ChatGPT and enterprise leaders

Share on

Share on

14:07 Minutes The average reading duration of this insightful report.


Primers: Primers are quick short form business reports that educate leaders on key emerging technologies.

ChatGPT and enterprise leaders

Access full research and
ignite the inspiration within

Download Primer

Discover the transformative impact of ChatGPT in the business world. Explore its potential in natural language processing, AI's role in enterprise strategy, and how leaders can leverage this technology for growth and innovation.

Explore a sneak peek of the full content

What’s inside:

  1. Evaluating ChatGPT’s value for enterprises short and medium-term.
  2. Understanding NLP in a business context with real-world examples.
  3. Recognizing limitations and strategic planning.

The value for you as an enterprise leader

Short-term deployment strategies
In the immediate future, the emphasis is on implementing ChatGPT Plus and its API in selective functions within enterprises. This phase aims to measure the return on investment by integrating ChatGPT in various divisions, particularly in areas like code development and marketing. A key benefit of this approach is the potential enhancement of employee productivity through accelerated learning and execution, leveraging ChatGPT’s advanced capabilities.

Value Of Enterprise Leaders

Medium-term licensing and training
For a 6-month to 1-year outlook, the focus shifts to licensing GPT3.5 and tailoring it with company-specific intelligence. This move aims to bypass the limitations of the general-access SaaS model and utilize ChatGPT’s full potential. By customizing the AI with domain-specific data, enterprises can create distinctive products or services, thereby gaining a competitive edge.

Short Term Vs Mid Term

Considerations and limitations
Key considerations include the confidentiality of data, the competency and adaptability of employees, and the initial costs and resources required for deployment. The hardware prerequisites, licensing costs, and additional expenses for model training are also crucial factors. The strategy involves a careful balance of immediate benefits against long-term investments, ensuring that the integration of ChatGPT aligns with the enterprise’s overall objectives and capabilities. Download Complete Research

The full model of GPT3 has 175B parameters. It translates to ~1TB of memory and requires a high-end GPU like NVDIA A100 & a highend CPU like Intel Xenon.

The technology behind ChatGPT (Natural language processing)

Evolution of machine learning and deep learning
The foundation of ChatGPT’s technology lies in the evolution of machine learning, a key subset of artificial intelligence where computers are trained to emulate human performance. Initially, machine learning powered simple applications like search and recommendation engines. Over time, it evolved into deep learning, which uses neural networks for more complex tasks. These neural networks, comprising units called artificial neurons, mimic the human brain’s functioning, processing data through interconnected nodes. This advancement is evident in modern applications ranging from chatbots to intelligent assistants.

Breakthrough with transformer models
A significant leap occurred in 2017 with Google’s introduction of transformer models. These models, central to ChatGPT’s technology, excel in processing entire sentences and generating text. They operate using an encoder-decoder mechanism and focus on the ‘attention’ principle, determining the relevance of each word in a context. OpenAI’s investment in these models led to the development of GPT (Generative Pre-trained Transformer) series, culminating in ChatGPT.

Factors Contributing to ChatGPT’s Success

  1. Massive dataset: Trained on a vast array of internet sources, ChatGPT’s dataset includes a staggering 499 billion tokens, offering a broad base for learning and response generation.
  2. Large-scale model: With 175 billion parameters, ChatGPT dwarfs its nearest competitor and demonstrates more nuanced understanding and response capabilities.
  3. Computational power: The use of Azure Supercomputers enables ChatGPT to process and learn from its extensive dataset efficiently.
  4. Refined algorithms: OpenAI’s continual refinement of the transformer model has significantly enhanced deep learning capabilities.
  5. Human feedback reinforced learning (HFRL): This training methodology incorporates human input to fine-tune the AI’s responses, making them more accurate and contextually appropriate.

This section highlights the technological advancements behind ChatGPT, illustrating its journey from basic machine learning applications to sophisticated natural language processing capabilities. Download Complete Research

The limitations

Evolution Of Learning Models

Accuracy and misinformation
ChatGPT’s training on extensive internet data poses risks of inaccuracy and misinformation. It often lacks the latest updates and struggles to differentiate between fact and fiction, leading to potential misinformation, especially for non-experts.

Contextual understanding and bias
Another limitation is its inability to interpret emotions or hidden intentions, potentially resulting in inappropriate responses. Furthermore, biases in its training data can skew ChatGPT’s outputs, reflecting these inherent biases in its responses.

Operational costs and legal implications
Maintaining ChatGPT involves significant costs due to its complex system requiring regular updates. Additionally, legal challenges, such as copyright issues, can arise from its text generation capabilities.

ChatGPT has been trained on massive amounts of data from the internet, hence knows only the internet (which as humans we know can have inaccuracies and biases)

Environmental impact
The energy consumption for running ChatGPT is substantial, contributing to environmental concerns. The costs, both financial and environmental, of operating data centers and processing large datasets are significant, highlighting a need for more sustainable practices.
This section highlights ChatGPT’s main challenges: accuracy and bias issues, high operational costs, legal risks, and environmental impact. It emphasizes the need for addressing these concerns for its effective and responsible use. Download Complete Research

Lead Authors@lab45: Arvind Ravishunkar, Dinesh Chahlia, Nitin Narkhede, Noha El-Zehiry
Contributing Authors@lab45: Aishwarya Gupta, Anindito De

Latest stories

Generative AI startups: Landscape & trends

Share on

Share on

12:49 Minutes The average duration of a captivating reports.


Primers: Primers are quick short form business reports that educate leaders on key emerging technologies.

Generative AI startups: Landscape & trends

Access full article and
ignite inspiration within

Download Primer

Generative AI is forming a new economic ecosystem, reshaping the behaviour of key players in the IT industry, generating opportunities for super-scalers, and unveiling numerous niches for startups. The outlines of this new IT landscape are emerging, prompting a closer examination.

What's inside

  1. Introduction: The rise & impact of generative AI
  2. The technology & business stack of GenAI
  3. Business niches for GenAI startups
  4. Future trends in GenAI
  5. Appendix 1: Reality & expectations of GenAI
  6. Appendix 2: Startups across the GenAI tech stack

Introduction: The rise & impact of generative AI

Generative AI has caused significant disruption, expanding its offerings and services well beyond traditional AI domains. This has led to an explosion of potential use cases for customers who aren't AI experts. Unlike before, customers no longer require a team of AI experts, curated data, or precisely measurable outcomes to adopt AI tool and gain immediate benefits. The interaction with GenAI is so seamless and intuitive that the onboarding for new customers is frictionless, eliminating barriers to adoption and facilitating rapid technology spread. The high variability in potential inputs and priming of generative models allows for a diverse range of applications impacting nearly every imaginable aspect of people activities. This is a foundation of a new era of Artificial Intelligence.

In this primer we leveraged our knowledge of 50+ GenAI-related and VC-backed startups to reconstruct the technological stack of the forming GenAI space.

The technology & business stack of GenAI

Large tech companies are leveraging their existing technological and capital advantages to create the framework for the GenAI market landscape, which we are going to explore in this section.

While offering of the LLMs on the current scale and heavy focus on unstructured data are somewhat new, the other elements of the tech stack closely mirror those needed for any large computational modeling. Established companies in the field of traditional AI are at an advantage, as they can expand and repurpose preexisting software, infrastructure, and services. Download Complete Research

Business niches for GenAI startups

While large players are occupying a sizable portion of the GenAI tech stack, there remains more than enough room for GenAI startups to flourish. The landscape of AI and ML is continuously evolving, with new startups, technologies, and methodologies emerging regularly.

Bottom-right (AIOps): Here, startups may offer tools for easier adoption of LLMs, facilitating the initial process of customizing and implementing these models.

Ascending (Integration): Moving upwards represents the process of integrating LLMs into various applications and business operations. Startups could offer integration services, templates, or frameworks to streamline this, or build an entire end to end app for a selected market niche.

Moving left (Service platforms): As we move leftwards, the focus shifts from core LLM functionality to auxiliary services. This could range from platforms offering specialized training data, to marketplaces for LLM apps, to optimization tools. These firms may automate the need for certain experts.

This taxonomy can serve as a foundational overview for anyone looking to understand the current state of the LLM ecosystem. It’s also worth noting that the landscape of AI and ML is continuously evolving, with new startups, technologies, and methodologies emerging regularly. Let’s inspect each block in greater detail:

Future trends in GenAI

The future of the GenAI landscape is going to be defined by several processes:

  1. Consolidation of major players
  2. Rise in open-source adoption
  3. Surge in service platforms
  4. Expansion of skill marketplaces
  5. Segment-specific applications
  6. Regulatory oversight and standardization

Appendix 1: Reality & expectations of GenAI

While enhancing the users with great capabilities, the LLM-based service is neither a freebie, nor a cornucopia. Each implementation of LLMs carries its own advantages and downsides. In this section of the Appendix, we discuss what can and cannot be realistically expected from a GenAI model in each of the most popular use cases.

We start with primary properties of a pre-trained LLM model, underlying its strong sides and functionalities as well as build-in flaws. And we move to the current ways of augment LLM model to work around the flaws. Download Complete Research

Appendix 2: Startups across the GenAI tech stack

The table of 78 startups we have based our analysis on is presented in this section.
The states of startups are set to the August of 2023.

Author@lab45: Rinat Sergeev

Top trending insights

Generative AI on the cusp of disruption

Share on

Share on

11:39 Minutes The average duration of a captivating reports.


Primers: Primers are quick short form business reports that educate leaders on key emerging technologies.

Generative AI on the cusp of disruption

Access full article and
ignite inspiration within

Download Primer

By acquiring 100+M users in first 3 months of launch, ChatGPT has brought the field of Generative AI (GenAI) into mainstream awareness. The adoption of ChatGPT and similar applications have positioned Generative AI (as well as “Deep Learning”) as the newest disruptive tech after cloud computing.

What's inside

  1. An overview of Generative AI, the market for it and the reasons for the hype
  2. A qualitative assessment of its time to mainstream across various industries
  3. Enterprise use cases and its limitations

What is Generative AI?

Generative AI, a field of Artificial Intelligence, refers to computational models that are trained on massive amounts of input data (300bn words in the case of ChatGPT). They can synthesize data, draw inferences and create new outputs in the form of text, images, video, audio, new data and even code.

Two architectures have made GenAI immensely valuable

  1. Generative Adversarial Networks (GAN), that became popular in 2014 are used for generating images & videos.
  2. Transformer Models, proposed by Google in 2017, are used for generating text.

ChatGPT, for instance, uses the transformer model along with Human Feedback Reinforcement training to generate high quality outputs. Training a model requires intensive computational power (supercomputers were used to train GPT3) and significant investments (OpenAI being a key example). But once a model is trained, it can be optimized for a larger user base. Download Complete Research

The market for Generative AI

The market is expected to grow from $8B in 2022 to $109B in 2030 at a CAGR of 34.6%. Key facts as follows

  • Software segment accounts for 60%, service segment is the fastest growing.
  • Media & entertainment is the biggest user of Gen AI accounting for 18% of revenue,  BFSI is the fastest growing at 36% CAGR.
  • North America is the biggest market with 40% share and APAC is the fastest growing region.

Why GenAI is here to stay?

Need for content synthesis
We generate ~2.5 quintillion bytes of data every day on the internet. This not only makes searching for information tough but also makes inferring tougher, for regular users. GenAI tools can search, synthesize and compose an answer.

Democratization of content creation
We are moving from a search and retrieval economy to an infer and compose economy. People used to prompt algorithms to search and retrieve information but now they can prompt algorithms to infer and compose information.

Instant economy
Digital natives prefer tools that enable instant creation of content e.g. Tik Tok. ChatGPT can generate a word in 350ms after processing database of 300B words.

Access to massive computational power
The ability to instantly process and compose information using cloud computing.

Evolution of deep learning neural networks
Large Language Models have become openly available. These models help organize much of the internet’s information and develop patterns to mimic human decision-making.

Download Complete Research

An early assessment of time to market & key GenAI companies.

Industry To Mainstream Gen Ai

Gen Ai Companies Capabilities

Enterprise use cases of Generative AI

GenAI will fundamentally change several functions in enterprises leading to improved productivity and performance of employees. A Few areas of business where it will have the biggest impact are as follows:

Content creation
Gen AI will lead to more automation in content creation. It will not only reduce the cost of content creation but also increase the quality & variety of content created. Generative AI based DIY Apps are expected to emerge for marketing and design functions.

Content personalisation
Marketing touchpoints like newsletter, websites, videos, metaverse etc. will get hyper-personalized. This will improve brand engagement and conversion ratio of the sales funnel.

Drug discovery
Drug Discovery is a time-consuming process that can extend to 5-12 years. Gen AI can help identify potential drug candidates and test their effectiveness using computer simulations, thus saving time in the process. It has already led to tremendous real-world value, when the first mRNA covid vaccines were developed by programming mRNA molecules to express the specific antigen response. By 2025, more than 30% of new drugs and materials could be systematically discovered using GenAI techniques, up from zero today.

Software development
IT products and services could see the biggest impact. Below are some scenarios that may unfold.

  • Reduced time on testing and coding
    Gen AI has the capability to create, test and debug the code in real time. In typical product development cycles, coding and testing takes 30-40% of the time. Gen AI will cut this significantly, thus reducing time to market.
  • Improve programmer performance
    Code can now be generated with a simple prompt command. This will allow even less-tech savvy programmers to generate a better code. Gen AI can also translate the code from one programming language to another. However human intervention will still be needed to customize the code for specific vertical / client use.
  • Automate recurring tasks
    Manhours will be freed from repetitive tasks, as automation is easier with GenAI. Tasks like report generation, log analysis etc will fall in the domain of automation.
  • More secure and reliable IT infrastructure
    Gen AI can track performance and security of IT infrastructure in real time. It can pre-empt any failures, by generating early warning signals and hence improve reliability of operations.

    Download Complete Research

Lead Authors@lab45: Siddhant Raizada, Nagendra Singh, Tommy Mehl, Arvind Ravishunkar
Contributing Authors: Aishwarya Gupta, Anindito De

Co-create for collective wisdom

This is your invitation to become an integral part of our Think Tank community. Co-create with us to bring diverse perspectives and enrich our pool of collective wisdom. Your insights could be the spark that ignites transformative conversations.

Learn More