AI Industry Research Report: Generative audio will soon break out after generative text?

Before the global craze broke out, artificial intelligence had been brewing for many years. To properly predict future AI innovations, we went back to the source and examined patent and venture capital deal activity for AI applications.

We collected 175,072 artificial intelligence patents published by 193 World Intellectual Property Organization (WIPO) members between 2012 and 2022, divided into five major categories. For venture capital investments, we used data on 24,310 transactions in 92 economies from 2012 to 2022 from the OECD Artificial Intelligence Database.

In absolute terms, venture capital activity and the number of patents in the field of artificial intelligence have surged since 2012. During this period, the number of venture capital transactions increased tenfold to 3,884, and the value of transactions in 2022 is almost 50 times that of 2012, reaching $83 billion. At the same time, the number of artificial intelligence patents increased sevenfold in 2022, reaching nearly 37,000.

Judging from venture capital transactions and patents, more than two-thirds of artificial intelligence innovation in the past 10 years has been concentrated in applications in transportation, industry, and consumer industries.

Next, we expect that generative audio will appear on a large scale in 2024 and disrupt industries such as gaming and film production.

Artificial intelligence comes out of hibernation

With the release of OpenAI’s ChatGPT in November 2022, interest in artificial intelligence has surged, not just among ordinary people, but also among enterprises. In 2022, "artificial intelligence" was mentioned more than 715,000 times in company documents, compared with only 135,000 times in 2020. According to our survey, by April 2023, 41% of Americans have heard of ChatGPT, and nearly 60% say their work already uses ChatGPT to some extent.

Additionally, in news on May 18, ChatGPT is now available on iPhone after releasing its first app version in the United States. Google also announced on May 5 that it would begin rolling out generative artificial intelligence tools integrated into its search products. Before going further, let us briefly define artificial intelligence - artificial intelligence is a machine-based system that, given a given set of goals by humans, can make predictions, recommendations, and even decisions that may affect reality or virtual environment.

Artificial intelligence has actually been brewing for several years. AI-related venture capital deal activity and published patents have been quietly surging since 2012. For example, the number of venture capital investments increased from 332 in 2012 to 3,884 in 2022. In 2022, the value of venture capital deals will reach $83 billion, up from $1.8 billion in 2012. At the same time, the number of patents in artificial intelligence has increased sevenfold since 2012.

The barriers to entry in the artificial intelligence industry have been lowered, which has stimulated business innovation in the field. For example, since 2018, training costs for image classification systems have been reduced by 64%, while training time has been reduced by 94%. Many startups and venture capital firms have also emerged, including OpenAI, Anthropic, Stability AI, A121 Labs, Midjourney and Cohere. Anthropic reportedly recently raised $450 million in Series C funding, with participation from Alphabet, Salesforce and Zoom. Additionally, Builder.Ai, a London-based artificial intelligence company, raised $250 million in its latest round of funding.

In order to correctly predict what is coming in artificial intelligence, we also studied patents in this field. We estimate that it will take approximately two to three years for a venture-backed company to bring a product to market, and even longer, such as seven to ten years, for patent approval. As far as patents are concerned, after a company's patent is protected, they need to market, select business partners, etc., which takes about one to three years.

Through Patentscope of the World Intellectual Property Organization, we collected 175,072 artificial intelligence patent entries published by 193 members of the organization between 2012 and 2022. These patents are divided into five major categories, including sectoral applications, service platforms (horizontal platforms), automated machines (autonomous machines) and semiconductors (semiconductors). On the venture capital side, we used 24,310 deals across 92 economies covered by the OECD AI database.

1. Artificial Intelligence Landscape - Bringing Creativity to Life

How did the stock market react?

Stock prices of companies that quickly adopted ChatGPT and other new artificial intelligence surged. For example, on January 31, 2023, the artificial intelligence software company C3. ai launched a "generative artificial intelligence product suite" that integrates OpenAI, Google, academic research and other artificial intelligence technologies. The company's stock price rose nearly 22% that day, and has risen 143% since the beginning of this year.

Shares of SoundHound, an artificial intelligence audio and speech recognition software company, are up 66% year to date.

Not just artificial intelligence software companies, artificial intelligence affects almost all companies in the value chain. Shares of Nvidia, a semiconductor company that makes computer chips that can run artificial intelligence models, rose 110%. Meta also performs well in the artificial intelligence race among big tech companies.

Funds flowing to artificial intelligence

In 2022, global venture capital investment in artificial intelligence will reach a peak of US$83 billion, up from US$1.8 billion in 2012. Databricks and MIT found that most companies are already deploying AI broadly. In the sample, the proportion of companies not using artificial intelligence by 2022 is less than 6%.

Businesses Will Benefit from Artificial Intelligence Growth

For many years, artificial intelligence research has been limited to academia, but now it is being applied in the commercial field. Academia is the first mover, and it usually takes 15 years from academic publication to obtaining a patent.

However, it is becoming increasingly clear that the time is ripe for an explosion of artificial intelligence innovation. From 2012 to 2022, 49% of venture capital deals were completed in the past three years. Likewise, 52% of AI-related patents in our database were issued within three years.

2. Application fields: Artificial intelligence will become mainstream in 2023

Amazon founder and executive chairman Jeff Bezos said: "We are now using machine learning and artificial intelligence to solve problems, and what is happening now is like something described in a science fiction novel."

In the past 10 years, nearly four-fifths of the venture capital transactions related to artificial intelligence were in the field of industry applications. Another 8% occurred in automated machines and automobiles, and 13% occurred in semiconductors. The proportion of AI-related patents is similar.

Industry Application: Artificial Intelligence is not limited to one industry

Currently, 79% of venture capital deals and 61% of patents occur in what we consider “industry applications.” By our definition, this includes consumer, industrial, information technology, transportation, healthcare, and financial services.

From 2012 to 2022, the number of venture capital transactions increased from 270 to 3,006, an increase of more than 10 times. In 2022, the total transaction volume will reach US$62 billion, much higher than the US$1.3 billion in 2012. Patents issued in these areas are also booming. Since 2012, the number of relevant patents has increased sixfold. The companies with the most patents are recognized technology giants such as IBM, Samsung, Intel, LG Electronics and Qualcomm.

Service platform: AI tool

Since 2012, we have seen the number of patents for horizontal platforms increase nearly 6 times, including artificial intelligence core (AI core), natural language technology (natural language technology), artificial intelligence automation platform ( AI automation platforms), computer vision software. Service-oriented applications (Horizontal applications) are key modules in the development and deployment of artificial intelligence. Companies with such patents include AT&T, IBM, LG Electronics, Baidu and established technology companies.

Natural language processing (Natural language processing) is an artificial intelligence tool for large language models (LLMs) that became popular with the emergence of ChatGPT. Some examples of natural language processing include translation, automatic error correction, automatic text summarization, chatbots, and processing of information needs in the financial sector.

Computer vision involves using artificial intelligence to analyze visual data and make meaningful predictions, including facial recognition, geospatial analysis, and more.

Microsoft’s Inner Eye technology can already help detect tumors and abnormal cells. In their latest research, Microsoft shows how its technology can help clinicians. Among the five companies with the largest number of computer vision patents, two are Chinese companies, namely Baidu and Huawei.

Who needs a driver’s license? Autonomous driving has arrived

Over the past decade, 8% of venture capital deals and 16% of patents have occurred in the field of "automated machines." This type of technology covers two key areas, namely self-driving cars and intelligent robots. This area has received a lot of attention over the past decade. This largely coincides with the success of Tesla, which has been experimenting with self-driving car systems.

From 2012 to 2022, the number of venture capital deals in the automated machines field increased from 22 to 363. Intelligent robots fall into this category. These robots are capable of operating to a certain extent without the need for human intervention and input. Artificial intelligence is often used to assist with robot model training and adaptation. Companies such as Samsung, LG Electronics, Intel, and China’s DJI and Didi have strong positions in robotics patents.

For several years, self-driving cars have also been the focus of discussion. In addition to Tesla, other companies are also expanding research and investment in the technologies required for self-driving cars, so the significant increase in venture capital investment and patent applications is likely to occur in this context.

Semiconductor Chip

Nvidia CEO Jensen Huang said: "It is clear that artificial intelligence will impact every industry. I think every country needs to ensure that artificial intelligence is part of its national strategy, and every country will be affected."

Semiconductors accounted for 13% of AI-related venture capital transactions and 4% of published patents, including processor design, edge AI software, intelligent sensors & equipment devices).

From 2012 to 2022, venture capital deals in the semiconductor field increased from 40 to 515. In 2022, the total number of transactions in this field will reach US$9.4 billion, up from US$200 million in 2012. Likewise, the number of patents in this field increased sevenfold from 2012 to 2022.

Since the 1970s, chips have experienced rapid innovation, with chip performance doubling every few years. Producers must keep up with developments or risk being left behind. As technology becomes more complex, new tools are necessary, and this is where artificial intelligence can help.

3. New opportunities in 2024 and beyond

Generative audio could be the next big thing

Artificial intelligence is now able to create human voices and synthesized audio based on text input in different languages, accents and dialects. It uses generative artificial intelligence popular in ChatGPT.

From the first quarter of 2020 to the fourth quarter of 2022, the number of company documents mentioning "generative audio" increased more than 13 times.

As music and sound research and development deepens, we may see an explosion of new competitors. Soon, people will be able to input text or images to generate audio content without needing an audio expert or a computer expert.

This could impact a range of sectors including gaming, communications, music, journalism and healthcare. Companies with the most generative audio-related patents include Sony, Amazon, Huawei, ByteDance, Adobe, Apple and Tencent.

What audio can be generated?

Generative audio is divided into four main categories, including synthetic voice, speech interaction, music generation and audio editing. First, through deep learning, artificial intelligence can synthesize real human voices with high quality, including pitch, tone and rhythm. Voice interaction includes technologies such as virtual assistants, such as Apple’s Siri and Amazon’s Alexa. As technology develops, there may be more and more “natural” interactions between humans and artificial intelligence.

Music generation technology makes it easy to create a wide variety of new music. Examples include Spotify's DJ and Aimi Studio. Artificial intelligence music is already posing a challenge to the music industry.

Finally, audio editors can use artificial intelligence to improve audio recording quality.

in conclusion

Artificial Intelligence is here, and new innovations are just around the corner. As measured by indicators such as patents and venture capital, innovation in the field of artificial intelligence has exploded over the past decade. Judging from venture capital activities and patents, more than two-thirds of artificial intelligence innovation in the past decade has been concentrated in transportation, industrial and consumer applications, followed by automated machines and service applications, and the semiconductor industry.

Beyond 2023, we expect a surge in generative audio innovation. With the arrival of ChatGPT and DALL-E, we have seen generative AI capture the world’s attention. Together, these technologies will transform industries such as gaming and film production.

Author | BT Finance

Source | Deutsche Bank

The above is the detailed content of AI Industry Research Report: Generative audio will soon break out after generative text?. For more information, please follow other related articles on the PHP Chinese website!