Conference Presentation
Ori Goshen, Co-founder & Co-CEO of AI21 Labs: AI - All The Work!
- AI21 is releasing its Maestro system immediately to build enterprise-grade knowledge agents that automate mission-critical workflows using model-agnostic systems capable of handling multi-step tasks.
- The platform enables developers to define instructions, requirements, and tools while dynamically optimizing computational budgets based on time or cost constraints.
- Maestro generates multiple execution plans for tasks, selecting the optimal option based on resources and utilizing step-by-step validation to prevent error compounding.
- Enterprise outputs include validated results accompanied by a report card detailing requirement satisfaction and system confidence levels.
- Agents provide full visibility into actions and reasoning through detailed execution graphs, such as tracking sequential web and file searches.
- The system is designed to bridge current LLM accuracy gaps, aiming to achieve consistency and robustness for complex scenarios like support ticket resolution and compliance work.
- AI21 plans to internally utilize Maestro ("dogfooding") to further refine capabilities and address the rigidity of static chaining.
- Companies are expected to replicate the Uplovin model, potentially doubling revenue while maintaining flat headcounts by leveraging AI.
- Significant output reductions through layoffs are predicted in the near term, with analogies drawn to a 30% AI-generated code milestone.
- A "super productivity" effect is anticipated where the gap between ideation and execution approaches zero, driven by an exponential increase in possibilities.
- Even a 10% automation rate is expected to deliver unprecedented value, shifting the narrative from human replacement to empowering humans to solve more problems.
- The economy is forecasted to expand massively as barriers to content creation collapse, potentially generating hundreds of new content formats and services.
- An explosion of individuals and companies is expected to fulfill the pyramid of human needs, mirroring the 64 million new video creators who emerged post-2000.
- Companies automating specific task percentages are expected to pursue previously unconsidered opportunities, transforming the broader business landscape.
- The "knowledge work" sector, currently comprising approximately 76% of the U.S. economy, is expected to continue expanding as AI processes the "six billion tasks" defining all work.
- The industry will transition from binary outcomes of replacing human intelligence to a process where humans "speed run" building things and focus on underserved opportunities.
- Primary corporate aspiration is shifting toward "super productivity" rather than achieving AGI or ASI, acknowledging these concepts as philosophically intriguing but secondary.
- Future growth is driven by the "third angle" of AI impact, focusing on doing more with the same people or more people rather than utopian or dystopian extremes.
- LLM capabilities are expected to evolve from 90% single-step success rates to reliable multi-step task execution without compounding errors.