The Rise of Cheap Intelligence: What Jev’s Success Means for the Future of Automation

The recent release of TypeSafe AI’s Jev model has sent shockwaves through the developer community, with many hailing it as a game-changer in software automation. This new transformer-based model produces probabilities and calibrated decisions that inform intelligent behavior in software systems, eschewing language altogether.

The significance of Jev extends far beyond its immediate applications. As Diogo Almeida, TypeSafe AI’s founder, notes, the falling cost of intelligence will lead to widespread deployment – and Jev is at the forefront of this trend. Almeida’s previous work on reinforcement learning from human feedback was instrumental in creating chatbots that dominate our current AI landscape.

However, he became frustrated with the limitations of these models, which often sacrifice automation for human language optimization. “We have lightning in a bottle,” he said, “but it is not useful.” Jev addresses this problem by producing calibrated decisions rather than text output, eliminating the risk of hallucination and allowing users to define outputs in advance.

The Hallucination Problem

Almeida’s experience with chatbots highlighted the trade-off between language optimization and automation. These models produce human-like responses but often lack practical application. Jev’s design shifts this focus by prioritizing calibrated decisions over text output, making it an attractive solution for developers seeking to create sophisticated automation systems.

The Future of Automation

Jev’s success has far-reaching implications for software development. Almeida envisions a future where smart software is ubiquitous, with intelligence deployed in a distributed and emergent way – similar to the early internet. This would enable developers to create complex automation systems without breaking the bank or relying on expensive large language models.

Moreover, Jev’s low cost and speed make it an ideal tool for real-time workload sorting and routing, revolutionizing software development and deployment. Its potential applications extend beyond software development, with possible uses in finance, healthcare, and transportation where automation is increasingly crucial.

The Rise of System One Models

Almeida’s decision to focus on intuition rather than reasoning in Jev’s design has raised questions among observers. TypeSafe AI refers to Jev as a “System One model,” but experts speculate that it may be built on top of an open-weight large language model. Almeida’s emphasis on synthetic data and calibrated decisions suggests a fundamental shift in how we approach AI development.

The Emergence of New Modalities

TypeSafe AI has announced plans to build more versions of Jev in new modalities, which could accelerate the adoption of cheap intelligence. However, this may lead to competition and innovation – and potentially even fear or hype – as companies seek to capitalize on the trend.

The Real-World Impact

Jev’s impact will be felt beyond the developer community, with far-reaching implications for industries that rely heavily on automation. As Almeida hopes, widespread deployment of Jev could lead to the emergence of intelligent systems that are both distributed and ubiquitous – but it also raises questions about accountability and responsibility in software development.

The growing reliance on cheap intelligence creates uncertainty about who will be accountable for its applications. Almeida notes wryly, “the main product of frontier labs is fear or hype” – a cautionary note as we navigate the implications of this new technology.