Artificial Intelligence Market Report

This report provides an overview of key terms and theories related to Artificial Intelligence (AI), examines the global AI market, highlights notable AI projects and companies, and offers insights into the concept of Artificial General Intelligence (AGI)
SUMMARY

Artificial Intelligence: Concepts and the Global Market
February 29, 2024

Global AI Market Overview


  • Artificial intelligence (AI) is a comprehensive term that includes computer systems that mimic human intelligence.
  • Machine learning (ML) is a branch of artificial intelligence that focuses on creating systems that learn and evolve from the data they produce.
  • Neural networks (NNs) are a type of machine learning algorithms.
  • Deep learning (DL) is the architecture of neural networks, one approach to constructing and training.
  • Machine learning is represented by 4 main types, each with advantages and disadvantages, applications, and popular algorithms.
  • Classical learning can be supervised (classification, regression) and unsupervised (clustering, dimension reduction, pattern search).
  • Ensemble methods are represented by stacking, bagging, and boosting.

FACT

The United States and China have collaborated most on AI research over the past 12 years. Pattern recognition, machine learning, and computer vision are the most popular topics for AI research.

Estimated training cost of select large language and multimodal models (in millions of dollars), 2022


  • Large language models are getting more extensive and more expensive.
  • GPT-2, released in 2019 and considered by many to be the first large language model, has 1.5 billion parameters and costs $50,000.
  • One of the flagship language models, PaLM, released in 2022, has 540 billion parameters and costs around $8M. PaLM is about 360 times larger than GPT-2 and costs 160 times more.

The number of incidents involving the misuse of AI is growing rapidly. According to the AIAAIC database, which tracks incidents involving the ethical misuse of artificial intelligence, the number of incidents in 2021 has increased by a factor of 26 since 2012.


Analysis of legislative documents from 127 countries showed that the number of bills containing "artificial intelligence" passed into law increased from 1 in 2016 to 37 in 2022. Analysis of parliamentary documents on AI in 81 countries also showed that the number of mentions of AI in global legislative procedures has increased almost 6.5 times since 2016.

Importance of AI solutions to organizations overall success worldwide in 2022



  • In 2022, many business leaders (94%) said that AI will be critical to their company's success in the coming years.
  • Critical success is the difference between a company's competitiveness and performance in the coming years.

AI is used by businesses in many different ways. Overall, the most common uses of AI include process automation (39%), machine vision (34%), text recognition (33%), and virtual assistants (33%).


In 2022, the most common uses of AI were service optimization (24%), creating new products based on AI (20%), customer segmentation (19%), customer service analytics (19%), and product enhancements (19%).

Importance of AI solutions to organizations overall success worldwide in 2022


  • While the adoption of artificial intelligence significantly increased from 2017 to 2018, it has leveled off considerably since 2019.
  • This is primarily because, by 2022, AI had already been recognized as an indispensable tool for optimizing business operations.

FACT

In 2022, Stanford University AI Index Report respondents cited cybersecurity as the most significant risk to AI adoption. Regulatory compliance, personal privacy, and accountability were the next most frequently cited risks.

  • In every sector in the US for which data are available (except for agriculture, forestry, fishing, and hunting), the number of AI-related job postings has increased on average from 1.7% in 2021 to 1.9% in 2022.

  • In 2022, the top three countries for jobs requiring AI skills were the United States (2.1%), Canada (1.5%), and Spain (1.3%).

Global AI Market Segments and Quantitative Metrics

AI & ML market size estimate ($B) by segment



  • The artificial intelligence and machine learning market reached $197.5B in end-user spending in 2022, led by the vertical applications segment and with significant contributions from semiconductors and autonomous machines.
  • The global AI & ML market is projected to reach $400.0B by the end of 2025 with a CAGR of 103%.
  • In 2023, the total number of AI financings and new AI companies funded declined.
  • However, overall investment in AI has increased significantly over the past decade.
  • There were 6.1 times more deals involving AI companies in 2023 than in 2014.

Investments in AI, $ billion

Private investment in AI by geographic area, 2023, $ billion



  • The United States leads the world in total private investment in AI.
  • In 2023, the amount of investment in AI in the U.S. was about 2 times greater than in the next largest country, the UK.

§The number of venture deals across the AI & ML vertical amounted to 7,063 in 2023, growing at a CAGR of 20.3%. The total volume of venture investments in the AI & ML market amounted to $80.74B in 2023.

AI & ML market size estimate ($B) by segment

The AI & ML market is divided into 4 main segments:

Horizontal platforms


  • AI automation platforms
  • AI core
  • Computer vision
  • Natural language technology (NLT)

Vertical applications


  • Consumer AI
  • AI in financial services
  • AI in healthcare
  • Industrial AI
  • AI in IT
  • AI in transportation

Autonomous machines


  • Autonomous vehicles
  • Intelligent robotics & drones
Semiconductors

  • Edge AI software
  • Intelligent sensors and devices
  • AI chips
  • 1
    Horizontal platforms
    • Horizontal platforms empower end users to build and deploy AI & ML algorithms across a variety of use cases. These platforms directly apply scientific AI & ML research advances to commercial applications. Companies in this segment have differentiated AI & ML approaches and are built with AI & ML from the ground up – this is also referred to as AI-first.
    • The Horizontal platforms market was estimated at $32.4B in 2022 and is expected to grow to $75.7B by 2025 with a CAGR of 32.7%.
  • 2
    Vertical applications
    • Vertical applications in AI & ML address specific problems within industries and are not always AI-first. These solutions typically differentiate based on the quality of the dataset used to train the industry-specific model and the industry expertise of the data scientists identifying decision-making areas that AI & ML models can enhance. As a result, many of these startups help automate specific functions within their industry but cannot cross-apply their AI & ML to other industries.
    • The Vertical applications market was estimated at $94.0B in 2022 and is expected to grow to $212.2B by 2025, with a CAGR of 31.2%.
  • 3
    Autonomous machines
    • Autonomous machines can perform tasks in human-present environments without explicit human control. These machines synthesize ML, computer vision, and datasets of the physical world, such as navigation. The segment requires the design of complex hardware with software “brains”.
    • The Autonomous machines market size was $41.1B in 2022 and is expected to grow to $58.1B by 2025 with a CAGR of 12.2%.
  • 4
    AI semiconductors
    • AI semiconductors have gained widespread adoption as AI & ML model training, and inference require hardware with maximum computational efficiency and customized processing for AI calculations. AI requires a high volume of contemporaneous calculations that run in parallel and benefit from specialized chips. Using general-purpose chips can cost thousands more than using an AI chip for these purposes.
    • The AI & ML semiconductors market was $43.6B in 2022 and is expected to grow to $72.7B by 2025 with a CAGR of 18.6%.

In terms of operational and financial performance, product development stage, volume of raised capital, investor base, as well as other indicators, the following tech companies are leaders in their segments:

  • By now, the AI ​​industry has separated itself from the traditional IT sector. However, it continues to grow by 30-50% per year. Large AI companies have reached billion-dollar valuations and offer their solutions to a wide range of corporations from various industries around the world.
  • The segmentation and profiling of companies within the AI ​​industry will intensify in the coming years due to the introduction by corporate clients of new services based on or employing AI; similar to the traditional IT sector, the number of “AI for AI” companies (companies providing specific services and focusing exclusively on AI companies as clients) will increase.

Notable AI Tools Examples

  • Chat GPT

    ChatGPT is an AI language model designed for interactive and informative conversations. It is developed by OpenAI and trained on extensive text data to provide helpful information, answer questions, and assist with various tasks.


    Key metrics:
    • Accuracy rate: ~85%
    • Users: 100M+

    Description:
    • ChatGPT is powered by GPT-3.5, an advanced LLM developed by OpenAI.
    • It leverages deep learning techniques, including transformers (175B parameters), to understand and generate human-like text. For this purpose, it was trained on a massive chunk of data (300B tokens) and fine-tuned.
    • A straightforward interface, similar to messenger interfaces, greatly influenced the speed of adaptation and popularity of the solution.
    • OpenAI generates income from ChatGPT through API fees ($0.002 per 750 words) and a subscription plan ($20/month), offering priority access, faster response times, early feature updates, and commercial license fees.
    • It has attracted a large number of partners, including Duolingo (for a particular subscription plan called "Duolingo Max"), Stripe (for fraud detection and customer support), Bing (for a search engine), and more.
    • The main competitors are Big Tech companies and their LLMs: Google's LaMDA, PaLM, and Facebook's LLaMA. These models have similar architectures. While some outperform ChatGPT technically, none have gained popularity.
  • Midjourney

    Midjourney has developed a text-to-image model for Discord. Users can make image requests accompanied by text and choose from four options.


    Key metrics:
    • Website traffic, march '23: 41.4M
    • Active users: 1.1M+

    Description:
    • Midjourney uses a diffusion model that generates new images and audio by adding noise, reconstructing with a neural network, and learning from data. It uses a text query and diffusion steps to create images within the training data distribution.
    • The company works to improve its algorithms, releasing new versions of the models every few months.
    • On March 30, 2023, Midjouney closed access to the free trial. Now, only paid plans are available. The service offers three subscription levels: Basic ($10/m), Standard ($30/m), and Pro ($60/m). The main difference is the number of generated images and additional features.
    • DALL-E from OpenAI and Stable Diffusion from the Stability.ai studio group are the main competitors. Although they have similar functionality and image quality, they possess different interfaces, with Midjourney's main advantage being its unique interface.
  • Waymo

    Waymo, Alphabet's subsidiary, pioneers unmanned vehicle technologies. The company's two main products are the ride-hailing service Waymo One and commercial cargo transportation Waymo Via.


    Key metrics:
    • Disengagement rate: 0.126
    • Miles, without human assistance: 1M+

    Description:
    • Autonomous taxis are available in three cities: Phoenix, San Francisco, and Los Angeles.
    • Waymo's fleet comprises all-electric Jaguar I-PACE vehicles with lidars, sensors, and the advanced VectorNet system.
    • The VectorNet system addresses the limitations of previous neural models by representing information as vectors instead of pixels. This approach eliminates irrelevant data and captures previously overlooked relevant data.
    • Waymo has partnered with companies such as Stellantis, Mercedes-Benz Group AG, Lyft, AutoNation, Avis, and Intel.
    • Competitors in this field include Cruise (GM's autonomous vehicle division) and Tesla, which focuses on selling self-driving cars rather than providing a mobility service.
  • DeepL

    DeepL is an online machine translation service by DeepL GmbH. DeepL provides translation services using convolutional neural networks. DeepL is available as a web-browser version and desktop and mobile applications for iOS and Android.


    Key metrics:
    • 29 languages and 700 language pairs
    • Over 20,000 companies worldwide use DeepL to communicate with team members and customers

    Description:
    • DeepL offers free translation of texts up to 5000 characters. With over 500,000 customers, DeepL Pro includes a Starter plan (€7.49/m) and an Advanced plan (€24.99/m). A paid subscription called "DeepL Pro" is available for professional translators, companies, and developers.
    • Commercial customers can utilize the paid DeepL API to integrate DeepL into their software, streamlining business processes and enhancing customer satisfaction, such as instant translation of international service requests.
    • DeepL participates in blind translation quality testing competitions alongside Google, Microsoft, and Facebook. Compared to major technology companies, DeepL Translator outperforms in terms of translation quality, thanks to advancements in network architecture, training data, training methodology, and network size.
  • Looka

    Looka, a graphic design platform, utilizes artificial intelligence to generate unique logo options and branding by analyzing user preferences and design trends through machine learning algorithms.


    Key metrics:
    • Logos made: 10B+
    • Users: 20M+

    Description:
    • With over 1,000 design rules, Looka offers a streamlined process where users answer questions to tailor their logo. The platform also provides visual identity and branding tools, including website design and social media layouts.
    • No design skills are required, and users receive a complete set of logo files for different formats and platforms.
    • Looka monetizes its services through a subscription-based model. Users pay for different access levels and features based on their chosen subscription plan.
    • Looka competes with other AI-based logo developers, notably Canva. What sets Looka apart is its extensive collection of fonts and icons and a straightforward algorithm that enables the creation of unique logos without relying on pre-made templates. This gives Looka an edge in delivering personalized branding solutions to its users.
  • Pictory

    Pictory is an AI platform that automates video content creation. It uses vast resources like voices, articles, visuals, and scripts to customize and create videos efficiently. Pictory condenses long videos into summaries and offers voice-over options. SEO parameters are adjustable, optimizing titles, descriptions, and timings. It's an all-in-one solution for seamless video creation and promotion.


    Description:
    • Pictory is a versatile professional platform offering features like clip creation, translations, and transcriptions.
    • Users can choose from subscription plans: Standard ($29/m) with some limitations, Premium ($59/m) with more capacity, and Enterprise (custom) for larger organizations. The free version is available for three uses.
    • Pictory's main competitor is InVideo, a cloud-based video creation platform that uses AI technology to enable users to easily and quickly create professional, high-quality videos.
  • Melobytes

    Melobytes is an AI-powered online platform that enables users to explore their creativity in music and visual media. It offers intuitive tools for creating and editing music, making music composition accessible to all, regardless of musical background.


    Description:
    • Melobytes is ideal for individuals without formal training or access to traditional instruments, including those with disabilities.
    • It caters to a diverse user base, including amateur musicians, students, content creators, and professionals seeking a convenient way to compose music.
    • Melobytes operates on a subscription model, with options for 7-day, 1-month, and 1-year packages, granting unlimited access to all apps and additional perks.
    • Key competitors in the field include AIVA and Soundraw.
  • GitHub Copilot

    GitHub Copilot is an AI tool by GitHub and OpenAI that assists developers in writing code. It leverages the OpenAI Codex model, a descendant of GPT-3, to understand and generate code based on user input. It offers comment analysis, code completion, and defining algorithms. The training dataset includes natural language and billions of lines of source code from public GitHub repositories.


    Description:
    • While the generated code quality is considered "average," fine-tuning with competitive programming and continuous development datasets improves accuracy.
    • GitHub Copilot is popular among developers working on code-intensive projects. The product has already proven its worth with impressive business cases. For instance, GitHub Copilot accelerated Duolingo engineers' development speed by 25%, while Elanco employees' onboarding time decreased by 1,000 days thanks to GitHub Copilot.
    • It offers two pricing plans: $10 for individual users and $19 for businesses.
    • Main GitHub Copilot's competitor, Tabnine, is a powerful code completion tool leveraging AI to provide intelligent suggestions and snippets as you code. Through extensive analysis of open-source code, Tabnine learns patterns and practices, offering personalized and precise suggestions that align with your coding style.
  • AlphaGo

    AlphaGo, developed by DeepMind, is a revolutionary computer program for playing Go. It made history by defeating a professional Go player and a world champion, marking a significant milestone. Go is an exceptionally intricate game, with more positions than atoms in the observable universe. Unlike chess, Go presents a formidable challenge for computers to conquer.


    Description:
    • AlphaGo's key innovation is its use of deep learning. By training neural networks on professional games, AlphaGo can predict moves made by professionals.
    • It has improved through self-play and reached the top level in 2015, surpassing other programs without relying on extensive search algorithms. DeepMind continues to refine AlphaGo's algorithms and release new versions.
    • Other notable AI solutions in gaming include OpenAI Five, a Dota 2 AI team that has won a world champion title, and Libratus, an AI developed by Carnegie Mellon University for poker. Libratus utilizes "regret minimization" to learn and enhance its gameplay.

Artificial General Intelligence (AGI)

Artificial General Intelligence (AGI), or Strong AI, is a theoretical description of a particular form of AI with human-like intelligence, self-awareness, and the ability to solve a wide range of intellectual problems, learn, and plan future actions.


The 3 types of Artificial Intelligence:

  • Artificial Narrow Intelligence (ANI, Narrow AI) – specializes in one area, solves one problem
  • Artificial General Intelligence (AGI, Strong AI) – can perform most of the tasks that a person is capable of
  • Artificial Super Intelligence (ASI) – surpasses the intelligence of any person and can solve complex problems instantly

Key characteristics and capabilities


A true AGI must be able to perform human-level tasks and abilities that no existing computer can achieve. Today, AI can perform many tasks, but not at the level of success that would allow it to be attributed to human or general intelligence.


The core concepts of AGI include:

  • Autonomy: AGI systems can operate independently, making decisions and solving problems without human intervention.
  • Learning: AGI can acquire knowledge and learn from experience, adapt to new situations, and improve its performance over time.
  • Generalization: AGI can apply knowledge and skills acquired in one domain to other unrelated domains, demonstrating flexibility and adaptability.
  • Creativity: AGI possesses the ability to generate novel ideas, discover new patterns, and solve complex problems in innovative ways.

Future prospects:

  • The timeline for AGI development remains a subject of ongoing debate among researchers and experts.
  • Some argue that it may be possible in years or decades, others maintain it might take a century or longer, and a minority believe it may never be achieved.
  • Additionally, there is debate whether modern deep learning systems, such as GPT-4, are an early yet incomplete form of AGI or if new approaches are required.

Download the full version of the report

I am interested in:
By clicking the button, you consent to the processing of personal data and agree to the privacy policy.