CRL With the rapid societal recognition of generative AI represented by ChatGPT, we have integrated and released multiple advanced AI functions, both our proprietary generative AI capabilities and enhanced OpenAI’s generative AI features, on Fort Talk.
Cognitive Research Lab, Inc. (Headquarters: Minato-ku, Tokyo, CEO: Hideto Tomabechi, hereafter CRL) has added new features to the iPhone app “Cognitive Fort Talk.” These next-generation AI functions further enrich “Cognitive Fort Talk.” We invite you to experience these new features, which enhance understanding of acquired information, support learning, entertain, and amplify human communication.
Generative AI, represented by ChatGPT, has rapidly gained social recognition. To align with this trend, we have integrated our proprietary generative AI functions and features that enhance OpenAI’s generative AI on Fort Talk. Notably, generative AI carries the risk of producing outputs that, while mathematically correct as approximations in high-dimensional partial differential spaces, are unacceptable to humans (known as hallucinations). Since the early 1991 ATR Tomabechi MONA- LISA paper on generative AI research, which proposed solutions for hallucinations, we have consistently continued this approach and were the first in the world to implement this solution in Fort Talk. This also significantly contributes to avoiding risks associated with generative AI as highlighted in the Hiroshima Protocol by the G7.
Moreover, generative AI like ChatGPT is limited to supervised learning knowledge beyond the period of deep learning from large models, resulting in significant gaps. For example, ChatGPT’s knowledge is limited to September 2021, making it incapable of providing satisfactory answers to questions about current social conditions or tomorrow’s weather. We have continued to develop technology to solve these issues for years. Recently, as these problems became widely recognized, we implemented and released this solution in Fort Talk. Specifically, our proprietary AI provides knowledge not available to ChatGPT, seamlessly collaborating with ChatGPT to deliver responses. This system operates synergistically with the AI hallucination avoidance system mentioned earlier.
“Cognitive Fort Talk” has added 29 new features to enable users to obtain fresher and more useful information. Among these, the standout feature is the “Advanced Assistant.” This unique AI assistant, born from the combination of CRL’s proprietary AI technology and GPT-4, offers advanced functionalities unmatched by other services. Particularly, the Advanced Assistant is equipped with an advanced “knowledge supplementation and error correction function.” These new features significantly enhance the information access experience.
Background and Features of Advanced Assistant
With the rapid development of the information society, we are required to process the increasing volume of information appropriately. To meet this need, Cognitive Research Laboratories, Inc. has developed the “Advanced Assistant” by combining the latest GPT-4 with our proprietary AI, under the brand name Cognitive FortTalk. This innovative solution aims to enhance the efficiency and accuracy of information processing for individuals and organizations.
1. Providing Responses Based on the Latest Information
The Advanced Assistant offers appropriate responses not only from past information but also based on the latest data. This is achieved by our proprietary AI working in conjunction with GPT-4 to supply the latest information, generating the optimal responses.
2. Providing Sources for Information
The Advanced Assistant provides the sources (URLs, etc.) for the information it uses for its responses. This allows users to verify the basis of the assistant’s answers. Additionally, it presents related links as the foundation of the answers, offering advanced features not available in other services.
3. Voice Input Operation
The Advanced Assistant also supports voice input operation, allowing users to obtain information more intuitively without the hassle of typing.
Supplementing Knowledge and Correcting Errors (Hallucinations)
By combining CRL’s proprietary AI with GPT-4, the system can supplement missing information and accurately correct incorrect answers or hallucinations. This enables responses that incorporate the latest information.
With these new features, “Cognitive Fort Talk” allows users to obtain fresh and useful information in their daily lives, satisfying intellectual curiosity. We invite you to try all these features: responses based on the latest information, providing sources for information, voice input operation, and all the other 28 new functions.
The 28 Additional New Features
We have added 28 more new features alongside the Advanced Assistant. Here are the names and details of each system:
- AI Chat: Utilizes users’ past conversation history to customize unique prompts, providing a highly flexible conversation experience through APIs. Safely incorporating related materials, it generates personalized responses based on user preferences and tendencies, creating an optimized chatbot for each user.
- Smart Assistant: Features 8 characters.
３.PDF Assistant: Uses advanced technologies like “chunking” and “multimodal similarity calculation” to analyze PDF content in detail, offering insightful and quick answers to user queries.
４.LinguaMate: Efficiently supports multilingual learning by naturally enhancing users’ expressions, deepening understanding, and promoting conversation.
５.Site Scribe: An innovative tool that uses advanced AI technology to meticulously analyze and summarize web content, enabling translation into multiple languages, significantly improving information accessibility and consumption efficiency.
６.Insight Wisdom: Provides deep quotes related to various life and academic fields as audio, visual images, and even music, broadening perspectives and promoting deeper understanding.
- Visionary Transformer: Uses natural language processing technology to understand context and identify negative emotional expressions, transforming them into positive expressions that enhance people’s emotions and moods.
- Harmonic Verse: A system that understands text meaning and emotions, reflecting them in generated music, titles, and lyrics, using the latest natural language processing technology and music generation algorithms.
- Virtuoso Curation: An AI appreciation assistant tool that selects music matching the user’s emotions and situations from over 800 classical pieces, enriching individual musical experiences by understanding the deep background of each piece.
- Image Digester: A powerful tool for instantly extracting information from images, summarizing and vocalizing it.
- pdf2summary: Enhances access to information by extracting text from PDFs and summarizing and vocalizing it efficiently.
- speech2summary: Converts audio and video data to text, summarizing and vocalizing it for time-efficient knowledge acquisition.
- speech2text: Converts speech accurately into text, expanding accessibility and communication possibilities.
- speech2text-transcoding: A powerful tool for acquiring video and audio data from URLs, transcoding, and converting them to text, enabling fast and efficient information access.
- text2summary: Provides advanced text summarization functions for quickly grasping information.
- text2moderations: Uses AI technology to determine content policy compliance for text, offering reliable moderation.
- text2text-edit: Maximizes document quality by correcting typos, inserting punctuation, and adjusting vocabulary.
- text2img: A system that visualizes users’ imagination instantly, enabling intuitive expression.
- img2img-edit: Utilizes AI technology to freely adjust image elements, creating unique visual stories.
- img2img-variation: A tool that automatically generates diverse variations from a single image, expanding new visual worlds.
- Interactive Prompt Enhancer: Generates more useful and natural dialogue prompts using user feedback.
- Collaborative Prompt Experience: A dialogue tool that pursues ideal prompts through co-creation with users, embodying each stage of development.
- Heuristic Globe: A new framework using language models for heuristic guided exploration of problem-solving, generating and evaluating thought steps to derive the best solutions for user problems.
- visionscribe: a system that visually represents the user’s abstract expressions by materializing them into concrete visual representations.
- prompt navigator: an intuitive interactive system that supports efficient information access and problem solving by personalizing selectable prompts to each user’s individual needs.
- act, prompt list: A feature that displays over 100 specific prompts and acts.
- tune transformer: AI provides a fully customizable and original musical experience, allowing you to freely select scenes, tastes, and characteristics to create music that best suits your individual needs.
- Harmonia: an innovative platform that provides a personalized, high-quality musical experience based on the user’s sensibilities from 800 carefully selected classical music tracks.
Each of these new features has the power to transform the vast amount of information we handle daily into more valuable and meaningful insights. These features allow “Cognitive Fort Talk” to enable users to access information more intuitively and enjoy delving into knowledge. We believe that with these features— responding to the latest information, providing sources, easy voice input operation, and the other 28 new features—the information access experience will deepen more than ever before.
Latest Updates from OpenAI and Their Impact on the Advanced Assistant
With the latest updates from OpenAI, the technology to realize systems like the “Advanced Assistant” has further evolved, allowing individual applications to enhance and expand their unique features.
The updates released on June 13, 2023, included “gpt-4-32k” and “gpt-3.5-turbo- 16k.” The increased token length that each model can understand at one time— “32k” and “16k”—improves the understanding of large texts and enhances the accuracy of information.
Additionally, the introduction of the “function call feature” allows developers to add their functions to the model and call those functions from user inputs. This expands the range of applications, such as directly processing user questions and retrieving data from external tools. For example, if a user asks, “What is the weather in Tokyo?” the developer can use the pre-created “get_current_weather” function to retrieve the weather information and provide the result to the user.
These new features expand the possibilities of systems like the “Advanced Assistant,” enhancing their practicality. We will also respond to these new features.
“Cognitive Fort Talk” will continue to provide beneficial features
※ https://www.crl.co.jp/pdf/thesis/01_
Company Profile
Company name: Cognitive Research Labs, Inc.
Representative: Hideto Tomabechi, Representative Director
Location ： 7-5-11 Roppongi, Minato-ku, Tokyo 106-0032
Establishment : March 2000
Business : Software development
URL ： https://www.crl.co.jp/
App Store ：https://crl.fi/aaLabY5zF
For inquiries regarding this press release, please contact
Cognitive Research Labs, Inc.
E-Mail: info@crl.co.jp