Conjecture

Conjecture is an AI safety research startup founded in 2022 by Connor Leahy, Sid Black, and Gabriel Alfour. Conjecture is working to address the "AI alignment problem", the potential for powerful AI systems to misalign with human values and goals. They believe that the current AI paradigm, focused on scaling large language models (LLMs), is fundamentally unsafe and are developing an alternative approach called Cognitive Emulation (CoEm).

CoEm aims to build AI systems that emulate human reasoning processes, making them more interpretable and controllable. Conjecture's research focuses on developing AI macros that are trained for specific tasks and can be combined to solve more complex problems. They prioritize transparency, aiming to build AI systems that show their workings, enabling users to understand and correct their outputs. Conjecture is based in London and has received funding from prominent investors in the AI field.

Conjecture offers a hybrid work model where employees work together in person four days a week in their London office, with the option to work remotely on the remaining day. The company sponsors visas and offers relocation support, along with generous benefits.

Conjecture offers a comprehensive benefits package, including visa sponsorship and relocation support for employees. They also provide generous benefits and allow employees to work together four days a week in person at their sunny office in central London.

Conjecture is dedicated to ensuring the safe and controllable development of advanced AI technology by building AI systems that are designed to follow the same trusted reasoning processes as humans. They believe that the current AI paradigm, with its focus on scaling black-box models, is too risky and that Cognitive Emulation, their proposed alternative, offers a safer path to powerful and reliable AI. They are committed to building the infrastructure and systems that enable a future where AI benefits humanity without posing existential risks.

Conjecture fosters a culture of intellectual curiosity, collaboration, and dedication to building safe AI. The company emphasizes transparency, both internally with employees through an open information hazard policy and externally through a public blog and social media presence. They value rapid iteration and experimentation, encouraging their team to take risks and learn from their mistakes. Conjecture prioritizes a collaborative approach, encouraging open communication and knowledge sharing.

Conjecture does not explicitly discuss their approach to diversity, equity, and inclusion on their website.