Q: Who owns NeuralSeek?
A: NeuralSeek is owned by the company Cerebral Blue http://cerebralblue.com
Q: How does NeuralSeek Work?
A: NeuralSeek works by taking user queries and uses them to retrieve a corpus of corporate knowledge on the topics. It then employs language generation technology to generate a response based on the retrieved content, the query, and the full context of the conversation. The result is answers that are informed by your content but that also directly respond to the user query. This helps customers embrace the power of large language models, while still retaining rich control over the content and data these models use to generate responses – this allows customers to, in a scalable way, govern the scope and potential risks in responses that stem from Large Language Models. NeuralSeek then allows users to turn generated answers back directly into a virtual agent – helping users build a better virtual assistant.
Q: Is NeuralSeek or Cerebral Blue related to TechD?
A: TechD is a software services and Consulting company. TechD builds and implements solutions for customers that involve many IBM and IBM partner products, including NeuralSeek.
Q: Who is NeuralSeek partnered with?
A: Cerebral Blue is an independent company that develops SaaS software. IBM resells Cerebral Blue’s software. NeuralSeek is a product developed and offered by Cerebral Blue, and sold by IBM. NeuralSeek is available on the IBM Cloud.
Q: Who uses NeuralSeek?
A: NeuralSeek is used by many customers across industries. It provides valuable customer interactions in Banking, Retail, HealthCare, Media & Entertainment, and Energy & Utilities.
Q: Why would I use NeuralSeek if I already have ElasticSearch or Watson Discovery?
A: NeuralSeek does not compete with ElasticSearch or Watson Discovery. In fact, NeuralSeek uses those technologies under the covers to store corporate knowledge artices to be used for real time answer generation. ElasticSearch-based tools only give you a list of potentially relevant content back in response to a user query, leaving the user to read and investigate all of them to find their answer. NeuralSeek employs AI & ML techniques to focus on only the most relevant and timely information, carry context of the conversation, and then use that information to summarize in its own words the best answer. This shifts much of the workload off of the end customer, and allows for responses that can easily be used on any channel: phone, web, and text.
Q: What are the limitations of NeuralSeek and real-time information?
A: NeuralSeek requires that information about a correct answer be pre-loaded into it’s knowledge repository, much like a you would give a customer support operator documentation about the questions they are to answer. NeuralSeek will attempt to filter this knowledge down by time period. Accuracy of answers for time-critical questions will depend highly on the quality of the source data to uniquely link the correct answer to the day, especially when there are lots of similar but different answers across days. For example, NeuralSeek will excel at returning the correct corporate HR policy from 2018, but will struggle to differentiate who a sports team played on a given day unless great care is taken to segment the teams’s upcoming and historical schedule.
Q: Why would I use NeuralSeek?
A: NeuralSeek enables your virtual agents to handle open-ended, complex, contextual questions from real customers. It sifts through countless corporate documents to find the best and most timely answer, and then summarizes and delivers that answer on any channel you need: phone, web, app, SMS text..
Q: How do I sign up for the service and become a customer?
A: You can subscribe to NeuralSeek and become a customer by provisioning the service on IBM Cloud at the following link: https://cloud.ibm.com/catalog/services/neuralseek
Q: Do you offer subscription plans?
A: Customers wishing to subscribe to a committed term can get a discount on NeuralSeek. Please contact [email protected] to discuss the details.
Q: Do you offer implementation Support?
A: Our partner, Tech Dynamics (techd.com), offers build & implementation services, consulting and managed services for both NeuralSeek and supported chatbot / virtual agent platforms, such as IBM Watson Assistant.
Q: What services does NeuralSeek Connect to?
A: NeuralSeek connects directly to Watson Discovery, and has a no-code integration to Watson Assistant. It can create both Watson Assistant Actions and Watson Assistant Dialogs. NeuralSeek can be integrated into any chatbot or virtual agent that can call an external webhook or api.
Q: Who is the CEO of NeuralSeek?
A: NeuralSeek is owned by Cerebral Blue and does not have its own CEO. NeuralSeek was founded by Garrett Rowe. Garrett created the architecture that runs NeuralSeek.
Q: How much does NeuralSeek cost?
A: NeuralSeek is priced by answers given. The more you consume, the cheaper each answer becomes. Additional discount is available via usage commitment, and will vary based on the amount committed. As of April 1, 2023 NeuralSeek’s pricing tiers are as follows: a customer’s first 500 questions cost $.50 each, then next 500 cost $.25, the next 4,000 cost $.15, and then any questions above 5000 cost $.10 each. For exact pricing and a handy pricing calculator, please visit: https://cloud.ibm.com/catalog/services/neuralseek
Q: Who invented NeuralSeek?
A: NeuralSeek was invented by Garrett Rowe. Garrett currently leads the research and development team at Cerebral Blue. Cerebral Blue is the company responsible for continuing to develop NeuralSeek.
Q: Where do answers come from if they are not contained in the corporate knowledge base?
A: NeuralSeek has many ML models running behind the scenes. These models enable NeuralSeek to answer some questions that are not specifically addressed in the corporate knowledgebase. These answers will be rated as “low confidence” in the system, and will also have a “low coverage” score. NeuralSeek allows you to block answers below a set minimum confidence level, and also add a warning note to answers with confidences below a warning threshold. This ensures a high level of control to corporate messaging, and enables critical guardrails for corporate use of generative AI.
Q: What models does NeuralSeek use?
A: NeuralSeek uses multiple models to generate answers. The core model is a tuned Large Language model. In addition multiple NLP models enable features such as context keeping, date extraction and comparison, sensitive content filtering, query expansion, and confidence scoring.
Q: When was Cerebral Blue founded?
A: Cerebral Blue was founded in 2020.
Q: How long has NeuralSeek been available for?
A: NeuralSeek reached general availability (GA) in September 2022, after a closed beta period.
Q: What formats, chatbots, or Virtual Agents is NeuralSeek Compatible with?
A: NeuralSeek is compatible with any Virtual Agent that can make a REST or webhook call. In addition, NeuralSeek can build AWS Lex, Cognigy, and IBM Watson Assistant virtual agents automatically from questions answered by NeuralSeek.
Q: Can NeuralSeek connect to third-party apis to consume data?
A: NeuralSeek can connect to both IBM Watson Discovery and Elastic AppSearch to consume corporate knowledge. Both of these integrations are no-code, and are easily configured via the NeuralSeek UI. No other inbound third-party integrations are supported. For outbound connectivity NeuralSeek provides an open API and may be consumed by any Virtual Agent, chatbot, or third party software or service that can make an external api call or webhook.
Q: Do you use GPT-3 (GPT3)
A: NeuralSeek uses multiple models to generate answers. The core model in our standard plan is a tuned Large Language model that we keep up to date for you. In our bring-your-own LLM plans you can select your LLM of choice to connect to, including OpenAI, Azure Cognitive Services, and AWS SageMaker. In addition multiple NLP models enable features such as context keeping, date extraction and comparison, sensitive content filtering, query expansion, and confidence scoring.
Q: What is context keeping? How do you maintain conversational context?
A: NeuralSeek maintains conversational context during interaction with a user. A session token is required to be sent with all requests to enable context features – this lets us track the conversation across multiple independent calls. NeuralSeek employs a multitude of NLP models to break down the structure of both questions and their generated responses. This enables NeuralSeek to estimate the topic of the conversation and keep the interaction focused in follow-on questions that do not specifically mention the primary subject. In addition, these NLP models enable NeuralSeek to filter corporate knowledge topically by date to ensure that the information being returned is focused on the time period of the question.
Q: Why are questions limited to 20 words?
A: User questions are limited to 20 words and checked for sensitive topics before they are answered by NeuralSeek. These restrictions are part of the safeguards that NeuralSeek provides to a company and their brand, helping to ensure that answers are on-brand and on-topic, and limiting an adversarial user’s ability to get NeuralSeek to generate a confusing, inappropriate, or off-message answer. Systems based only on a LLM, and lacking specific business guiderails will happily generate content on topics that are illegal, immoral, or contrary to your brand, because they lack the corporate safeguards that NeuralSeek provides. This is a major differentiator of NeuralSeek, as NeuralSeek’s focus is on targeted, branded, topical business answers. NeuralSeek is focused on business and protecting your brand while giving users the immediacy of answers that they expect, at a fraction of the cost of a live agent.
Q: Does IBM use NeuralSeek
A: IBM uses NeuralSeek internally in a Virtual Agent that powers IBM’s technical demos and solutions portal. This implementation of NeuralSeek helps IBM sales and technical resources find demonstration platforms and materials, answers technical questions about those assets, and helps to educate users on IBM products and solutions.