Archive — US — techcrunch
A new kind of AI model from a ChatGPT inventor is thrilling developers
Diogo Almeida, an OpenAI researcher, has launched TypeSafe AI with Jev, a new transformer-based model that produces calibrated decisions instead of text outputs. Unlike traditional LLMs, Jev is faster and cheaper, appealing to developers for software automation. Users report significant improvements in speed and accuracy, prompting high demand despite its early stage. Almeida envisions a future of widespread smart software driven by the lower cost of intelligence.
Key points
- Diogo Almeida, after leaving OpenAI, created TypeSafe AI to address the limitations of LLMs.
- Jev model focuses on producing probabilities rather than text, making it fast and cost-effective.
- High demand from developers led to capacity issues with TypeSafe's API.
- Software developers reported Jev significantly outperformed LLMs like OpenAI’s ChatGPT in speed and cost efficiency.
- Jev can act as a supervisor for LLMs to prevent errors and improve task routing based on model needs.
Why it matters
- The rise of AI models like Jev reflects a trend towards more specialized AI systems following the LLM boom.
- Companies are focusing on automation, paralleling the increased efficiency trends seen in IT with the integration of AI tools.
- The shift from text-based outputs to decision-making models aligns with emerging debates around efficiency versus creativity in AI development.
- Almeida's approach ties into larger discussions about the AI's role in enhancing automated workflows rather than just conversational capabilities.
Jev represents a pivotal shift in AI capabilities, moving from generating text to making actionable decisions, promising to transform software automation significantly and marking a new chapter in AI application trends.