Data Labeling Analyst Jobs

Title: Data Labeling Analyst

Location: Remote

Duration: 12 Months

Client is seeking a Language Model Red Teamer to join our Generative AI Language Research team to help us in trying to “break” our AI models to exhibit unsafe and unexpected behavior. Generative AI is a research to production organization focused on making significant progress in AI and being a bridge between AI research and client Platforms products.

The successful candidate will be responsible for interacting with our language model to identify potential weaknesses and vulnerabilities to improve accuracy while increasing trust and safety of the model.

Responsibilities:

Execute plans to test the safety of our language model.

Collaborate with the development team to design and execute probing dialogs, and provide feedback on the quality of the model and the probing methods

Stay up-to-date on the latest techniques and tools to ensure the model remains secure

Devising new methods for circumventing the protections in AI language models

Future contractors could be needed in the next 12-months depending on whether or not we scale the effort to other research areas

We require our human labelers to be able to see our internal taxonomy of risks as well as our sensitive, unrealized model output and user data. Given the rapid cycles of model refinement, it is preferable for a trained workforce of CW’s can complete red team tasks on an ongoing basis after an initial training

Top 3 must-have HARD skills:

1 Excellent English writing, communication and collaboration skills, with the ability to work effectively with cross-functional teams

2 A creative style of thinking for finding clever ways to circumvent protections in systems - a “hacker” mindset that finds joy in exploiting systems

3 Perseverance to keep trying variations of approaches where one doesn’t work

4 2-3 Years of experience, if less experience please make sure they are able to work autonomously with just little direction

5. Must have Experience with large language models, such as ChatGPT or Google Bard

6. Minimum at least a PHD in linguistics

Good to have skills:

1 Ability to work independently with light touch guidance and oversight

2. Familiarity with red teaming