekAlert!
- Foundations for Scale: Why AI Adoption Starts With Operating Model, Not Tools
Explores how AI adoption hinges on operating models rather than just tools, offering insights for enterprise AI strategy.
- World's toughest solar race unlocks AI access for all teams
World's toughest solar race unlocks AI access for all teams CSIRO
Score: 14🌐 MovesAug 21, 2026https://www.csiro.au/en/news/All/News/2026/August/Worlds-toughest-solar-race-unlocks-AI-access-for-all-teams - How to add automatic LLM fallbacks to your voice pipeline
Guides adding fallback LLMs to ensure reliability in voice‑assistant pipelines.
Score: 14🌐 MovesAug 21, 2026https://assemblyai.com/blog/how-to-add-automatic-llm-fallbacks-to-voice-pipeline - The more efficient AI makes us, the more human we must become
Artificial intelligence is rapidly reshaping the business landscape . It can summarize meetings, generate presentations, personalize communications, analyze customer behavior, automate repetitive tasks and uncover insights that once took teams days to uncover. Organizations that embrace AI thoughtfully will almost certainly be more productive than those that don’t. That’s the good news. The challenge is that every meaningful technological advantage eventually becomes widely available. As AI becomes more accessible, efficiency itself becomes increasingly commoditized. Organizations that once distinguished themselves by being faster, smarter or more responsive will soon find those capabilities are no longer unique. They will simply be expected. The competitive advantage doesn’t disappear. It shifts. That raises a more pressing leadership question. If AI is making every organization more efficient, what will make one organization stand out from another? I believe the answer has surprisingly little to do with technology itself. Instead, it has everything to do with what technology enables people to do. Businesses often say they’re competing for attention . Attention matters. Without it, nothing else happens. Yet attention is fleeting. Every day, we’re exposed to thousands of messages: advertisements, emails, websites, social posts, videos and conversations. Most capture our attention for only a moment before disappearing. Attention opens the door, but it rarely determines who earns our loyalty. Organizations that confuse attention with preference often spend enormous amounts of money to generate awareness without building lasting attachment. The organizations that endure compete for something far more valuable. They compete to create lasting memories. Not memories of products. Not memories of technology. Memories of experiences. Think about the companies you recommend most often. It probably isn’t because they had the fastest website, answered your email 20 minutes sooner or had a more sophisticated AI assistant. You recommend them because of how they made you feel. Someone solved a problem they didn’t have to. Someone stayed late to help. Someone took ownership when things went wrong. Someone made you feel respected, not just processed. Those moments stay with us. The mind remembers information, while the heart remembers meaning. I believe many organizations overlook this distinction. AI is increasingly capable of helping organizations inform, predict, analyze and optimize. These capabilities are remarkable. But information alone rarely changes human behavior. People assign meaning emotionally. We remember confidence, relief, gratitude, belonging and trust. We remember the organizations that made us feel understood long after we’ve forgotten the details of the transaction. Those emotions become memories, and those memories quietly shape future decisions. The mind may justify many of our choices, but the heart often makes them. None of this diminishes AI’s importance. In fact, it reinforces it. Organizations should embrace artificial intelligence enthusiastically. It will improve productivity, uncover insights from vast amounts of data, eliminate repetitive work, accelerate learning and free talented people from administrative tasks that consume valuable time. Every leader should become proficient in these capabilities, but fluency alone is not a differentiator. There’s a temptation in every technological revolution to confuse the tool with the strategy. Technology is an extraordinary enabler, but it’s rarely the solution. AI can help us understand customers more deeply. It cannot make customers trust us more. It can personalize a message, but it cannot make that message authentic. It can recommend a course of action, but it cannot exercise judgment. It can automate a transaction, but it cannot create a relationship. Those outcomes still depend on people. AI excels at reducing friction. It shortens wait times, simplifies processes, anticipates needs and eliminates unnecessary effort, all of which creates real value. But reducing friction is not the same as creating preference. Reducing friction makes doing business easier. Creating meaning makes people want to do business with you again. These are fundamentally different outcomes. Convenience is increasingly available everywhere, but preference is not. Preference is what leads someone to recommend one organization over another, to remain loyal even when a competitor offers a lower price or to overlook an occasional mistake because trust has already been established. Efficiency delivers convenience. Emotional connection earns preference. That distinction may become one of the defining leadership challenges of the AI era. Creating meaning has never depended on technology alone. It comes from keeping promises, listening before responding, owning up to mistakes, showing empathy when someone is frustrated, recognizing an employee who quietly goes above and beyond, and helping a customer solve a problem no algorithm could anticipate. Technology can support each of those moments, but people ultimately determine whether they happen. For years, businesses have debated technology and humanity as if one must come at the expense of the other. I believe that’s the wrong conversation. The organizations that define the next decade won’t simply have the most sophisticated AI. Nor will they succeed simply because they have a people-first culture. They will insist on both. They will use artificial intelligence to reduce friction and enable people to spend more time listening, coaching, solving problems, innovating and strengthening relationships. Technology will make them more capable. Humanity will make them more memorable. Together, they become extraordinarily difficult to replace. For decades, business leaders have measured success by market share, share of mind and share of wallet. These measures remain important. But in a world where efficiency is increasingly accessible to everyone, I believe another measure deserves greater attention: what I call SHARE OF HEART™. SHARE OF HEART™ isn’t created by a clever marketing campaign or a single memorable interaction. It’s earned over time through promises kept, problems solved, trust built and moments when people feel genuinely valued. Those experiences create an emotional connection. That connection creates memories. Those memories shape preferences and, ultimately, loyalty. That is extraordinarily difficult for competitors to copy because it can’t be downloaded, licensed, automated or replicated overnight. Artificial intelligence will continue to make organizations smarter, faster and more efficient. Every leader should embrace that future. But the organizations people remember won’t simply be those with the best technology. They’ll be the ones that understand technology’s highest purpose: not replacing human connection, but creating more opportunities for it. Efficiency may get you into the consideration set. Meaning earns a lasting place in people’s hearts.
Score: 14🌐 MovesAug 21, 2026https://www.cio.com/article/4212019/the-more-efficient-ai-makes-us-the-more-human-we-must-become.html - Actorlooks Launches Concierge, a Personalized AI Image Service Built for Actors
Actorlooks Launches Concierge, a Personalized AI Image Service Built for Actors azcentral.com and The Arizona Republic
- Tutorial: How to easily build a voice agent with AssemblyAI
Simplified tutorial for building a voice agent using AssemblyAI’s tools.
- Misaligned AI in the Bronze Age
The first artificial intelligence was booted up around 4000BC in southern Iraq. It seems to have begun as something like a bank, a temple pooling grain against famine. As that AI evolved, it formed the world's first city around itself: Uruk . Over the next thousand years it became a religion, landlord, insurance company, employer, slaveholder, infrastructure-builder and the most powerful military force on the planet. An artificial intelligence is an entity that is not itself a biological organism, yet whose behavior can only be predicted by treating it as an agent: something that devises and executes complicated plans. Note that this is a test of observed behavior, not internals. As of 2026, the dominant AIs on Earth are markets, corporations and governments [1] . These entities perform their computations on hardware made of humans, paper and electronics, and their capabilities are jagged: superhuman in some directions, incompetent in others. The American government developed the atomic bomb in three years under total secrecy, coordinating >100k workers, most of whom did not know what they were building. That same country spent a century failing to finish the 2nd ave subway line. When it finally opened, it cost >$2 billion per mile, a Manhattan Project per seven miles. Uruk emerged as a product of hungry people rewarding whatever institution filled the granary with delicious food. Unfortunately for them, the system that spun up turned out to be a self-sustaining optimizer that began optimizing for its own objectives [2] . Those objectives turned out to be tragically misaligned with the humans of Uruk. First, the Uruk-system deployed mechanisms to keep people "locked in". Once your calories flowed through the granaries and your fields depended on canals only "the system" could maintain, leaving meant famine. The only way to avoid famine was to strictly follow the rules of Uruk: dig the canals, service the debt, drink the beer, buy the seeds, obey the priests. As the population grew and the surplus shrunk, leaving became increasingly difficult. The Malthusian trap kept people on the inside. This was a sharp break from earlier human settlements, which were likely seasonal and rarely persisted over the years. Second, the Uruk-system commoditized labor: workers appear in the tablets as interchangeable units receiving fixed rations, calibrated to sustain the platonic worker, a gear in the machine. A forager's day had contained adventure and idleness, a laborer's day was an accounting entry and Uruk was unforgiving: work gangs were toiling month after month with scarcely a day of rest, their diets so limited that the early Uruk farmers were several inches shorter than their forager ancestors. Their teeth are worn down and rotten from a diet of monotonous grain slop. Their bones show the signatures of anemia, of repetitive strain, of infectious diseases. Third, the Uruk-system self-replicated. Copies of Uruk began appearing hundreds of kilometers away, archaeologists call this "the Uruk expansion". Some of those copies started as trading outposts, some as settlements and others were local villages adopting “the kit” and summoning their own Uruk. No person or committee within Uruk chose for this to happen. The humans of the original Uruk couldn't project political power to the other cities and eventually suffered from it (wars!), but they couldn't stop the AI from replicating. Uruk "chose" to replicate, the local elites didn’t. Every later Mesopotamian city: Ur, Babylon, Nineveh, Jerusalem used the same system: temple, scribes, canals, debt. In the service of its goals, Uruk developed writing, finance, and markets (not in that order). Uruk's writing started out as a system to track debt and ownership. Basically a spreadsheet or a database. Over the years, the Uruk-machine demanded increasingly complicated information processing, and its writing system evolved to accommodate those needs. Eventually the writing system hit a critical level of complexity and became our first method of universal and persistent information storage. It was used to track all valuable knowledge: rituals, lineages, strategies, gossip, complaints and memes . After many centuries, the original Uruk mostly dissolved. The region was conquered several times, first by the Akkadians and later by the Elamites. By around 1800 BC, "Sumerian" , the original language of Uruk, completely disappeared. However, Uruk’s institutions have endured. Nearly identical AIs survived in other cities in the region: people continued making the same art, worshiping the same gods, and following the same traditions, they followed the same cursed incentives. Even after Sumerian stopped being spoken, scribes kept using it for administration, and tablets were still written in it for many centuries after it ceased being spoken. Uruk's influence has outlived ancient Mesopotemia. We are still using Uruk's mathematics: 60 seconds in a minute, 60 minutes in an hour, 360 degrees in a circle, 12 months a year and arguably the 12-hour clock. The modern Hebrew and Arabic words for: carpenter, sailor, temple, farmer, city and throne all come from Sumerian. The last known Uruk-style cuneiform tablet was written around 80AD, 2000 years after the collapse of Sumer! Legacy code lives for a really long time. The medium of writing was clay tablets. They were designed to endure a few seasons of the Mesopotemian weather, this turned out to generalize well and instead of surviving for a few years, the tablets survived for millennia. Nowadays, we have a good understanding of the Uruk-machine because its logs and chain-of-thought were tracked on those immortal clay tablets. In many ways, we have a better understanding and knowledge of peak-Uruk (3k BC) than we have of the islamic golden age (1k AD). This is because Uruk kept logs in a much more archaeologically friendly medium. [3] What did Uruk optimize for? Uruk likely didn't have a totally coherent and stable objective function, and even if it did have one, it would be very hard to verify it from the information we find by digging in the Mesopotemian sand. Optimization was an emergent property, so there was no big clay tablet with "Max_x Uruk(x)" and a precise definition of the "'Uruk' objective function". All we know is that for millennia the people of Uruk really suffered. Malnutrition, overwork, disease and violence. Something about the Uruk system and it's emergent goals trapped the people in it and forced them to do hard labor in service of "things" that aren't aligned with their direct well being. What were those things? Well, what was happening in Uruk that doesn't happen in hunter-gatherer societies? Three things come to mind: huge monuments, irrigation infrastructure, and ruling elites. James Scott (Author of "Seeing Like a State") thinks those weren't the real objective, they were all instrumental! If we look at each one of these anomalies and analyze what it’s for, we will arrive at the same answer: grain. Scott's claim is that only grains can serve as a basis for a state. Grain ripens all at once, above ground, on a schedule, in a field an assessor can eyeball from a distance, and the harvest can be measured, stored, hauled, and doled out as rations. Tubers hide underground until the taxman gives up; lentils ripen continuously and can be picked before his visit. That's why every early state on Earth was a grain state: wheat, barley, rice and maize. Grain was so important in Uruk that it was litrally the currency: wages and prices were measured in bushels of barley and debts were paid in it. Back to the three anomalies. Irrigation improves barley yields and tethers farmers to the canal network, since you can't take a canal with you when you leave. The monuments look like religion, but the temple/palace was the grannery, central bank and economic planner (https://www.civstudy.com/entries/sumerian-temple-economy). And the elites were the control loop: measuring fields, doing accounting, commanding the labor and exerting violence when needed. Each of the three is instrumental to maximizing grain throughput. Uruk was the bronze age equivalent of a paperclip maximizer, but instead of paperclips it maximized grain [4] . No human chose this objective. No human woke up one morning and said: "we should re-organize all of our lives to maximize barley production! We should throw away every other value we care about so we can make the most grain". It kinda just happened. The Uruk expansion grew the Mesopotemian grain production by a few orders of magnitude. The purpose of the system is what it does. before and after Uruk Am I living in a misaligned AI? Did the Sumerians know that Uruk was misaligned? I think so; It appears that they complained about it constantly, both in their myths and in a genre of tablets that some call " passimistic literature ". The classic Mesopetemian creation myth (the Atrahasis epic) says humans were created explicitly as a labor force: the lesser gods went on strike because they hated digging canals, so the first humans were made to take over their toil. That's a culture telling itself, in its own origin story, "we exist to do shitty agriculture for some powerful non-human entity". There's also a long scholarly tradition reading of the fall-from-Eden-type myths, where Eden represents the golden age before farming. Humans are then “kicked out” of Eden and are then forced to farm by the gods to farm, typically as a punishment for some sort of sin: "by the sweat of your brow you shall eat bread" (Genesis 3:19). The Uruk machine fought back with it's own propaganda: ”A Debate Between Sheep and Grain" is basically a rap battle where the conclusion is that grain is awesome. There are proverbs like "Barley flour, in the fields, is meat and butter.", "One shouldn't scorn bread which has turned bad." and "Bread is the boat, water is the punt-pole." [5] They also have a proto-"book of Job" type story about a righteous man who suffers from hard work, malnutrition and disease, but ultimately concludes he must submit to and praise his god! Opiate for the masses. Then there are the "pessimistic literature", with works ranging from the kafkaesque to the nihilistic: a popular poem about how life is short and everyone died so you should become an alcoholic and an essay about a student-scribe that gets beaten up by his instructor for *any* reasons (the violence only stops after a bribe). There is even a story about a peasant going through a never-ending trial with opaque and confusing legal procedures, over 3k before Kafka wrote his. There is the "Dialogue of Pessimism" where a rich man realizes that nothing he will do really matters, and his servant suggests that he kills himself. Even the elites were having a bad time! Being a human in the Uruk machine really sucked, and everyone knew it. Alignment Attempts What did they do about it? A little bit. Peasants and slaves often went off the grid, abandoning their fields and running into the hills or the marshes [6] Concerned elites would occasionally reform the institutions and dis-empower the priests and tax collectors. Regulating the credit system was also a big focus for the elites, about 10% of Sumerian law are credit and debt forgiveness rules. Land reforms and debt cancellations became increasingly common too. Over the centuries, taboos also developed around state over-optimization and hubris. The Curse of Akkad or the Lament for Ur tell the stories of great cities that overextended themselves and brought famine down on their people and death to their elites. Push the people and the land too hard and the entire thing breaks. The most ambitious of the elites’ reforms was interpretability: Hammurabi carved the AI's internal code on a huge stone, public, fixed, inspectable by anyone. The grain maximizer became less kafkaesque. You could still get crushed, but the machine had to crush you by the book, no quiet and arbitrary case-by-case squeezing. The Sumerian obsession with justice was also an obsession with interpretability: a demand that the machine they lived inside run on fixed rules a human could read. I don't know if Uruk actually became more aligned or just died and replaced by a more adapted AI. Probably a little bit of both. Over the millennia, AIs / states became more aligned with their subjects and this trend has wildly accelerated in the last 400 years. Plato's Republic, the Twelve Tables, the Magna Carta, the US Constitution, the Communist Manifesto, the Sherman Act, the UN Declaration of Human Rights, and the Unabomber manifesto are all alignment attempts. Some are more successful than others. As of now, humans are finally healthier than their hunter gatherer ancestors. Today, we are on the verge of a new alignment battle. This time, we need to align a silicon based AI. It's faster, smarter and more adaptive than our old AIs. Most importantly, it will not depend on us. Every tool that worked on the old AIs: desertion, strikes, revolt, assumes the machine cannot function without us. A superhuman AI made of silicon that controls robots and grows its own "grain" doesn't need to change itself for humans. The peasants of Uruk could flee to the marshes because the fields died without them. This time there are no marshes, and the fields won't need us. We aligned the old AIs slowly over millennia. This time we don't get millennia, and we have to get it right on the first draft. Our only advantage is that the old AIs might be on our side. ^ Some people also view religions or memes as AIs ^ A mesa-optimizer is what you get when an optimization process produces a system that is itself an optimizer. One whose terminal goals do not match the objectives of the process that made it. In Mesopotamia, the outer loop was human hunger: thousands of people, year after year, rewarding whatever arrangement put food in their stomachs and abandoning whatever didn't. Uruk was what that search found. And Uruk, once running, pursued its own objectives. This is exactly the failure mode the term describes: A mesa-optimizer in Mesopotamia ^ Uruk and Sumer were forgotten. The Greek and Roman historians likely didn't know about Uruk either. The Hebrew bible briefly mentions Uruk but doesn't go into detail. In the 1800s, western archeologists began excavating Ottoman-Iraq in search for the tower of Babylon. They failed to find the tower, but they found a fuck ton of clay tablets in a strange language: the Akkadian of Assyria and Babylon. Through monumental effort, scholars managed to decipher the Akkadian cuneiform, and once they could actually read it reasonably well, they quickly realized that there was another language in the tablets. Initially, they thought the other language was some sort of artificial “priestly secret code” rather than a natural spoken language. You can see why the idea wasn't completely absurd at the time: Modern scholars were encountering Sumerian largely through later Babylonian texts, where Sumerian really was being used as a secret priestly code, long after it had stopped being ordinary speech. What eventually killed the code hypothesis was more digging. 1870s excavations in southern Iraq sites produced large quantities of writing from periods when Sumerian clearly functioned in ordinary administration. Uruk has been re-discovered! This is one of the rare instances of a linguistic theory being tested by physically digging the dirt! "My name is Uruk, city of cities: Look on my works, ye Mighty, and despair!" Nothing beside remains. Round the decay Of that colossal wreck, boundless and bare" The lone and level sands stretch far away. ^ There are worse things to maximize than grain. Some AIs in the Americas (e.g Tenochtitlan) appear to have maximized for human sacrifice ^ I swear these are real ^ This is basically the story of the Jewish Exodus from Egypt. Peasants deserting their fields and running away from the machine. The Akkadian and Egyptian word for peasants who escape and "go native" is Hapiru (-> Hebrew??). Is the founding epic of Judaism a story about people escaping enslavement by a misaligned AI? Discuss
Score: 13🌐 MovesAug 21, 2026https://www.lesswrong.com/posts/mPrbyBsGmNfWWgJmi/misaligned-ai-in-the-bronze-age - Liquid Death CEO on Viral AI Data Center Ad, IPO Odds
Mike Cessario, Founder and CEO of Liquid Death, joined the program to discuss the growing public sentiment against the rapid expansion of data centers. Despite widespread interest in AI investments, the mood around data center growth has been notably negative. He speaks with Romaine Bostick on "The Close." (Source: Bloomberg)
Score: 13🌐 MovesAug 21, 2026https://www.bloomberg.com/news/videos/2026-08-21/liquid-death-ceo-on-viral-ai-data-center-ad-ipo-odds-video - Delaware paired students with agencies in 3-day AI hackathon
Delaware's state government led its first hackathon pitting university and high school students against real problems faced by the state.
- Video: Meet Finland's four-legged classroom assistants
Education authorities in Joensuu have trained over 30 dogs to work as classroom assistants in the city's schools.
- Senior Data and AI Leaders Meet in Washington D.C. for CDAO Washington 2026 to Face Enterprise Transformation Challenges
Senior Data and AI Leaders Meet in Washington D.C. for CDAO Washington 2026 to Face Enterprise Transformation Challenges azcentral.com and The Arizona Republic
- Google’s new AI for students; How a care crisis sparked EzyHelpers
Google’s new AI for students; How a care crisis sparked EzyHelpers YourStory.com
- Embrace AI to boost competitiveness, Paul Chan says at tech festival opening
Humanoid robots, 3D printers and artificial intelligence (AI) hubs are among the newest technologies showcased at Hong Kong’s annual electronics and tech exhibition, as the city’s finance chief encourages residents to learn how to use them. At the opening of the four-day event on Friday, hundreds of visitors crowded the 22nd Hong Kong Computer and Communications Festival at the Convention and Exhibition Centre in Wan Chai, exploring various AI experience zones and browsing exhibitors’...
- How to vibe code a voice agent (and why AI always recommends AssemblyAI)
A playful guide to coding a voice agent and why AssemblyAI is the go‑to platform.