Intelligence is Free, Now What? <br> Data Systems for, of, and by Agents
AI inference costs have dropped up to 900x since early 2023, making advanced capabilities accessible at under $0.10 per million tokens. This shift raises questions about the future of agent-based systems and data infrastructure.
- AI inference costs have dropped between 9x and 900x since early 2023, with a median decline of 50x.
- GPT-4-class capabilities now cost under $1 per million tokens, with some providers charging less than $0.10.
- The accessibility of low-cost AI is accelerating the development of autonomous agent-based systems.
- Open-source models are keeping pace with frontier models in terms of cost reductions.
The cost of running AI models has collapsed in recent years. In early 2023, GPT-4-class capabilities cost around $30 per million tokens. Today, the same performance can be achieved for under $1, with some providers offering rates below $0.10. Benchmarks show inference prices have fallen between 9x and 900x annually, with a median decline of about 50x. Even frontier models are becoming significantly cheaper with each new generation, and open-source alternatives are keeping pace.
This dramatic reduction in costs is not just a financial milestone; it signals a fundamental shift in how AI systems are deployed and scaled. The accessibility of advanced intelligence at such low prices is accelerating the development of agent-based systems, autonomous programs that can perform tasks, make decisions, and interact with users or other systems. The implications for data infrastructure are profound, as organizations now face new challenges and opportunities in managing, storing, and processing the outputs of these agents.
The article explores what this new era of "free intelligence" means for the future of AI. It questions whether we are approaching a point where even Nobel-level intelligence could become commoditized, and what that would mean for industries, researchers, and society at large.
Source: Intelligence is Free, Now What? <br> Data Systems for, of, and by Agents. Read the full piece at the source.
Developers can now experiment with advanced AI models at minimal cost, enabling rapid prototyping and innovation.
Businesses can deploy AI at scale without prohibitive infrastructure costs, unlocking new use cases and revenue streams.
Investors should watch for startups and infrastructure providers that leverage these cost reductions to disrupt traditional AI markets.
The democratization of AI intelligence could reshape industries and daily life, making advanced capabilities accessible to all.
- inference
- The process of running a trained AI model to generate predictions or outputs from new input data.
- agent-based systems
- Autonomous programs or entities that can perform tasks, make decisions, and interact with their environment or users.
New artificial intelligence model reveals invisible multiple sclerosis lesions - News-Medical
Grants support new research on role of AI in the arts - Binghamton University
How novice coders can develop AI programs for military applications - MIT News
AI ResearchLiquid AI Open-Sources Antidoom: A Final Token Preference Optimization (FTPO) Method that Reduces Doom Loops in Reasoning Models
Where Medicine and Artificial Intelligence Converge - New York Institute of Technology
Discord admits AI moderation bug wrongfully banned users over harmless images
Discord’s AI moderation system wrongly banned over 8,000 users in two months due to a bug that misclassified harmless images like spreadsheets and chessboards as harmful content.
Meta AI app adds image generation, also powering WhatsApp and Instagram features - 9to5Mac
Meta’s AI app now generates images and powers visual features in WhatsApp and Instagram, expanding its multimodal capabilities.
AI ToolsBuilding a physical API
A developer built a Rust-based system that converts raw CAN bus traffic from a robot into structured telemetry via a physical API, enabling real-time monitoring and control.
Government AI can’t scale — and it’s not the models - Federal News Network
A new report argues that government AI adoption fails not because of model limitations but due to systemic scaling challenges.
Palantir CEO Alex Karp is paranoid about Anthropic and OpenAI. That doesn't mean he has nothing to fear. - Fortune
Palantir's CEO expresses concerns about competition from Anthropic and OpenAI, suggesting the company must adapt to survive in the AI landscape.
BusinessCopilot goes cheap as Microsoft phases out OpenAI and Anthropic models to cut costs
Microsoft is swapping OpenAI and Anthropic models for its own MAI models in Copilot to reduce costs, potentially impacting performance for users.