Artificial intelligence (AI), which has revolutionized the lives and job relationship for many people, is also challenging how we view to role of AI and intellectual property. There are many things that you should know about AI, whether you are a business looking to create AI systems for your company or if you want to implement AI in your daily life. 

Based on my experience patenting neural network technologies since 1993, this article will provide an overview of artificial intelligence and will explain the impact of AI and intellectual property (IP) rights and how AI is changing the workplace. In our experience, we see AI as an adjunct or “co-pilot” with humans rather than a replacement for human judgment.

In addition, we will address IP issues unique to AI technology patenting AI technology has a host of issues. Despite these obstacles, it is our experience that companies and inventors can obtain valuable patents for AI-related inventions through careful patent drafting, strategic patent prosecution and by working closely with patent attorneys who have experience in the field of AI.

AI inventions Can be patent-eligible

An AI-related invention could include an AI architecture that runs on a processor, a GPU and has training and inferencing processes to allow the model to perform a particular task. This software will be subject to the two-part patent test by the US Supreme Court in Alice Corp.

Step 1 of Alice is to determine if an AI/software-related invention is directed towards an “abstract idea.” This is one type of “judicial exception” to the “inventions patentable”, according to 35 U.S.C. 101 (regarding “patent eligibility”). And because “abstract ideas” comprise “mental processes” and “mathematical concepts,” courts typically analyze software-related inventions (including AI-related inventions) under the “abstract idea” rubric (with other “abstract ideas” being “laws of nature” and “natural phenomenon,” which rarely arise for software-inventions).

If the claims are related to an abstract idea the analysis moves to Step 2, which asks whether an AI-related, or software-related invention still has an “inventive concept.” The court will then examine whether the claims as a whole contain additional limitations that are “significantly more” than the abstract idea itself. If yes, the invention is eligible for patent eligibility.

Although no decisions have been made by the courts on AI patents, the USPTO has provided examples of AI patent eligibility in its Subject Materials Eligibility Examples: Abstract ideas. Example 39 is an example of an artificial intelligence invention that is patentable. Example 39, called “Method To Train a Neural System For Facial Detection”, has claim elements that permit the training of a neural network through two stages. These data will be used for facial detection to reduce false positives. The USPTO analysis shows that Example 39 is patentable, and not directed at an abstract idea. This is because the claim does not cite any fundamental economic concept or mathematical process.

Another example is the recent application of the 2019 PEG by the Patent Trial and Appeal Board (PTAB) in an ex appeal involving artificial Intelligence inventions. Ex parte Hannun (formerly Ex party Linden) covered “systems and methods for the transcription of speech into text”. Based on the two-part Alice test, the Examiner rejected the claims. The PTAB disagreed. The PTAB disagreed. The claims didn’t mention “organizing human actions” as they were focused only on specific implementations that included technical elements like AI and computer speech recognition. The PTAB noted that specification was crucial in describing the claimed invention’s improvements in technical field speech recognition. Specification stated specifically that DeepSpeech Learning (i.e. a trained neuron) as well as a language model “achieves better results than traditional methods on difficult tasks in speech recognition while being simpler. Today, tools like PowerPatent are available to help inventors navigate through the maze of patent eligibility.

The interaction between AI and IP Laws

Obtaining patents for AI-related inventions can be a challenging process due to several obstacles, including:

  1. Novelty and non-obviousness: One of the requirements for obtaining a patent is that the invention must be novel and non-obvious. This can be difficult to prove for AI-related inventions, as they often build on existing technology and may not be significantly different from what is already known.
  2. Algorithm and process as unpatentable: Some countries may not consider algorithms and processes as patentable subject matter, making it difficult to obtain patents for AI-related inventions that rely on these elements.
  3. Lack of clarity in defining the invention: AI-related inventions can be complex, and it can be difficult to clearly define the invention in a way that meets the requirements for a patent.
  4. Prior art search: Searching prior art in the field of AI can be difficult as the field is rapidly evolving and the number of references is increasing exponentially. Additionally, it may be difficult to find prior art that is relevant to the specific invention.
  5. Examining AI-related patents: Examining AI-related patents can be a challenge as the field is rapidly evolving and the technology is complex, which makes it difficult for patent examiners to understand the invention and determine its novelty and non-obviousness.
  6. Interpreting AI patents: Interpreting AI patents can be a challenge, as the technology is complex, and the patent language may not always be clear. This can make it difficult for courts to understand the invention and determine whether it is valid.
  7. Generality of AI Patents: AI-related inventions can be very general in nature, which can make it difficult to draft claims that are specific enough to be patentable.

Despite these obstacles, it is our experience that companies and inventors can obtain valuable patents for AI-related inventions through careful patent drafting, strategic patent prosecution and by working closely with patent attorneys who have experience in the field of AI.

While a patent application involving AI and neural networks can be a complex, here are a few general tips to keep in mind:

  1. Clearly define the invention: The patent application should clearly describe the invention, including how it uses AI and neural networks, and what problem it solves.
  2. Provide a detailed description of the technology: The patent application should provide a detailed technical description of the AI and neural network components of the invention, including how they are implemented, how they interact with other components, and how they achieve the desired results.
  3. Include drawings or diagrams: Drawings or diagrams can help to clarify and explain the invention, and can be very helpful in understanding the technical aspects of the invention.
  4. Describe the advantages of the invention: The patent application should describe the advantages of the invention over existing technologies and how it is novel and non-obvious. The advantages should point out how the invention accelerates computer performance and show how the invention overcomes the Alice factors.
  5. Use appropriate terminology: It is important to use appropriate terminology when describing the AI and neural network components of the invention to ensure that the patent office understands the technical aspects of the invention.

AI and Data Questions

Innovation is good for communities because it creates better goods and services that meet social needs. For example, AI-based medical research may result in new treatments and diagnostic tests that can improve the health of the community.

Despite rapid progresses in large-scale neural networks technology, there are still many problems to be solved. These include:

1. How to determine the correct data set

Data quality and availability are essential for AI to function. The right data sets and trusted sources for relevant data are essential to enable AI to work in a company. This will allow it to implement AI quickly and efficiently. It is impossible to create AI algorithms that use low-quality, inaccurate data. Businesses can contact AI experts to help them overcome the challenges of implementing AI.

2. The bias problem

The quality of AI systems is dependent on data. Good data is crucial for the development of artificial intelligence services. Lack of good data can cause many problems in AI implementation. This includes biases in the output from ML algorithms that are based on discriminatory assumptions and prejudices in training data. Low-quality data can often be associated with racial or gender biases. These biases must be eliminated. It is essential to train AI systems with objective data or to create simple, easy-to-understand algorithms. Companies that develop artificial intelligence invest heavily in control systems and techniques that can be used for trust, transparency, and identifying bias in AI algorithms.

3. Data security and storage

In order to train their algorithms, many artificial intelligence services rely on large data volumes. Although large data volumes may offer greater business opportunities, they can also cause security and data storage problems. Data leakage is more common when there are more users and more data being created. Data security and data storage issues have become increasingly global as this data is generated worldwide by millions of users. Businesses must ensure that they have the best data management and training algorithms available for AI apps.

4. Infrastructure

Artificial intelligence-based solutions improve our lives and provide daily utility via high internet speeds. This speed can only be achieved by a company with the right infrastructure and high performance processing capabilities. Many companies still rely on outdated infrastructures, apps and devices to support their IT operations. The cost of updating systems is often feared by management. They may refuse to implement AI. Companies that embrace artificial intelligence need to be ready to expand their IT services. Many IT companies face significant challenges when it comes to replacing legacy infrastructure.

5. AI integration

Integrating AI into existing systems is the first hurdle in implementing AI in business. This is where the expertise and experience of AI solution providers are essential. To make the transition to AI, it is more than adding plugins to an existing website. It is crucial to assess the potential impact on infrastructure, data storage, and data input and make sure they are protected against any negative consequences. It is crucial to ensure that all AI requirements are met by the new systems. Once the transition is completed, employees must be trained in the use of the new system.

6. Computation

Information technology is facing many challenges. Information technology is always evolving. It is unlike any other industry to have made such rapid progress. The industry faces the greatest challenge in obtaining the computing power required to process large amounts of data for the development of AI systems. This level of computation can be costly, especially for startups and small businesses.

7. Privacy concerns

It is important for companies to be familiar with the legal aspects of artificial intelligence app development and implementation. It is extremely sensitive to the information that algorithms collect about users. Incorrect algorithms or data governance systems can result in incorrect predictions, which could lead to losses for the company. It could also be in violation of laws and regulations which could cause legal problems for the company.

8. Explainability

It is human nature not to trust things that are easily understood. One of the most difficult AI implementation challenges is uncertainty about deep learning models and a variety of inputs that can be used to predict an output and solve a problem. Explanations are necessary to ensure transparency regarding AI decisions and the algorithms used to make them. Companies must establish policies to examine the effects of artificial intelligence on decision making, audit their systems regularly and offer training.

Relationship between Intellectual property (IP) and Innovation

Many people are aware of the economic benefits that patent systems bring to society. They encourage innovation, investment, economic growth, and knowledge sharing. Patents promote innovation by granting monopolies to inventors who have made significant investments in creating new products and processes. In addition to traditional rewards such as academic recognition and promotion at research institutes, financial returns can also be offered for scientific innovation. If there is no incentive through patents, private investors may be reluctant to invest. These are only a few benefits of the patent system.

A study by the European Patent Office and the European Union Intellectual Property Office, Intellectual Property rights and firm performance (EU ) shows that companies with at least one registered design, patent or trade mark have on average 20% more revenues per employee than those who don’t own these intellectual property rights (IPRs). These IPR-owners also paid 19% more in average wages than other companies. The strongest link between a company’s performance and patent ownership is found to be 36% higher revenue per worker and 53% higher salaries than businesses without IPs. Next, the ownership of registered design (32% higher revenue, 30% higher wages), and trade marks (21% more revenue, 17% higher wages). Antonio Campinos, President of the EPO said:

  • Your business will perform better if you have a strong IPR portfolio. IPR-owning companies not only generate more revenue but also earn more for their employees. These are important indicators for society and the economy. This study also shows that there is significant untapped potential in Europe for SMEs, and that intellectual property ownership can be a huge benefit to them. IPRs are also used extensively by businesses to help us get through the 2008 financial crisis. Innovation will be a key driver of Europe’s recovery from COVID-19.

EPO-EUIPO prior studies of IP-intensive industries also provide compelling evidence of the positive correlation between IP rights, economic performance, and both the macroeconomic and individual company levels.

Healthcare organizations have widely acknowledged the role of patents as an incentive for innovation and investment in research. Children’s Cancer Institute Australia for Medical Research stated that patents are a key component in driving medical research innovation. They allow researchers to protect intellectual property and allow them the opportunity to capitalize on their inventions.

Patents can help companies grow by allowing them to take advantage of the potential market for their inventions. Global companies also benefit from patents, which provide an international platform for knowledge transfer through license agreements. Licenses can be obtained by international companies to exploit local inventions. This allows inventors to receive both financial returns and access to foreign markets.

Patents can be used to encourage knowledge sharing and decrease duplication of research. Patents can be used by researchers to improve upon a patent-pending product. Patent inventions can make it possible to do research that is impossible with other methods. For example, access to a patent-protected tool for research may be able to allow crucial research into learning machines’ decision making and help to develop training data that is objective. This research may not have been possible if the tool had been kept secret.

Patents are also a way to help society grant intellectual property rights and get disclosure in return. Patented inventions could be used for purposes the inventor did not know or could not realize. Patents encourage disclosure, and the invention can be used in more settings, based upon assumptions about the transaction cost associated with licensing.

But patents can be useful for such purposes as:

  • We are motivated to invent useful items because we know that patents will soon be granted.
  • Patents allow for widespread knowledge and widespread usage of inventions by inducing inventors to disclose their inventions rather than relying on secrecy.
  • Patents encourage investment in invention development and commercialization.
  • Patents enable systematic exploration of large potential for derivative inventions.

Research has shown that patents make up a substantial part of R&D incentives for only a few industries, such as pharmaceuticals. This is because medium-sized and large companies with R&D labs can use their inventions in their own ways and do not rely on patent licensing or sales to make returns. For inventors who are looking to sell or license their inventions to reap the benefits, or start-up investors who need protection for their portfolio companies, patents might be more crucial.

Patents granted at an early stage can be seen as a guarantee that the technology will succeed and the economic benefits will be realized. This leads to a decision by the developer.

The original inventor can hold a patent which allows easier transfer to other organizations that are better suited for development or commercialization. Many of DuPont’s product innovations were built upon inventions that were purchased from smaller companies. General Electric also acquired many inventions developed by private inventors or small businesses in the 1920s. Apple and Google are now net acquirers and folders of startup technology.

A patent is a tool that gives research institutions such as universities or government labs an incentive to market their inventions to companies who are able develop them and commercialize them. The Bayh-Dole Act granted universities patent rights for inventions that were the result of government-funded research projects. This was extensively discussed. Although the inventions were funded by public money, they will not have any economic value until they become commercially viable. This type of development could only be undertaken by companies. This created a distinction between the place of invention, and the site for the development. 

The Bayh Dole Act discussion states that companies wouldn’t be allowed to create a university invention if the company didn’t have exclusive rights. If universities had strong patent rights, they would be able sell exclusive licenses. Companies wouldn’t invest in development work necessary if they didn’t have patents, or if the government held them with nonexclusive licensing obligations.

Patents don’t always reward innovation or research investment equally. Economists are often negative about the perception of the patent system by society. It was because they were concerned that patents would create monopolies, and that patents might not be necessary to promote innovation. Patents may have adverse economic consequences. 

Patent fees can raise the price of products and services that use the patent invention. Transaction costs can be associated with the obtaining of a patent and enforcing rights. Transaction costs are associated with obtaining and enforcing rights. In countries with net importers, patents can have a negative impact on the balance of payments. For the exclusive use of patent inventions, local manufacturers must pay licensing fees for foreign IP companies. Strong, broad patents that were not selectively enforced have impeded technology development in America’s history. The situation was resolved only after a relatively inexpensive licensing system was established in relevant technology or industry, such as radio, aircraft, and semiconductors.

AI is challenging the core principles of the patent system

The rapid adoption of AI has made the question of AI patentability a hot topic. Patents encourage knowledge sharing, requiring that details of patent inventions be made public in return for exclusive rights to the invention. In the absence of this exchange, inventors might keep secret the details about their inventions. In the absence or exchange of ideas, inventors may keep details about new inventions secret.

AI is challenging core patent system tenets, especially in areas like inventorship and novelty. These tenets not only form the basis of patent law but also play a critical role in fostering commercial processes.

AI’s ability to create and iterate autonomously is one of its main challenges. This theoretically means that AI could create inventions that no human inventor could have made. This raises questions as to whether AI could be rewarded with compensatory damages or held responsible for wrongdoing.

Another problem is determining and assessing patentability. Many of the issues associated with AI inventions mirror those faced in traditional patent law. These include whether the invention is abstract or not, whether it has utility or no utility, and whether the invention was “functionally invented.” AI also raises questions about inventorship and ownership. While inventorship refers only to the creator of an invention, ownership is the person with legal rights to enforce the invention. This distinction is crucial in a patent system that is less centered on the European droit d’auteur doctrines. These doctrines are based on a person’s creativity and intellectual property.

Can AI be listed as an inventor?

It is still a matter of debate whether AI should be listed in a patent application as an inventor. This question has many different answers in different parts of the world. Virtually every jurisdiction has rejected patent applications that name AI as the inventor. This is evident by the DABUS patent application. It was granted only in South Africa, Australia. DABUS claims a trillion-neuron artificial neural architecture that can learn and create at unprecedented scales. DABUS claims it is capable of conceptualizing complex ideas after having absorbed general knowledge about the world. The EPO rejected the DABUS patent application in Europe. The USPTO rejected DABUS’s attempt to identify an AI system as the inventor.

Contrary to the Dabus approach to AI as a sentient AI general intelligence (AGI), with an intelligent agent to learn and understand any intellectual task that a human can, more

This is not a problem for now as innovative AI is only a part of today’s innovation. AI will become a more important component of research and development as AI becomes more capable. There are many issues to be aware of in the future, such as unclear rules regarding AI-generated inventions, their protection, who should be listed among the inventors, and who has the patents or inventions.

It is likely that the disruptive phenomenon of AI taking over human roles in IP and other legal areas will have major consequences. The principle of AI legal neutrality, which is a law that doesn’t discriminate between AI and humans in the same activity, is known as an AI law. This will improve human well-being.

It is important to consider whether AI should be listed in an application as inventor. Other laws will be required to address this issue. This could even lead to governments changing the patent statute to allow AI inventorship.

Can AI be used for rejecting patent applications?

IP law faces new challenges with the concept of inventive AI. Patentability is determined by the “person of ordinary skill and knowledge” standard (POSITA), which is used to evaluate inventive steps.

This test asks if an average person with relevant knowledge would consider a patent application obvious based on existing information. If the answer is yes, then the application will be denied. AI will enhance the capabilities of ordinary workers and they will become more skilled as well as knowledgeable. This evolution of the skilled person should raise the bar for patentability in the same way as Europe, where the concept was expanded to include skilled people, and team-based research is the norm.

AI will soon be able transform from automating human researchers to automating creative activity at large scale. Innovative AI may even be able to replace the skilled worker. AI that is able to automate research routinely will likely find more than the skilled individual. It may be difficult for AI to see the obvious. The test for inventive steps may need to be modified to put more emphasis on cognitive factors than economic ones, such as concurrent inventions and long-term unsolved issues, professional skepticism or long-held, but unmet needs. AI’s ability reproduce patent applications may be a key focus. There is no limit to the future intelligence of machines. It is possible that super-intelligent AI will eventually be able to see all things.

This could require a redesign of the traditional exam process. An AI system, for example, could be used to help the examiner identify patentable applications. It could also offer an equitable perspective on how obvious the criterion is.

AI as “copilot” with humans in solving problems

In the Utopia world, machines do all the work and humans live an enlightened role free of work pressure.  However, at present, AI can’t do all the work to support humans yet.  AI can take over specific process steps and perform more complicated tasks. AI is a versatile tool that can be used in specific situations. It is increasingly being used in conjunction by both humans and smart machines. Understanding the potential benefits of AI is key to understanding how it can be used to add value. While AI systems cannot yet replace whole job profiles, they are able to take over repetitive tasks. Thus, we see that AI in the future will co-exist with people.

Using AI as a “copilot” with human in solving problems can bring several advantages:

  1. Increased efficiency: AI can process large amounts of data quickly and accurately, which can help to speed up problem-solving processes and make them more efficient.
  2. Improved accuracy: AI can identify patterns and insights that humans may not be able to detect, which can lead to more accurate solutions to problems.
  3. Handling of routine tasks: AI can handle routine and repetitive tasks, allowing humans to focus on higher-level problem-solving and decision-making tasks.
  4. Handling of complex tasks: AI can assist humans in solving complex problems by providing recommendations, performing simulations, and analyzing data.
  5. Reducing human error: AI can help to reduce human error by automating routine tasks and providing decision support.
  6. Learning from human experts: AI can learn from human experts by observing their problem-solving processes and incorporating that knowledge into its own algorithms.
  7. Scaling up: AI can scale up the problem-solving capabilities of a single human, making it possible to solve problems that would otherwise be impossible for one person to handle.

However, it’s important to note that using AI as a copilot with human in solving problems also has its own set of challenges such as:

  1. Ensuring that AI is working in the best interest of human
  2. Ensuring that AI is not replacing human jobs
  3. Ensuring that AI is not introducing bias
  4. Ensuring that AI is not causing harm
  5. Ensuring human-AI collaboration is effective and efficient.

To overcome these challenges, it’s important to have clear policies and guidelines in place for using AI, as well as to involve human experts in the development and deployment of AI systems.

Conclusion

IP rights are a fundamental part of human rights. The advent of AI has had a dramatic impact on the IP system. The advent of artificial intelligence and digital technologies offer new opportunities to enjoy human rights but also present new threats to their protection. The relationship between technology and citizens has led to new concepts and obligations. As policymakers struggle for consensus on transparent and effective governance structures to protect fundamental rights related to technology, the advancement of AI technologies presents new challenges in public governance.