AGENTIC-SYSTEMSJuly 15, 20261 min read

Optimization Is Not All You Need

A critical examination of the 'optimization culture' in AI development, arguing that while optimization can measure the improbability of text, it cannot distinguish between creative invention and error.

Computer Science > Artificial Intelligence

Title:Optimization Is Not All You Need

View PDF HTML (experimental)Abstract:In 2019, OpenAI released two million GPT-2 outputs-ungrammatical, half broken-to aid the detection of machine-generated text. The alignment that produced their more fluent successors is usually regarded as an engineering achievement; we read it instead as the newest expression of optimization culture: the conviction, older than the technology, that measurable improvement along predefined axes exhausts the question of value. Tracing that conviction through the stack-pretraining, decoding, preference tuning, benchmarking, interface-and back through its genealogy in the audit society, we arrive at the limit: an optimization procedure can measure how improbable a piece of generated text is; it cannot tell whether that unlikelihood is error or invention. A procedure that cannot make that distinction has nonetheless, within half a decade, assumed the authority to set the protocols of legitimate language. Held for centuries by academies and schoolrooms, grammars and examiners, this authority has been given over to loss functions, reward models, benchmarks, and system prompts: an apparatus that executes the office of judgment with no capacity for judging.

Bibliographic and Citation Tools

Code, Data and Media Associated with this Article

Demos

Recommenders and Search Tools

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Source: arXiv cs.AI Recent

More in this category

agentic-systems

Good Benchmarks

As AI models advance rapidly, establishing high-quality benchmarks is more critical than ever. A recent paper outlines the core criteria that define a truly effective and practical AI evaluation system.

agentic-systems

Cost-Governed RAG: Unified Per-Tenant Cost Attribution Across Retrieval and Generation in Multi-Tenant LLM Systems

Researchers have introduced Cost-Governed RAG, a novel architecture that enables precise, per-tenant cost attribution across both retrieval and generation layers in multi-tenant LLM systems. By leveraging a codebook-oblivious vector index called TurboVec, the system achieves 99.96% cost attribution accuracy while reducing infrastructure costs by up to 9 times.

agentic-systems

The Emerging Paradigm of Geospatial Foundation Models: From Pre-Training to Agentic Reasoning

This paper explores the paradigm shift enabled by Geospatial Foundation Models (GeoFMs), detailing their types, operationalization strategies, and the future of Agentic Geospatial Reasoning.

NOW LET US Related – Operationalising Multi-Dimensional Evaluation for Conversational Agents: A Scalable, Governed Pipeline with Selective Re-evaluation and Model Benchmarking

agentic-systems

Operationalising Multi-Dimensional Evaluation for Conversational Agents: A Scalable, Governed Pipeline with Selective Re-evaluation and Model Benchmarking

This paper presents GenAI Evaluation, a governed, configuration-driven pipeline designed for large-scale, multi-dimensional evaluation of retail conversational agents. The framework processes approximately 50,000 daily records, achieving high alignment with human judgment across key metrics like helpfulness, truthfulness, and tone.

agentic-systems

Representing and Generating Levels Over Time through Playtrace Reconstructive Partitioning

Researchers introduce a novel 'cake' representation and the Playtrace Reconstructive Partitioning (PRP) algorithm, allowing AI to generate game levels by encoding dynamic playtrace information over time. Tested on Sokoban, this approach outperforms existing PCG methods while preserving solution diversity.

agentic-systems

Connected by Construction: Learning Tractable Near-Tour Marginals for Traveling Salesman Problems

Researchers have introduced C2TSP, a breakthrough unsupervised learning pipeline for the Traveling Salesman Problem (TSP). By focusing on learning structurally meaningful latent objects rather than relying solely on the final decoding stage, C2TSP opens up new possibilities for AI-driven combinatorial optimization.