Industry Spotlight. AI & ML

In addition to influencing pop culture and apocalyptic predictions of a machine uprising, AI & ML have a significant place in the world of venture investments

In March 2023, Bill Gates called artificial intelligence a revolutionary development that would be on par with personal computers, smartphones, and even the Internet. His statement was prompted primarily by the success of generative AI. The topic is being shouted from all sides, and it is impossible to ignore it without a great deal of willpower. However, AI and ML are not new to venture capital investing.

Over the past few years, investors have poured money into the field, hoping to make a profit and be part of the sci-fi revolution. For a long time, the industry lacked the push it needed to spread the technology and create real-use cases. Many products did not find product-market fit, were the amusement of scientists, or simply did not provide enough impact to warrant the resources to implement them. In addition, in most cases, AI & ML were not stand-alone products, but an add-on to various projects. But things have changed.

Today, AI and ML are at the forefront of the attention of venture investors, who are doing their best to catch up. We are no exception, so we invite readers to the epicenter of the trends: the AI & ML Industry Review column.

The topic of AI & ML is complex and goes far beyond the realm of business. Stanford University, for example, took 400 pages to cover the subject. However, we want to focus on the venture component. This article should help angels get up to speed on the business from the ground up, as well as provide important and relevant information with numbers from this dynamic industry.

And now, at the initiative of ICLUB, an international network of angel investors founded by TA Ventures, we took part in the creation of an article that will help angels understand the topic of AI and ML investments from scratch, as well as look at the numbers and dynamics of the industry at a glance. 

We also recommend reading their Angel Staff newsletter, which helps to navigate through the complex world of startups and their perspective from the side of business angels. Now, let's move on to the topic of the episode.

What is AI, and ML, and what is their difference?

AI and ML are not new technologies – they were invented in 1951 and 1952 respectively, and not everyone has built up a solid understanding of these terms.

Artificial Intelligence (AI) is a field of computer science that involves the creation of software that can perform tasks that would normally require human intelligence. For example, learning (Elsa, accent training app), problem-solving (Jasper, AI copywriter), decision-making (AlphaGo, gaming tool), and pattern recognition (DeepText, social media pattern search engine).

The meaning of machine learning is derived from its name, specifically the method of teaching computers to learn and make predictions based on data. Traditionally, developers write code that tells a computer how to perform a specific task. With machine learning, however, developers can train the computer to recognize specific interconnections and algorithms. In this way, the computer learns to recognize patterns on its own. The most popular examples are Watson, PyTorch, Apache Mahout, and Open NN.

The difference between them is easy to see. In Machine Learning, developers create Artificial Intelligence with models that can “learn” from data patterns without human interaction. So, AI grants the machine the capability to independent reasoning, while machine learning is a tool that helps AI to achieve this.

Here's a simple example to illustrate the difference: imagine you want to create a system that can distinguish between pictures of cats and dogs. Using AI, you could program a set of rules to help the computer make this determination. However, this would require you to explicitly define what a cat or dog looks like, and the computer would have to follow those rules exactly. Thus, you can use Machine learning.

You would need to feed the computer a large dataset of labeled cat and dog pictures and let it learn the patterns and features that distinguish the two animals on its own. This would enable the system to make predictions about new images it hasn't seen before and continuously improve its accuracy over time, which can be considered independent intelligence.

In short, Artificial Intelligence is a broader field that encompasses the idea of a machine that can mimic human intelligence. In contrast, Machine learning is a specific technique used to achieve artificial intelligence.

AI and machine learning is being used in almost every industry, including healthcare, finance, transportation, and manufacturing, primarily to automate processes, reduce errors, and improve efficiency. They can revolutionize the way we work and live, increasing economic output by 38% by 2035 and leading to $14 trillion in growth across 16 industries in the 12 most developed economies.

The Market

The AI & ML sector is now attracting increased attention, but even without this surge, the industry has never been an outsider. Even without the generative AI craze, investment in startups has been growing steadily year after year. 

In 2022, investors closed 2,956 deals worth $45.8 billion, according to CB Insights. That's down 34% from 2021, but higher than any previous year. The average deal size was $20.7 million. The top geography was the United States, which accounted for the majority of deals and dollars. 66% of all investments went to early-stage startups.

Analysts at PitchBook did similar research. According to their information, investors closed 6727 deals worth $78 billion in 2022. The most affected segments were autonomous vehicles, processor development, automation platforms, and smart sensors. As noted, an interest in hardware has declined. And what is on the rise is the consumer segment, which includes generative AI.

A recent major study by Stanford University reports that global private investment in AI was $91.9 billion in 2022, a 13-fold increase from 2013.

Despite the differences in some figures, all sources confirm the positive trend of investments in the sector. This factor should have a positive impact on the industry's future, even after the hype dies down. If we talk about the global size of the market, it is about $200 billion, says Grand View Research. Moreover, it is projected to grow at a CAGR of 37.3% from 2023 to 2030.

Separately, let's talk about the elephant in the room – generative AI. PitchBook estimates the current market size of this segment alone at $42.6 billion and predicts it will grow to $98 billion by 2026. Here, investment has already tripled to $2.3 billion in Q1 2023, according to Axios Pro. In general, the train is fueled and running at full speed. 

Still, some observers are skeptical of the novelty, noting some characteristics of a financial bubble. Generative AI could suffer the same fate as Web3 with cryptocurrencies. Here, the demand exceeds the supply. 

There are extremely few projects on the market to invest in, forcing people to throw money everywhere. For example, according to CB Insights, only 110 deals were registered in 2022, but for $2.5 billion. Such activity will eventually lead to inflated company valuations and the presence of unprofitable projects in the market.

Either way, generative AI has become the real engine of an entire industry. Because of this monster, global spending on AI-enabled systems will reach a record $154 billion by 2023, according to IDC. Everybody wants a new toy.

Industry Developments. What should you know first?

When writing this unit, we relied on a Stanford University study that can be called, without overstating the case, one of the most authoritative documents on AI in recent times.

First, the vast majority of AI startups are located in the United States. For the industry, the United States is a key geography that has held a strong lead for years, both in terms of capital and the number of deals. North America is outpacing the competition for a number of reasons. In China, for example, there are regulatory and legislative issues. The Chinese AI sector has been virtually frozen lately. In Europe, there is another reason: a lack of equipment and capacity for computing.

The number of companies adopting AI for their own needs continues to grow. A global study showed that in 2022, around 50-60% of companies have adopted AI in one way or another. In 2017, for example, there were only 20% of them. Interestingly, this is not happening for the sake of trends. Businesses are confirming that the adoption of AI has resulted in lower costs and higher revenues.

In terms of cases, companies most often use AI to optimize service operations, create new products based on artificial intelligence, segment customers, analyze customer behavior, and improve products, or in other words, integrate AI into them.

Now for startups. The companies attracting the most investment are those developing AI solutions for medicine and healthcare, cloud solutions, data processing and management, fintech, cybersecurity, privacy, and retail.

The least amount of capital is invested in facial recognition, ed tech, fitness and wellness, geospatial, legal tech, entertainment, and agritech.

When we talk about trends in AI, it's important to remember that in most cases, AI itself is not a standalone product. Rather, it is an extension that significantly impacts business and even changes the rules of the game. Each industry uses AI differently, taking into account specific tasks and other requirements. 

For example, in biotech, artificial intelligence and machine learning can help with drug discovery or finding new types of proteins. Therefore, it is appropriate to talk about AI in terms of a specific context, i.e., an industry.

Industry developments. What the press is saying?

We scoured popular media publications for 2023 and selected the most frequently mentioned predictions. Sources include TechTarget, Fireflies, Emeritus, and Dataversity.

Automated machine learning (AutoML), namely improved tools for data labeling and automatic tuning of neural network architectures.

Conceptual design. Bringing AI to the creative sector on an industrial scale. For example, if you give AI the input text “avocado chair,” it will design a chair in the shape of an avocado.

Multimodal learning. This is the ability of AI to analyze multiple sources of data simultaneously. For example, text, vision, speech, and IoT sensors.

Models that can achieve multiple goals. Not to be confused with the previous point. In this case, the AI analyzes one or more data sources but considers multiple goals instead of just one to produce results. For example, profit growth plus environmental compliance. In the end, it comes up with the optimal solution.

Cybersecurity. New AI and machine learning techniques will play an increasing role in detecting and responding to cyber threats.

Improved language modeling. It is predicted that as generative AI grows, inaccuracies and even dangerous conclusions will increase. This will force more time to improve the technology.

Computer vision. Cheaper cameras and AI are expected to lead to a boom in computer vision for analytics and automation in the coming year.

Democratization of AI. Improvements in tools are reducing the level of expertise required to build AI models. This will make it easier to include domain experts in the development process.

Eliminating errors in ML. The goal is for AI to make predictions objectively. For example, people may now be discriminated against when applying for credit, buying goods online, or receiving medical care.

Digital twins. Their capabilities allow data to be used and predicted in new ways. For example, modeling the disease process, testing new drugs, customer behavior, and the economic impact of certain decisions.

Generative AI. Well, everything is clear here. It is noted that technology has allowed AI to go beyond being a tool for automating routine tasks. Now AI is a creative instrument and much more. For example, the technology can be used to produce synthetic data.

Collaboration between humans and AI is becoming the norm. From writing code, automated testing, deployment, and troubleshooting to creating content and automating routine tasks.

According to Gartner, the use of AI in software development will increase significantly by the end of the decade. By 2030, 80% of workers will interact with intelligent AI tools on a daily basis, up from less than 10% today.

The evolution of conversational AI. These trends include personalizing interactions, using no-code to reduce the burden on IT, and incorporating natural language processing, machine learning, and sentiment analysis to understand user intent.

Focus on ethical AI. AI training requires data, and that often means personal data. If people don't trust the AI or understand how it's making decisions, they won't share their data, and it all falls apart.

Quantum computing + AI. As the amount of data generated and stored grows, so does the need for more powerful and efficient computing technologies. Quantum computing has the potential to transform many industries, including finance, healthcare, retail, and logistics.

Low-Code & No-Code AI. Greater involvement of non-technical personnel in the world of AI will democratize its use across industries and businesses of all sizes.

Advanced Analytics. Allows AI to assist in data preparation and idea generation, applying machine learning (ML) and natural language processing (NLP) techniques to complement traditional data research methods.

Edge AI. This is the next frontier of computing technology that will decentralize the entire data analysis process. For simplicity, data sources will be equipped with the infrastructure necessary to process data instantly as it appears.

Increased use of cloud services. With the development of AI and machine learning-based cloud software, companies can track and analyze massive amounts of corporate data in real-time and make necessary adjustments to their business processes.

Business Intelligence (BI). Using advanced artificial intelligence and ML tools, today's BI platforms are able to maximize the value of correlations, trends, and patterns in data.

The rise of data-as-a-service (DaaS). Increasingly, data is being generated specifically for sale to stakeholders.

The Elephant in the Room – Generative AI

On November 30, 2022, Open AI released ChatGPT, a newer version of its language model that uses Generative AI. This advancement brought enormous interest to AI, and ChatGPT has become the fastest application that reached 100M users, outpacing TikTok, Instagram, Pinterest, and Spotify. But why is Generative AI such a hot topic, and why do users flock to it?

Generative Artificial Intelligence (such as ChatGPT) refers to the ability to generate new content across various mediums, such as audio, code, images, text, simulation, and video, after being trained on large amounts of data. Basically, Generative AI allows users to create any piece of digital content with significantly less effort.

Creating a Generative AI model has typically been a significant undertaking, to the point where only a select few technology powerhouses with ample resources have attempted it. OpenAI, the company responsible for developing ChatGPT, previous GPT models, and DALL-E, has received billions in funding from prominent donors. Google and Microsoft compete for the best search engine, with their Bard, and Bing that have similar technology. DeepMind (which is a Google subsidiary) and Meta, have already launched their Make-A-Video offering, which utilizes Generative AI. These companies employ some of the world's top computer scientists and engineers.

However, it's not just about talent. Training a model using nearly the whole internet's data comes with a high cost. Although OpenAI has not disclosed exact figures, estimates suggest that GPT-3 was trained on approximately 45 terabytes of text data, equivalent to about one million feet of bookshelf space, at the cost of several million dollars. These are resources that an average startup cannot access.

The use of Generative AI tools has the potential to generate an unlimited amount of entertainment and produce a diverse range of credible writing within seconds. And the results of this generation can be further improved. This has significant implications for a wide range of industries, including IT and software organizations that can benefit from the instant production of largely accurate code and organizations that require effective marketing copy. 

Essentially, any organization that needs to generate clear written materials can benefit from these tools. Additionally, organizations can utilize Generative AI to produce more technical materials, such as higher-resolution versions of medical images. Organizations can pursue new business opportunities and create greater value by saving time and resources using such tools.

Main Challenges for Business

Startups, as well as their customers, are concerned about a number of factors hindering the adoption of AI. According to the Stanford survey, respondents are most concerned about AI risk management (50%). This relates directly to the issue of regulation, as AI will undoubtedly be subject to some form of restriction in the future. And when it does, it is not clear what can be implemented and what will be prohibited.

Finding the data needed to train AI models remains a sore point (44%). Either there is not enough data, or there is enough data, but it is running out.

37% of businesses are still looking for proof of value from AI (37%), and 33% cannot choose the right product and technology. 42% are having difficulty implementing AI.

Advice for Investors

Though the appetite for AI business is significant, the real number of companies that actually somehow use AI/ML technologies is much smaller than the number of companies that use the term “AI” as a marketing tool.

Firstly, investors should conduct the investment due diligence accurately to identify whether the company really builds AI. In fact, 2830 European companies claim to make use of AI. However, a surprising 40% of them are not using any type of AI at all. It’s a common issue when investors just follow the trend with no rational considerations. 

As an example at the beginning of 2023, investors piled capital in relatively obscure public firms with AI in their names like BigBear.ai (300% price increase), C3.ai (100% price increase), BuzzFeed (100% price increase), and Guardforce (51% price increase).

It is important to understand what the startup’s product is doing, how it leverages AI/ML technology, and how advanced it is compared to its past approaches that are solving the problem now. Investors should also pay specific attention to cost structure and tech defensibility.

Secondly, investors should learn to spot differences between basic data analysis and Artificial Intelligence. Several SaaS and automation companies are positioning themselves as AI brands, however, all they really do is use data analytics to orchestrate applications and workflows. Their technology doesn’t get more intelligent over time, providing better analysis and better insights.

It is crucial to find out whether the company uses AI or basic data analysis and how it creates its own data exhausts (a large trail of proprietary data that they collect from interesting sources). Besides, watch for how it uses Machine Learning to make the product smarter and whether it reduces the need for a human in advancement loops.

Finally, investors should pay additional attention to the core technological research. Some companies do not have any technology at all, and just use a lot of manual labor from low-income countries, which is portrayed as AI to attract investors. Such investor deceiving techniques are rather rare than common, though it is still an existing issue that requires attention.

The Roosh Ventures and ICLUB Experience

We write about AI and invest in it. Everything is fair, so we propose to look at some real cases. The first three relate to Roosh Ventures, the rest to the club.

• Cresta. An all-in-one AI platform for modern contract centers.

• Alter. A digital AI avatar company.

• Regression Games. A full-featured AI platform for gaming developers and tournaments.

• Vochi (acquired by Pinterest). A professional video editor in your smartphone with an AI-based computer vision algorithm.

• ResearchRabbit. An app that uses AI technology to find similar material by topic, citation, and popularity to make it easier to conduct research in different fields (medicine, finance, law).