Availability
Q: When is the kHub platform expected to be available to the public?
A: We are targeting 2024 for the general availability of our services.
Please join our waitlist if you are interested in our services.
If you have any input, you can reach us by email at biz@khub.ai.
Target users
Q: Is the kHub platform intended to be used by individuals or businesses?
A: For the phase 1 of the kHub platform, we target those users who are influencers, those who needs a personal smart assistant, and domain experts who wish to inject his/her expertise into a specialist chatbot for use by others. For the phase 2, we expect to be able to take on business users who needs highly customized chatbots.
Skill requirements
Q: Do I need to have coding skill in order to use kHub?
A: No. You can create a kBot on the kHub platform simply by training it with uploaded text material, pointers to your media sites, or carrying on a conversation with your kBot in order to train it.
kHub usage cost
Q: How much do I need to pay for using kHub?
A: This is broken down as follows:
Users who visit kHub to use its kBots pay nothing.
Those who wish to create a kBot will bear the cost of the LLM usage (e.g., through getting their own OpenAI's ChatGPT API key).
This is necessary since at this time the cost of using certain LLMs is too high for our platform to bear.
We believe that this cost will come down over time, and we strive to make it more affordable as situation improves.
We may charge a subscription fee for those kBot owners who require additional services. The detail for this will be announced at a later time.
Monetization
Q: Can I make money from the kBots that I built on the kHub platform?
A: The short answer is Yes. Details for this will be announced at a later time prior to the platform launch.
Ads
Q: Why am I seeing ads on my kBot dashboard? Can I have them removed?
A: Our platform is partially ad-supported in order to reduce the cost for our users.
Users may opt to pay a subscription fee to get the ads removed.
Influencer vs Specialist vs Personal kBots
Q: What are the major differences among the various types of kBots offered?
A: An Influencer kBot provides in-depth support for integrating with owner's social media material over the Internet.
With a Specialist kBot, the purpose for for its owner to preserve and share his/her knowledge about a certain domain.
A Personal kBot is meant for personal use or by a small closely-knit group of people, with emphasis on providing ongoing support of daily activities.
Similarity or difference with genealist LLM chatbots
Q: How is a kBot similar or different with, say, ChatGPT, BingChat, or Google Bard?
A: All of the above are all a kind of LLM-based chatbot.
However, those other chatbots can be described as "generalist chatbots," which means that they can do many things, but they do not cater to your specific needs.
Such chatbots are first popularized by OpenAI's ChatGPT in late 2022.
On the other hand, each of our kBots is a kind of specialist that is trained to meet its owner's needs.
Each kBot has long-term memory for tracking of changes that are specific to its owner.
Each kBot is also specialized to solve a particular type of problem.
Utility of kBots vs generalist chatbots
Q: Generalist chatbots, such as ChatGPT, seems to work so well that I am able to use them for just about anything. Why would I need specialized ones like the kBots?
A: This is like saying that an operating system, e.g., Windows or Linux, is so versatile that we don't need any appliciation at all.
This is of course incorrect.
In order to provide a complete and polished solution for a given problem,
a suitably-designed user interface and many additional features must be implemented in an application, which runs on top the underlying OS.
That is, with the various specialized kBots on the kHub platform, users get to have solutions that better serve their needs.
LLM used
Q: Which LLM model that kBots are based on?
A: kHub is designed to be LLM-agnostic, meaning that we recognize that such technologies are still evolving rapidly, each with its own advantages and disadvantages. As such we do our best to design the kHub platform to adopt the best available.
Scope of specialist kBots
Q: The idea of "Specialist kBots" on kHub seems quite broad. What kind of specialists are supported?
A: This is limited by the capability of the latest available LLM technologies. For example, the current LLM technologies are mostly mono-modal, i.e., they understand only text.
We expect the LLM technologies to advance quickly to become multi-modal (e.g., incorporating images, audio, video, etc.), at which time we should be able to support more advanced kBots that are able to handle knowledge beyond textual.
Tweaking a kBot
Q: What do I do if my kBot is giving out wrong answers?
A: Just like any LLM-based chatbots, there is always some possiblity that a kBot may give the wrong answer (as judged by you).
When this happens we give you the tools needed to incorporate user's and your feedbacks into the retraining of your kBot, by engaging in a tutoring conversation with the kBot until it provides the correct answer (as judged by you).
Training a complex kBot
Q: I'd like to train a kBot to possesses deep understanding of a complex problem domain, is this feasible at all?
A: The current crop of kBots offered on the kHub platform are designed for users to get started quickly and easily on their own, without needing programming skills, and without hand-holding from us.
As a result such kBots are somewhat less powerful than a custom chatbot that requires complex training data, model fine-tuning, or integration with other systems.
We may support such complex and highly customized kBots on a case-by-case basis in the future.
We'd love to hear from you at biz@khub.ai if you have any input for us.
Custom chatbots
Q: Is your company in the business of building a chatbot to customer's specifications?
A: We are not in the business of building customized chatbots.
Whitelabeled kHub for enterprises
Q: Do you offer whitelabel solutions for businesses or organizations?
A: This is on our long-term roadmap, but we won't be offering such solutions until our platform is more matured.
We'd love to hear from you at biz@khub.ai if you have any input for us.
Risk to human
Q: Can the specialist kBots actually replace human experts?
A: Certainly not in the near term. Our specialist kBots are meant to provide pointers to quick answers and relevant source of information.
Such kBots certainly will not have the depth of a human expert in the near future.
Risk of abuse
Q: How do you prevent kBots from being used for abusive purposes?
A: We take every precaution to prevent kBots from being used for abusive purposes. Various measures such as captchas, content moderation, user verification, and blacklisting are used to identify improper use or nefarious actors.
Unverified kBot owners will be granted only limited rights, so that their kBots cannot be used for malevolent purposes.
Risk to privacy
Q: How do you protect the privacy of your users?
A: We take every precaution to protect the privacy of our users. However, since we may need to rely on external LLM services where we have somewhat limited control, we ask that our users avoid posting highly private and sensitive information.
For business users who need higher level of privacy protection, please send us an email at
biz@khub.ai and we
would love to hear your input on this matter.
Liability
Q: Who is liable if my kBot gives out highly improper or even illegal responses to a user?
A: Unless superceded by separate agreements with kHub.ai, as the owner of a kBot you are ultimately responsible for its behavior.
As such all kBot owners are advised to test out their kBot thoroughly before releasing them for public use.
We at kHub also strive to provide kBot owners all the tools necessary for them to train their kBots well.
Ownership of knwowledge
Q: Who owns the knowledge encapsulated within a kBot?
A: The answers for this question are broken down into the following categories:
We do not claim ownership of the raw media used by a kBot owner to train a kBot, even though those may be stored on our platform.
We do not claim ownership of those LLM foundation models that we acquire from external sources to support our operations.
The legal ownership of a fine-tuned LLM model depends on the terms of the agreement between the parties involved.
The legal ownership of any semantic embeddings created on the kHub platform depends on the terms of the agreement between the parties involved.
The kHub Terms of Agreement is not yet available at this point, since we are still at an early stage of our development cycle.