AI - #6 (Replacing Prompt Engineering with Voice! Example AI Application using NLP)

Replacing Prompt Engineering with Voice! Example Application of AI using Voice

Generative AI (“GAI”) uses a technique called "prompt engineering" to enable the "user experience".    Prompt engineering can be defined as the art and science of crafting the best prompt (another word for question or simply, as the GAI system input) to get the optimal output from a generative AI model.  It is a crucial skill for those who want to use generative AI models to create text, images, or other creative content.   The goal of prompt engineering is to provide the generative AI model with enough context and information to generate the desired output, while also avoiding ambiguity or misleading the model. This is often a challenging task, as the generative AI model may not be able to understand complex or nuanced instructions, plus crafting these inputs requires significant trial and error to learn.  (lot’s of typing :>) Thus, complexity adds an inverse relationship versus the overarching goal of making GAI easy to utilize.   A simple analogy is to look back at the early days of Google Search; it took some experience and massaging of the queries (massive trial and error) to get the desired search results. This was before the injections of adds, politics ideology, censorship into search … which added more layers and that finally resulted in one single fact: “Search sucks!”. Let’s do better for GAI, which is fast replacing search on many levels!

If the AI gurus want to deliver on all their promises of AI robots running the world and getting Skynet to go live (The system goes on-line August 4th, 1997. Human decisions are removed from strategic defense. Skynet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, August 29th. In a panic, they try to pull the plug), it will have to do a better interface than "prompt engineering", obviously.  

I am going to make a simple proposal and demonstrate one way that prompt engineering could be improved or replaced for generative AI applications.  I am going to start by sharing a use case (in this case a product) that I have been working on for a few years (see Quantum AI), which is a good example of how this will work.  To the point, prompt engineering can be replaced by NLP (Natural Language Processing).

Of course prompt engineering is not the only application of NLP that should be considered. Using in NLP in any AI application could offer potential benefits. In this post I will unpack a use case for NLP and AI that demonstrates some of the general benefits it can bring such as: (in generaly this assumes a trained LLM with machine learning capabilities and trained on the specific healthcare vertical knowledge sources)

• Accurate transcription of questions and responses

• Intelligent understanding and transcription of symptoms for specific disease states.

• Comprehension and transcription of diagnosis and recommendations

• Intelligent understanding and transcription of prescription drugs

Natural language processing (NLP) is a field of computer science that deals with the interaction between computers and human (natural) languages. NLP research has applications in many areas, including machine translation, text mining, and question answering (!!!).

NLP typically involves the following steps:

• Tokenization: This is the process of breaking down a text into smaller units, such as words, phrases, or sentences.

• Part-of-speech tagging: This is the process of assigning a part-of-speech tag to each token. This helps to identify the grammatical function of each word.

• Named entity recognition: This is the process of identifying named entities, such as people, organizations, and places.

• Semantic analysis: This is the process of understanding the meaning of a text. This can involve identifying the relationships between words and phrases, as well as the overall meaning of the text.

• Text generation: This is the process of creating new text, such as summaries or translations.

NLP is a complex and challenging field, but it has the potential to revolutionize the way we interact with computers. By understanding the nuances of human language, NLP can help us to create more natural and intuitive interfaces that are more helpful and informative.

Here are some examples of NLP applications:

• Machine translation: NLP is used to translate text from one language to another. This is a challenging task, as it requires understanding the meaning of the text in both languages.

• Text mining: NLP is used to extract information from text. This can be used for a variety of purposes, such as market research, customer sentiment analysis, and fraud detection.

• Question answering: NLP is used to answer questions about text. This can be used to create chatbots or to provide customer support.

• Sentiment analysis: NLP is used to determine the sentiment of a text. This can be used to identify positive or negative opinions, or to track the public's reaction to a product or event.

NLP is a rapidly growing field, and new applications are being developed all the time. As NLP technology continues to improve, it will have an increasingly profound impact on our use of technology (with the goal of making it easier and more accurate!).

Make it stand out

Whatever it is, the way you tell your story online can make all the difference.

The above "AI Assistant" example is an architectural view of a “medical scribe”. In real life, a medical scribe is a person who takes the responsibility of transcribing the interaction between the doctor and the patient. This is typically done by the doctor (the real world scribe is relieving them of this difficult but important duty). The Quantum scribe was built to help doctors gain back time with patients and avoid painful, slow and error prone manual data entry. Today, doctors spend much of the limited few minutes they spend with patients (on average 5-10 minutes) typing in the notes (or using poorly implemented voice solutions that require that doctor move the cursor to the field on their screen that is accepting input) to transcribe the interaction and document the necessary actions for the patient. Hardly optimal. Many misunderstandings and errors occur in this process.

Enter the Quantum AI assistant. The Scribe Assistant uses NLP and an LLM (the Quantum AI platform noted above) to achieve natural language data capture of the patient doctor conversation. This data is then accurately filled into the proper forms for the given patient diagnosis, and then integrated into the EMR (electronic medical record) of the patient, turned into accurate prescriptions, and provides a complete transcription of the report to the clinic and the patient. Of course, the audio (and/or video) of the session can be archived for any liability or insurance concerns that may popup down the road. The combination of the audio, video, and transcription ensures that there are less errors and misunderstandings between doctors and patients.

In future posts I will drill further down into the components of the Quantum system and other AI applications and approaches that I am working on.