Generative AI For The Rest Of US

Generative AI Applications Across Industries

Generative AI is making waves across a wide range of industries, from healthcare and finance to education and marketing.

"Generative models are changing the way we think about machine intelligence and creativity, and have the potential to transform industries from media to finance to healthcare." ~ Oriol Vinyals, Research Scientist at Google

Let's explore some of the key applications of Generative AI in these sectors and the impact it's having on the future of work.

Generative AI in Healthcare

Could an AI chatbot match your doctor's ability to diagnose your headache or stomach pain? Is GPT-4 as good and as empathetic as a mental health professional? In my two books, "OpenAI GPT for Python Developers", first and second editions, I have explored this potential as part of hands-on projects for my readers. The first example is a chatbot that was trained on a large dataset of drugs and their uses. The goal was to create a conversational AI that could help doctors and patients find the right drug for a specific condition. The second example was a mental health chatbot that was trained on a vast dataset from CounselChat, a platform where people can ask mental health questions and get answers from professionals. The chatbot was designed to provide support for people experiencing mental health issues like depression, eating disorders, and anxiety. Both chatbots were built using a fine-tuning approach on top of OpenAI GPT models. The results were promising, and with more fine-tuning layers, the chatbots could be prepared for use in the real world. I was impressed by the relatively easy overall process and the associated results.

A study titled "ChatGPT and Generating a Differential Diagnosis Early in an Emergency Department Presentation" suggests that ChatGPT may be capable of diagnosing patients as effectively as trained physicians. The study, serving as a proof of concept, highlights a 60% overlap in diagnoses between doctors and ChatGPT, with the AI achieving a 97% accuracy rate in its shortlists - outperforming doctors' 87%.

In reality, Generative AI is already being used in fields like drug discovery. Traditional drug development can take up to 15 years and cost between $0.5 billion and $2.6 billion. In contrast, Insilico Medicine used Generative AI to identify a potential drug candidate in just 46 days. That's faster than it takes to get a doctor's appointment! Atomwise, a San Francisco-based company, uses AI to expedite the process of drug discovery, while Variational AI, a Canadian company, has unveiled Enki, a foundational model designed for small molecules. Enki allows for the quick creation of new molecule structures by inputting a Target Product Profile (TPP) aimed at treating certain diseases, using generative algorithms trained on extensive experimental data. Deep Genomics, another notable company, provides an AI framework for simplifying the complex topic of RNA biology, which involves the control of gene expression via ribonucleic acid molecules. With the help of their advanced technology, Deep Genomics can identify the best treatment candidates for specific diseases.

Generative AI is also excelling in the field of medical imaging. Using convolutional neural networks, which enhance the analysis of images, AI automates the segmentation of scans and predicts anomalies. A notable example of AI-driven image analysis success is the development by Google's health research team of a system that, in some instances, outperforms radiologists in detecting breast cancer. With the help of large datasets of anonymized mammograms, this AI is engineered to identify signs of cancer potentially missed by experienced specialists. Another example to consider is the AI-enhanced medical imaging solutions that have been used to diagnose COVID-19. Huiying Medical in China developed this tool and claimed a 96% rate of accuracy in early virus detection.

Generative AI not only automates diagnostics but also predicts health risks and prevents diseases by analyzing patient data, identifying patterns, and offering personalized recommendations. Freenome, for example, is a company that leverages Generative AI in its multiomics platform to detect cancer at an early stage through advanced blood tests. Colorectal cancer is one of them.

In the field of personalized medicine, companies like Celcuity develop cell-based assays using a cellular analysis platform, providing specific cellular data of patients to offer unique insights into a patient's response to drug therapy.

When it comes to mental health, Limbic, a company that develops a Generative AI care companion, is a good example to highlight. Limbic is a mental health startup that provides personalized mental health support. The company states that its app makes treatments 40% more effective.

"Eventually, doctors will adopt AI and algorithms as their work partners. This leveling of the medical knowledge landscape will ultimately lead to a new premium: to find and train doctors who have the highest level of emotional intelligence." ~ Dr. Eric Topol, Director of the Scripps Research Translational Institute

In addition to these practical applications, the medical research community is leveraging Generative AI to accelerate the pace of research, optimize resources, and offer predictive insights. Reading and analyzing a research paper can take hours, but with the help of Generative AI tools like Elicit, researchers can quickly extract key information, identify trends, and generate insights from scientific literature in minutes.

The potential of Generative AI in healthcare is vast, and a collaborative approach between AI developers, healthcare providers, and regulatory bodies is essential to ensure its responsible and ethical implementation. Google's DeepMind is a prime example of a company that has made partnerships with several healthcare institutions to develop AI-driven solutions for medical research. One of their most notable projects is the development of a Generative AI model that predicts the 3D structure of proteins, a critical step in understanding how proteins function and interact with other molecules. This model is known as AlphaFold.

Generative AI in Finance

To stay competitive, BlackRock, the world's largest asset manager, has been aggressively embracing AI and its generative counterpart. This focus on AI extends beyond just using existing solutions. BlackRock is actively developing new, in-house AI-powered tools specifically designed to benefit both employees and clients. One example is a chatbot that can address customer questions, streamlining the client experience. BlackRock is also pioneering AI-powered "co-pilots" aimed at boosting employee productivity. This focus on unique solutions reflects a strategic decision by BlackRock. Rather than relying on standard industry tools, BlackRock intends to leverage its Aladdin platform - a data powerhouse that generated a staggering $1.4 billion in revenue in 2022 - to fuel the development of its AI tools. BlackRock launched a private preview of its first co-pilot for a select group of 10 clients using eFront, a software platform wholly owned by BlackRock and housed within the Aladdin umbrella. Following the conclusion of this preview phase, the co-pilot was made available to all 130 clients of eFront Insight - a group that includes major pension funds, funds of funds, insurers, and other asset owners.

JPMorgan Chase, another financial giant, is also making use of GenAI. The bank developed IndexGPT, a Generative AI software, to analyze securities and recommend investments based on customer preferences. This initiative positions JPMorgan as a frontrunner in offering GPT-like technology directly to consumers.

Generative AI can play a crucial role in fraud detection. By analyzing vast amounts of data, tools like Feedzai's Railgun can identify patterns and predict potential threats. With an average of $51 million lost annually to fraud, representing about 1.7% of total revenues, US fintech companies are currently facing significant losses. Mastercard, for instance, has developed its proprietary Generative AI model to help banks detect and prevent fraudulent transactions. Decision Intelligence Pro, the company's AI model, works by analyzing the immense volume of transactions flowing through Mastercard's network. Mastercard claims their technology has reduced false declines by 300%.

BlackRock, JPMorgan Chase, and Mastercard's decisions to develop rather than buy underscore a key takeaway: In finance, a competitive advantage lies in the creation of unique, custom-built Generative AI solutions instead of purchasing them. This trend of in-house development in finance is likely to become standard practice.

Other companies we can mention are Finpilot and AlgosOne. They are leveraging Generative AI to streamline the analysis of extensive documentation and optimize trading strategies, respectively.

In the insurance sector, a competitive space, Lemonade is an excellent example showing how GenAI can be a competitive advantage. Lemonade uses Maya, an AI-powered chatbot, to handle claims and underwriting and provide customers with a friendly experience.

The use of Generative AI extends to portfolio management, risk assessment, and investment recommendations. Kavout, for instance, offers a Generative AI platform for investment professionals. The platform uses deep learning algorithms to analyze market data, identify trends, and predict stock performance.

The potential of Generative AI in finance is vast, and its adoption is expected to grow in the coming years.

Generative AI in Education

Imagine a future of education where AI tutors personalize learning to match each student's style and pace. Imagine a world where teachers provide individualized attention based on real-time insights into every student's progress. Generative AI has already turned these visions into reality.

Take the example of Duolingo, leading the language learning industry by using Generative AI to analyze user data, understand their learning habits, and create customized lessons. Learning a new language can be difficult, but aligning with personal learning preferences will certainly make it easier.

"Technology will never replace great teachers, but in the hands of great teachers, it’s transformational." ~ George Couros, Teaching consultant

With the help of OpenAI's GPT-4, Duolingo has been able to create educational roleplays that simulate real-life scenarios. For instance, users can order a coffee at a café in Paris and specify their preferences in French: "Je voudrais un café au lait, s'il vous plaît." The waiter will then respond in French: "D'accord, voulez-vous du sucre avec ça?" and the discussion continues in French or the language of choice. This immersive experience adds a realistic touch to language learning and from a psychological perspective, nothing beats real-life scenarios. Users can also use a feature called "Explain My Answer" that provides detailed explanations for why an answer is correct or incorrect.

Similarly, Squirrel AI leverages Generative AI to create personalized learning pathways. By analyzing student data, the platform identifies educational gaps and implements specific interventions. Squirrel AI serves as an "AI Learning Intelligent Adaptive Learning System (IALS)" as it uses Generative AI to provide personalized and adaptive learning experiences. It breaks down academic subjects like mathematics into thousands of fine-grained knowledge components, which are interconnected in a knowledge graph structure.

ℹ️ The knowledge graph is a key component of Squirrel AI's AI-driven learning experience. It represents the relationships between different knowledge components and how they relate to each other. Each knowledge component is linked to specific learning materials like a text passage, an animation, or a video. As students interact with the system and generate learning data, the AI algorithms continuously update and optimize these relationships.

Another application of Generative AI in education is OpenAI's Codex, a Generative AI model that can generate code snippets in multiple programming languages. Codex can be used to create educational content, such as coding tutorials, exercises, and quizzes. It is currently powering a number of programming productivity tools, including GitHub Copilot, a code completion tool popular among developers of all levels. The coding co-pilot suggests code snippets as you type code. However, the innovative potential of Codex is overshadowed by an ongoing controversy over the rights and copyright implications of Generative AI. The Free Software Foundation, a non-profit group advocating for the use of free software and its distribution under copyleft licenses, has expressed concerns about potential copyright infringements linked to AI-generated code from tools like Copilot and Codex, especially in relation to the derivative work conditions of the GPL. The GPL software license, used by many open-source projects, is a copyleft license that requires derivative works to be licensed under the GPL. For instance, if you create a new application based on existing code with a GPL-license, GPL mandates that your new code should be licensed under the same license. The FSF argues that the way code is used to train AI models is unfair: