Opportunities in AI
Generative AI, perhaps one of the most profound human inventions of the 21st century, is changing the way
companies operate, particularly in the fields of content creation and business intelligence. As a tool for creativity, its power is unparalleled. However, Generative AI (GenAI) is often misunderstood, maligned, and mistrusted, in part because it’s ascendency has been so speedy.
OpenAI, the makers of ChatGPT, launched their first public AI interface in November 2022. In just eight months, ChatGPT had over 100 million users and by February 2023 had amassed over a billion monthly page visits.
ChatGPT isn’t the only game in town, either. Microsoft, Meta, Google, Amazon, and other tech giants all have
in-house and specialist GenAI tools, whether fully open source, available to developers via APIs or closed.
It’s big business too. The market value of GenAI was recently estimated to reach over US$105 billion by 2027, displaying a CAGR of 8.11%. Looking at AI in general, Fortune Business Insights predicted a CAGR of 21.6% for the AI market between now and 2030, with market capitalization hitting $2 trillion by the end of the decade. The business opportunities are vast, and we’re only at the beginning of the AI gold rush.
Furthermore, these tools are increasing in capability at incredible rate. By May 2023, Anthropic’s GenAI was capable of generating up to 75,000 words in a minute, compared to the 9000 it could manage just two months earlier, according to McKinsey.
When aligned with corporate strategy and user priorities, GenAI can be a hugely powerful adjunct to human experience and expertise. It is also a major resource for service developers and offers significant opportunities for B2B innovation.
Six Links in the GenAI Value Chain
The GenAI value chain has six links, running from the underlying computer hardware (accelerator chips) up to the services and apps built on top of underlying architecture.
Here is that value chain in brief:
Services: specialized domain knowledge of GenAI
Applications: B2B or B2C products built on foundation models
MLOps and Model Hubs: curating / fine-tuning of AI services
Foundation Models: Core models upon which GenAI apps built
Cloud Platforms: Access point to hardware architecture
Hardware: Accelerator chips which allow models to run.
Evalueserve is investing in all four mid- and upper-level services. Let’s look at them in a little more detail.
1. Foundation Models
Underlying any AI is its foundation model – this is the specific architecture (such as a neural network) which has been pretrained to create a particular type of output (e.g., text, images, sound etc.) It can then be made available to developers to build services on top, either with open access, or via APIs.
There may be opportunities to create ever-better foundation models to counter some of the common limitations of the pioneering GenAIs, such as Chat GPT’s tendency to ‘hallucinate’ information. Of course, the difficulty of training such foundation models is that they require access to massive datasets, whether public or private in scope.
Another challenge is the complexity of the required training. As a recent McKinsey article pointed out, “to improve its next output so it is more in line with what is expected, the training algorithm adjusts the weights of the underlying neural network. It may need to do this millions of times to get to the desired level of accuracy. Currently, such training efforts can cost millions of dollars and take months.”
2. Machine Learning Operations (MLOps) and Model Hubs
The engine that powers generative AI is machine learning. This is the method by which an AI is trained on a dataset and employs pattern recognition and probability heuristics to interpret what a user wants and then generate an output. LLMs (large language models) are the type of GenAI (such as ChatGPT) that create text, but there are GenAIs that create static and moving visuals, write code or even generate protein structures for industrial or healthcare purposes.
ML services can be developed for highly specific tasks, and there will be an increasing demand for such services as GenAI products become ever more prevalent. These services are built on top of existing GenAIs, and service either corporate or individual user needs.
MLOps is the business of enabling developers to create such tools by assisting with data preparation, model creation, testing and deployment. Model hubs are the places developers go to access the foundation models upon which they build their more specialist products.
Developing expertise in model development and data analytics builds confidence in app developer clients, even when they don’t have direct access to source code. Companies such as Hugging Face and Tensorflow have staked an early claim on this territory but there’s much room for innovation and growth within this intermediate stratum.
3. App Development
The upper level of the development hierarchy are the apps which utilize underlying architecture and analytic
models to offer a specialized type of output. These are popular both with individual and corporate users. Here are just a few examples of top-selling AI-enabled apps:
Amazon Alexa: probably the best-known audio virtual assistant.
Fyle: an expense management tool with accountancy software integration.
Jasper AI: a written content and copy creation tool.
Shazam: the best-selling app for music identification.
DeepL: an automated text translator that incorporates idiomatic vocabulary.
Rock Identifier: a tool which does exactly what its name suggests!
There’s huge potential for businesses to develop apps which they can retain in-house or license out. B2B developers can create bespoke tools leveraging the power of GenAI to automate a host of challenging data analytic or creative tasks, without making human expertise redundant.
4. ML-Empowered Market Services
Sales and marketing teams expend million of hours annually sifting through potential leads, qualifying, and validating them to reach the customers that matter. Analytic AI systems can automate much of the early-stage lead research that once required many emails, cold calls, and rejections.
GenAI chatbots have been developed which prompt users to reveal the purpose of their site visit then pre-validate them for future contact. Users know these interlocutors are chatbots, but don’t mind interacting when they provide speedy access to the information they’re seeking.
Evalueserve has developed several such systems with banking clients to help assess loan applicants and evaluate risk. AI not only allows for in-depth and accurate assessment but can suggest courses of action to take to minimize risk and deliver tailored products, thus minimizing future defaults.
AI can even identify patterns of repayment delinquency which indicate an early need for intervention, helping save banks millions in lost payments, and clients thousands in interest payments and penalties.
This is just one example of how an industry with a high volume of customer interactions can benefit from the analytic power of AI. Utilities companies, travel businesses, and entertainment providers might equally benefit from such products.
The GenAI future can only be glimpsed right now but it’s likely to exceed all expectations and carry us into areas of innovation we can scarcely imagine. Partner with Evalueserve to come with us on the GenAI journey.