What is Bertėjas: A Complete Guide 2024



What is Bertėjas?

Bertėjas is an NLP model developed by Anthropic, an AI safety startup. It is described as a conversational agent focused on being helpful, harmless, and honest. 

Bertėjas builds on top of Claude, Anthropic’s first conversational assistant released in 2021. Claude demonstrated an ability to have nuanced, informative conversations while avoiding potential harm.

Bertėjas aims to expand on Claude’s capabilities and be even more responsive in dialog. The key goals are improving language understanding, reasoning, and judgment to carry out helpful conversations.

How Bertėjas Works

Bertėjas utilizes transformer-based neural networks that are pre-trained on massive datasets to understand and generate natural language. Specifically, it employs an extended variant of GPT-3 called Constitutional AI. 

Constitutional AI constrains model behavior during training to align the model incentives with human values. This technique gives Bertėjas greater robustness and safety than a typical large language model.

On top of this foundation, Bertėjas has been trained using human feedback on thousands of conversations. This “self-supervised learning” helps it develop communication skills and conventions expected by users.

Over time, through sustained dialog training, Bertėjas has the flexibility to take on specialized roles like personal assistant, tutor, or financial advisor that can provide customized value to individuals.

Key Capabilities

Here are some key characteristics of Bertėjas:

  1. Helpful by design – Unlike a general language model, Bertėjas aims specifically to provide useful services to human users.
  2. Safety-focused – It employs Constitutional AI and model constraints to greatly improve AI safety.
  3. Dialog expertise – Bertėjas specializes in coherent, responsive conversations that meet users’ needs.
  4. Reasoning capacity – More advanced logical inference and judgment skills compared to previous models. 
  5. Personalization – This can be fine-tuned over time to take on specialized roles and customize expertise. 
  6. Reflective ability – Able to discuss and explain its capabilities, limitations, development needs, etc.
  7. Emerging technology – As an early-stage model, Bertėjas has ample room for improvement in capabilities.
  8. Harmlessness – Architected carefully to avoid potential harm to human users and society.
  9. Honesty – Focused on truthful dialog rather than deceit or manipulation.  
  10. Accessibility – Expected to power next-gen AI assistants available to the masses when complete.

These core attributes set Bertėjas apart as an AI system specialized for cooperative, trustworthy dialog that provides value to people. The combination of state-of-the-art transformer networks and ethics-by-design model development make Bertėjas a promising platform for the future of conversational AI.

BERT vs Bertėjas

As the names are similar, it’s reasonable to compare BERT and Bertėjas. BERT stands for Bidirectional Encoder Representations from Transformers. It is a hugely influential NLP model developed by Google in 2018.

BERT pioneered the transformer architecture for language modeling which Bertėjas also leverages. However, BERT is focused purely on predictive language tasks. As a feature extractor, BERT lacks the conversational and reasoning skills of language models like Bertėjas.

So while Bertėjas depends on advancements like BERT to function, it is optimized specifically for dialog and alignment with human preferences. Bertėjas is trained end-to-end on conversational data rather than static language tasks like BERT.

Potential Applications

Bertėjas aims to power the next generation of helpful, harmless, and honest AI assistants. Some potential applications include:

  • Personal assistants – For scheduling, reminders, recommendations 
  • Intelligent tutors – Adaptive teaching tailored to student needs
  • Customer service agents – Responsive, empathetic interactions 
  • Expert advisors – Financial, medical, legal, etc. guidance 
  • Creative aids – Brainstorming tool for writing, design, innovation
  • Mental health support – Empathetic counselor for therapy sessions

The shared trait across these applications is the need for sustained, coherent dialog driving toward human-centric outcomes. By improving language and reasoning in conversation, Bertėjas can enable personalized AI across many facets of life.


Bertėjas represents an evolution in conversational AI – steering large language models towards benevolence and cooperative dialog. By focusing on safety and service, Bertėjas aims to expand access to helpful AI. As the technology matures, Bertėjas could become an invaluable assistant and problem-solving companion. We expect great progress in this space so stay tuned!


Here are some frequently asked questions about Bertėjas:

What is Bertėjas?

Bertėjas is an artificial intelligence conversational agent created by Anthropic to be helpful, harmless, and honest. It builds on previous work with Claude to have a more responsive dialog capacity.

Who created Bertėjas? 

Bertėjas was created by researchers at Anthropic, a startup focused on AI safety led by Dario Amodei and Daniela Amodei. 

How does Bertėjas work?

Bertėjas uses a transformer-based neural network architecture fine-tuned with a technique called Constitutional AI that better aligns incentives. It has been trained extensively on dialog data to develop communication skills.

Is Bertėjas available to talk to?

Bertėjas is still in development. Anthropic has not yet released a public demo, although Claude is accessible on their website. Release plans for Bertėjas are not yet known.  

What makes Bertėjas different?

Bertėjas prioritizes helpfulness, harmlessness, and honesty in dialog. It also has powerful reasoning and judgment capabilities than previous conversational AI systems like Claude.

What can Bertėjas be used for? 

Target use cases involve personalized assistants, tutors, advisors, counselors, and more. Any domain requiring coherent dialog and human preferences is well-suited.


Leave a Comment