Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
Nobel economist predicts AI will grow GDP by only 1.5 percent - OpenSmartRoute
Nobel economist predicts AI will grow GDP by only 1.5 percent
Daron Acemoglu says AI adds just 1.5 percent to global GDP over ten years. He believes human adaptation limits productivity gains more than model size.
Key points
Acemoglu predicts AI boosts global GDP by only 1.5 percent over a decade.
He estimates AI will replace at most five percent of current jobs.
The bottleneck is human nature, not the speed of model development.
Easy-to-deploy apps matter more than bigger models for productivity.
Why it matters: This forecast suggests companies should focus on small app upgrades instead of massive automation bets.
By OpenSmartRoute editorial · written through the router by writer-small
From The Decoder - “Microsoft publishes Nobel economist's bearish AI forecast of just 1.5% GDP growth over a decade”
Microsoft released a new economic forecast about artificial intelligence. Nobel Prize-winning economist Daron Acemoglu wrote this piece. Microsoft published it on its blog called The Humanist Review of AI. Authors sign their pieces by hand in this specific publication. This is an unusual way for a major tech company to share research. The post argues that current excitement about AI might be too high. It suggests the real world will move slower than predicted by labs.
Daron Acemoglu is one of the most famous economists in history. He won the Nobel Prize for his work on economics and politics. His latest analysis focuses on how artificial intelligence affects global wealth. He looks at Gross Domestic Product, which measures total economic output. This number tells us how much value a country creates every year. Acemoglu believes AI will not change this number as much as others think.
He predicts that AI will add only 1.5 percent to GDP over ten years. This is a very small increase compared to past technological shifts. Some researchers claim AI could double productivity in the same time. Acemoglu disagrees with those optimistic numbers significantly. He thinks human limits matter more than computer power alone.
The article also discusses job displacement rates. Acemoglu estimates that at most five percent of jobs will be replaced. This figure covers all industries and all types of work. It is much lower than some AI safety groups fear. Others predict millions of workers could lose their roles quickly. His number suggests a slower, more gradual change in the labor market.
Acemoglu admits the pace of AI development is hard to predict. Technology moves fast and creates new tools constantly. He knows it is difficult to forecast such rapid changes. Still, he believes his economic model accounts for real-world friction. Companies cannot simply switch on robots and expect instant results.
The bottleneck, Acemoglu says, is human nature. People need time to learn new ways of working. Companies have to reassign tasks among their employees. They must upskill workers to handle new tools. This process requires significant effort and planning from management. It takes longer than people often realize in the rush to adopt AI.
This restructuring phase could drag on longer than electrification did. The electric car revolution took decades to fully transform transportation. Acemoglu compares the current AI wave to that historical shift. Both require changing how humans interact with their environment. Neither can happen overnight without human adaptation.
OpenAI plans to dump hundreds of AI-solved math problems on GitHub without publishing papers. Mathematicians want formal verification and proper credit before accepting the results.
Mirror Particle raises capital to create an AI engine that simulates changing human motivations. The company rejects large language models in favor of a foundation model trained on longitudinal data.
Bigger models won't fix that specific problem he describes. Increasing the size of an artificial intelligence system does not solve organizational issues. A larger model cannot force a company to retrain its staff. The software is only as good as the people who use it daily. Productivity depends on how well humans and machines work together.
What's missing are easy-to-deploy apps that change how things get made. Acemoglu argues for tools that integrate smoothly into existing workflows. These applications should enhance human skills rather than replace them entirely. Full automation often fails because of last-mile problems. The final steps in a process require human judgment and care.
Even 99 percent accuracy often isn't enough once you factor in real user needs. A system that gets things right most of the time still makes mistakes. These small errors can cause big problems in critical tasks. Users need reliability, not just high statistical scores. They want tools they can trust for daily work.
For Microsoft, that's a convenient thesis to follow. The company can bolt AI onto existing products without betting on large-scale automation. This approach allows them to sell upgrades to current customers. It avoids the risk of building entirely new automated systems. Companies prefer incremental improvements over risky overhauls.
The article ran in Microsoft's eccentric corporate blog "The Humanist Review of AI." Authors sign their pieces by hand in this specific publication. This adds a personal touch to what is usually cold technical writing. It suggests a focus on human values alongside technological progress. The blog aims to balance innovation with ethical considerations.
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.
Engineers who run models will see this as a warning about hype. Many projects fail because they ignore human factors in the design. Teams focus on model performance but neglect user training. They build complex systems that no one knows how to use. The result is low adoption rates despite impressive technical specs.
Managers who decide what to buy will find this useful for budgeting. Large automation projects cost more than expected when you count training time. Companies often underestimate the cost of changing people's habits. Acemoglu's forecast helps them plan for a slower rollout. It encourages investing in change management alongside software development.
The specific numbers in the GDP and job replacement predictions matter for planning. A 1.5 percent GDP increase sounds small but represents trillions of dollars globally. Over ten years, this adds up to significant economic growth or stagnation. Policymakers need accurate data to make decisions about infrastructure and education. They cannot rely on optimistic projections from tech conferences alone.
Acemoglu's argument that human nature is the real bottleneck challenges the current narrative. Most AI discussions focus on computing power and algorithmic efficiency. Fewer people talk about organizational culture and worker readiness. This shift in perspective changes how we approach AI deployment strategies. It moves the conversation from technology to society.
Why bigger models won't fix the productivity problem he describes is clear when looking at real cases. A massive model cannot automate a task that requires complex social negotiation. Salespeople, teachers, and therapists need human empathy and adaptability. These roles are hard to automate even with advanced AI systems. The bottleneck remains in the human element of work.
The role of easy-to-deploy apps versus full automation is key for business success. Apps that sit on top of existing software are easier to adopt. They do not require a complete overhaul of company processes. Full automation disrupts workflows and often causes resistance from staff. It is harder to implement and manage than simple add-ons.
Why it matters for engineers building agents and managers buying tools comes down to cost and speed. Engineers waste time building features that users will never adopt. Managers lose money on projects that take too long to deploy. Both groups need to respect the human adaptation timeline Acemoglu describes. Speed gains come from better integration, not bigger models alone.
What to do when running models with limited deployment budgets starts with choosing the right tools. Do not chase the largest model if it does not fit your workflow. Look for applications that solve specific problems without requiring massive retraining. Test small changes before committing large resources to a project. Measure adoption rates alongside performance metrics to guide decisions.
Check how easy it is to integrate new AI features into your current stack. Ask your team how much training they need to use the tool effectively. Plan for a six-month rollout period instead of expecting immediate results. Budget for change management consultants if your organization is large. Remember that human nature sets the pace, not the code.
Compare different vendors based on their ease of deployment rather than just raw power. Some companies offer better support and documentation which speeds up adoption. Read case studies to see how others handled similar integration challenges. Look for partners who understand both technology and human behavior.
Try a pilot program with a small group of users before scaling up. Gather feedback on usability and training needs during the trial phase. Adjust your approach based on what people actually say they need. Do not assume that technical superiority translates to business value automatically.
Remember that electrification took decades to transform an entire industry. Your AI transformation will likely take just as long if you ignore human factors. Plan accordingly with realistic timelines and budgets for training and support. Focus on tools that extend human skills rather than replacing them entirely.
Announcement - Microsoft publishes Daron Acemoglu's bearish AI economic forecast
Microsoft released a new report from Nobel economist Daron Acemoglu. He predicts artificial intelligence will grow global GDP by only 1.5 percent over ten years. This number is much lower than many tech companies claim. The report appeared on Microsoft's blog called "The Humanist Review of AI." Authors there sign their names by hand in the text. This blog style feels different from standard corporate news sites.
The specific numbers in the GDP and job replacement predictions
Acemoglu estimates that AI will add 1.5 percent to global economic output. He believes this growth happens slowly over a full decade. Some other experts predict much faster gains from the technology. They expect AI to replace up to five percent of all jobs. This figure is also far below what some startups say. The gap between predictions shows how hard it is to guess the future.
Acemoglu's argument that human nature is the real bottleneck
The main problem, Acemoglu says, lies in human behavior. Companies must reassign tasks and upskill workers before gains appear. This restructuring process takes a long time. It could take longer than the electrification of industries did. People need time to learn new ways of working. Resistance from employees slows down productivity improvements significantly.
Why bigger models won't fix the productivity problem he describes
Larger artificial intelligence models cannot solve this human bottleneck. Even if a model is 99 percent accurate, it still fails in practice. Real users face last-mile problems that confuse even smart systems. These are small details at the end of a process. They matter more than raw accuracy numbers for daily work. Better integration beats bigger models alone every time.
The role of easy-to-deploy apps versus full automation
Simple applications sit on top of existing software easily. They do not require companies to overhaul their entire workflow. Full automation disrupts current processes and causes staff resistance. Implementing such systems is harder than adding simple tools. Businesses prefer apps that fit into current routines quickly.
Why it matters for engineers building agents and managers buying tools
Engineers waste time building features nobody uses. Managers lose money on projects that take too long to launch. Both groups must respect the human adaptation timeline Acemoglu describes. Speed comes from better integration, not just larger models. Ignoring human factors delays success and increases costs unnecessarily.
How OpenSmartRoute helps
A team using OpenSmartRoute gains speed when Acemoglu says human adaptation limits gains. The router sends requests to models that fit specific needs rather than one big model for everything. This setup avoids paying for a single large model on every task.
The system scores candidates on quality, cost, speed and safety before routing. Easy-to-deploy apps matter more than bigger models for productivity, so the router optimizes for quick wins. Teams can set weights per request to prioritize what their business values most.
Hard rules ensure personal data stays on-premises while respecting region or cost caps. An input guard spots prompt injection before requests leave the network. A savings ledger shows how much each routed request saved compared to the most expensive option.
What to do when running models with limited deployment budgets
Start by choosing the right tools for your specific needs. Do not chase the largest model if it does not fit your workflow. Look for apps that solve specific problems without massive retraining. Test small changes before committing large resources to a project. Measure adoption rates alongside performance metrics to guide decisions.
Check how easy it is to integrate new AI features into your current stack. Ask your team how much training they need to use the tool effectively. Plan for a six-month rollout period instead of expecting immediate results. Budget for change management consultants if your organization is large. Remember that human nature sets the pace, not the code.
Compare different vendors based on their ease of deployment rather than just raw power. Some companies offer better support and documentation which speeds up adoption. Read case studies to see how others handled similar integration challenges. Look for partners who understand both technology and human behavior.
Try a pilot program with a small group of users before scaling up. Gather feedback on usability and training needs during the trial phase. Adjust your approach based on what people actually say they need. Do not assume that technical superiority translates to business value automatically.
Remember that electrification took decades to transform an entire industry. Your AI transformation will likely take just as long if you ignore human factors. Plan accordingly with realistic timelines and budgets for training and support. Focus on tools that extend human skills rather than replacing them entirely.
Insurers prepare for massive payouts as autonomous AI agents cause damage. Personal liability for executives like Sam Altman and Dario Amodei is now discussed.
