Skip to content
TheChatArchiveThe Chat Archive
Perplexitysonar-reasoning-pro-high· PerplexityHugging Face LMArena Search Arena 24K

what have Noam Shazeer did (esp research or innvoation) during the character ai period?

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Geminigemini-2.0-flash-grounding’s answer.
  1. User

    what have Noam Shazeer did (esp research or innvoation) during the character ai period?

  2. Perplexity Assistant

    Noam Shazeer's tenure at Character.AI (2021-2024) was marked by significant innovations in conversational AI, though his work during this period was more focused on product development and consumer applications rather than publishing academic research as he had done at Google.

    Founding and Development of Character.AI

    After leaving Google in 2021, Shazeer co-founded Character.AI with Daniel De Freitas, a former colleague[1][6]. The departure was reportedly motivated by frustration over Google's reluctance to release a chatbot he had developed, leading him to pursue his vision independently[6].

    This decision proved successful, as Character.AI quickly gained popularity:

    • The platform attracted 3.5 million users who engaged with AI characters for approximately two hours daily on average[4]
    • Within a year, Character.AI secured 150millioninfundingandreacheda150 million in funding and reached a 1 billion valuation[5]
    • By 2023, over 18 million characters had been created on the platform[4]

    Technical Innovations at Character.AI

    While at Character.AI, Shazeer focused on several key technical areas:

    Advanced Conversational AI

    Shazeer's work centered on developing AI systems that could maintain engaging, personalized conversations. Unlike many existing AI assistants, Character.AI allowed users to interact with AI personalities that could portray specific characters, from historical figures to fictional entities[4].

    Memory and Personalization

    One of Shazeer's primary research focuses at Character.AI was improving memory capabilities for AI systems. As he stated: "Some of the things we wanna work on the most are memory cause our users definitely want their virtual friends to remember them. There's so much you can do with personalization and you wanna dump in a lot of data and use it efficiently"[8].

    Voice and Multimodal Interaction

    Under Shazeer's leadership, Character.AI developed several innovative features:

    • Character Calls: A feature allowing users to have voice conversations with AI characters, supporting multiple languages including English, Spanish, Russian, and Chinese[6]
    • Character Voice: A text-to-speech system that gave AI characters distinct voices, with over 1 million user-created voices in the library[6]

    Consumer-AGI Approach

    Shazeer positioned Character.AI as a "consumer-AGI" company, focusing on making advanced AI accessible to everyday users[3]. His vision was to create versatile AI systems that could solve "millions and billions of different problems" while remaining highly usable for non-technical users[1].

    Business and Strategic Direction

    Rather than pursuing purely academic research, Shazeer took a more product-driven approach at Character.AI:

    • He emphasized getting "technology out there to billions of users" as his primary motivation for leaving Google[4]
    • Shazeer focused on creating AI that could provide value across numerous use cases, stating: "We believe the best use cases are things we haven't thought of yet. Our job is to get something very versatile and usable in the hands of billions of people"[4]
    • He worked on scaling large language models for conversational AI while addressing practical challenges like maintaining engaging interactions with reasonable computational resources[1]

    In August 2024, Shazeer and approximately 30 key team members returned to Google as part of a $2.7 billion deal, where he now leads AI efforts at DeepMind, particularly on the Gemini initiative[6][9].

  3. User

    have he published anything during that period

  4. Perplexity Assistant

    Based on the search results, there is no indication that Noam Shazeer published academic research papers during his tenure at Character.AI (2021-2024). His focus during this period appears to have shifted from academic publishing to product development and commercialization of AI technology.

    Publication History

    Looking at Shazeer's Google Scholar profile provided in the search results, his most recent significant publication mentioned is the PaLM paper from 2023 titled "PaLM: Scaling language modeling with pathways"[11]. However, this work was likely completed during his time at Google or in collaboration with Google researchers rather than representing research done at Character.AI.

    Focus at Character.AI

    Instead of academic publishing, Shazeer's work at Character.AI centered on:

    1. Building and scaling a consumer-facing AI platform that quickly attracted 3.5 million users who engaged with AI characters for approximately two hours daily[4][8]

    2. Developing product innovations such as:

      • Character Calls: A feature allowing voice conversations with AI characters in multiple languages including English, Spanish, Russian, and Chinese[6][7]
      • Character Voice: A text-to-speech system that gave AI characters distinct voices, with over 1 million user-created voices in the library[6][7]
    3. Improving memory capabilities for AI systems, as he stated: "Some of the things we wanna work on the most are memory cause our users definitely want their virtual friends to remember them. There's so much you can do with personalization"[8]

    4. Implementing his "consumer-AGI" approach focused on making advanced AI accessible to everyday users rather than pursuing purely academic research[3]

    During interviews in this period, Shazeer emphasized practical applications over theoretical research, stating: "We believe the best use cases are things we haven't thought of yet. Our job is to get something very versatile and usable in the hands of billions of people"[4][8].

    This shift from academic research to commercial product development aligns with his stated motivation for leaving Google: getting "technology out there to billions of users"[4].

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.