Elevating human-machine relationships with no-code, reusable AI

We are excited to bring Transform 2022 back to life on 19th July and virtually 20th July – 3rd August. Join AI and data leaders for sensible conversations and exciting networking opportunities. Learn more about Transform 2022


Our ability to discover and use tools is vital to human evolution. Computers as tools have certainly advanced humanity since their inception. As computing technologies advance, human-machine relationships also evolve. Initially only computer developers or programmers could operate a computer by giving machine (programming) instructions that the computer could understand and follow. With the development of the graphical user interface (GUI), the public can now operate computers without any code. The human-machine relationship, however, remains the same as the operator-machine relationship, during which humans must tell machines exactly what to do.

With the rise of Artificial Intelligence (AI) – computers with certain human skills – human-machine relationships can be completely redefined. For example, computers with human visual acuity can enhance security personnel to quickly identify objects in the mountains of surveillance images, or computers with human language skills can enhance paralegals to summarize a large number of text documents. However, teaching human skills to machines is a complex, time consuming process that requires deep skills and programming skills, not to mention the efforts to collect, clean and critique the large amount of training data needed to train machines with the desired skills. .

No-code, like GUI-powered computer operations, what if humans, security personnel, and paralegals alike could teach machines human skills without a code? Like in the movie HerWhat if we could adopt a turnkey AI assistant with built-in human skills and easily customize it without the code to meet our specific needs? This vision of no-code, reusable AI will definitely enhance our current operator-machine relationship into a supervisor-assistant relationship. New relationships will not only enable us to enhance rather than replace them with AI, but the no-code nature will also democratize human growth.

1. AI by human skill

Depending on the tasks to be achieved, AI systems are trained to have different human skills. Figure 1 illustrates the AI ​​system through human skills. Some AI systems use a A type of human skill, Such as human visual perception or linguistic skills, to perform a specific task, such as object identification or emotion analysis. In contrast, more complex AI systems are used Multiple human skills Together to achieve complex tasks. For example, a self-driving car should use multiple human skills to achieve its driving goals, such as human visual vision and decision making skills. Similarly, a conversation AI assistant should use multiple human skills, such as communication skills or specific human soft skills (e.g., active listening) to accomplish its tasks.

Diagram, timeline description is generated automatically
Figure 1. Example of an AI system with different human skills.

2. Multi-level reusable AI

It doesn’t matter if the AI ​​system requires one or more human skills to work, creating an AI system from scratch is always difficult and requires a lot of skill and resources. Like building a car, instead of building it entirely from scratch with raw materials, it would be much easier and faster if we could quickly customize and assemble pre-built parts and systems like engines, wheels and brakes.

While there are many no-code, reusable AI systems, due to the complexity of technology and the need for multi-level, re-use of no-code, complex AI systems, such as enabling a conversational AI system, is the most challenging. Reuse. Figure 2 shows an example of a 3-level architecture in support of a cognitive AI assistant, a new generation of AI assistants with multiple advanced human skills, including soft skills.

Diagram descriptions are generated automatically
Figure 2. Example architectural cognitive AI assistant with reusable AI at multiple levels.

Reuse of general purpose AI models

As shown in Figure 2, the following level is a set of general purpose machine learning models on which any AI system depends. For example, data-based neural (deep) learning models, such as BERT and GPT-3, are typically pre-trained on a large amount of public data, such as Wikipedia. They can be reused throughout AI applications to process natural language expressions. General purpose AI models however are insufficient to power the cognitive AI assistant. For example, general purpose models trained on Wikipedia may not handle general brief communication, such as managing a conversation or guessing the user’s needs from the conversation.

Reuse of specialty AI engine

To power an AI assistant with human soft skills, a special AI engine (medium level) is needed. For example, the Active listening The engine shown in Figure 2 enables the AI ​​assistant to focus on the conversation and gives it memory so that it can properly interpret the user’s input, including incomplete and obscure expressions in the context as shown in Figure 3.

Figure 3. Examples show how the Cognitive AI Assistant interprets the same user input in two different contexts and is able to respond accordingly.

Similarly, specialty AI engines such as the Reading and Conversation Communication Engine between the lines empower the AI ​​assistant with additional human skills. For example, reading between lines enables AI assistants to analyze user input during a conversation and to automatically predict the user’s unique characteristics (Figure 4). Conversation-specific communication engines enable AI assistants to better interpret user expressions during conversations, such as user input is a question that is a reflective statement, ensuring a variety of AI responses.

With careful design and implementation, all specialty AI engines can be made reusable. For example, an active listening conversation engine can be pre-trained with conversation data and pre-built with optimization logic to find various contexts of conversation (e.g., user giving excuses or asking clarification questions). Always trying to balance the user. Complete experience and work while handling user interruptions to guide the conversation.

Figure 4. Example showing how the Cognitive AI Assistant is able to analyze the text of a user’s conversation and automatically predict the user’s soft skills.

Reuse of complete AI assistants

In addition to reusing individual AI components / skills, the ultimate goal is to reuse a complete AI solution. In terms of creating AI assistants, re-use a complete AI assistant based on AI assist templates with predefined workflow and consistent knowledge base (top level of Figure 2). For example, the AI ​​Recruiting Assistant template includes a set of job interview questions and a knowledge base for answering job-related FAQs. Similarly, the AI ​​Learning Assistant template outlines workflows, such as checking student learning conditions and delivering learning instructions or reminders. Such templates can be reused directly to create turnkey AI accessories or can be quickly customized to suit specific needs as shown below.

3. No-code AI reusable AI

Each AI solution typically requires some customization, reusable AI no-code enables AI customization. Below are some examples.

No-code customization of AI support templates

Suppose the HR recruiter wants to create a custom AI recruitment assistant based on the existing AI template. Like using PowerPoint or Excel, the recruiter will use the GUI to customize interview questions (Figure 5) and job-related FAQs. No-code customization greatly simplifies the creation of powerful, end-to-end AI solutions, especially for non-IT professionals.

Automatically generate graphical user interface, text, application, chat or text message description
Figure 5. No-Code Customization of AI Recruitment Assistant for asking specific questions (T17). AI Assistant will automatically handle discussions on this topic.

Continuing the above example, assuming that the recruiter wants to ask questions to the applicants for the AI ​​assistant job. “What do you like most about your current job?If the applicant’s response is something like this, “Communicate with customers“Recruiter wants to ask AI follow-up question”Can you give me an example that you enjoyed interacting with your customer?Since the pre-built AI template does not handle this particular case, the recruiter will need to customize the AI ​​communication. Figure 6 shows how such customization can be done without coding.

Figure 6. No-code customization of AI assistant based on user response to question in T17 with follow-up question (T18). AI Assistant will handle the workflow automatically.

4. No-code defines reusable AI supervisor-assistant relationships

No-code enables reusable AI Everyone, Including non-IT professionals, to create their own custom AI solutions (assistants). The AI ​​assistant only needs to be notified What to do (E.g., asking users a set of questions) and then performing tasks Automatic (E.g., how to handle user interruptions). This transforms the traditional operator-machine relationship into a supervisor-machine relationship. When humans have to program / code a machine to teach machines, humans play a role. Operators / Developers Of machines. While humans now provide machines with high-level, no-code instructions, such as outlining tasks and teaching new knowledge, humans now Supervisor Of machines. This new relationship enables humans to do more with the help of machines.

5. No-code, future directions of reusable AI

No-code, reusable AI democratizes the creation and adoption of powerful AI solutions without the need for rare AI talents or expensive IT resources. In addition, no-code, reusable AI enhances human-machine relationships, enabling them Everyone The machine will be enhanced by powers. To create a no-code, reusable AI is a prime example of developing and adopting AI solutions, progress should also be made in some areas.

Explainable AI

The first area is to make reusable AI components / systems understandable. To help non-IT employees reuse pre-trained or pre-built AI components and solutions, it is important to unbox the “black box” and explain what is inside each component or solution, both pros and cons. Explainable reusable AI not only helps humans better understand and take advantage of existing AI components / systems and also helps avoid potential AI pitfalls. For example, HR will help recruiters understand how personal insights are predicted before they use such AI power to predict applicants’ insights.

Automatic AI debugging

The second area will be the basis of automatic AI debugging. As AI solutions become more complex and sophisticated, it is difficult to manually test potential AI behavior in a variety of complex situations. Non-IT users in particular will need help evaluating the AI ​​solution (e.g., AI Assistant) and improving it before formalizing it. Although there is some preliminary research on the profile of AI assistants, much remains to be done.

Responsive AI

The third area will ensure responsible use of AI, especially with the democratization of AI. For example, if a person can reuse only the AI ​​functional unit to retrieve sensitive information from users, how and who can protect users and their sensitive information? In addition to measuring specific AI performance such as accuracy and robustness, new steps and usage guidelines will be needed to ensure the design and use of reliable and secure AI solutions.

Michelle Zhou, Ph.D. Is the co-founder and CEO of Juji Inc.

DataDecisionMakers

Welcome to the VentureBeat community!

DataDecisionMakers is where experts, including tech people working on data, can share data-related insights and innovations.

If you would like to read about the latest ideas and latest information, best practices and the future of data and data tech, join us at DataDecisionMakers.

You might even consider contributing to your own article!

Read more from DataDecisionMakers

Similar Posts

Leave a Reply

Your email address will not be published.