Ai & IP - Current State and Future...a system of cells. interlinked. within cells interlinked...

Generative AI models have been compared to the revolutionary impact of the internet. But, just like predicting the impact of the internet in its infancy was impossible, it's impossible to accurately predict how the use of these tools will impact intellectual property (IP) in the future. At the end of the day, they are tools, and their use will determine the implications and impact in the future.

Generative AI has been applied in various fields for several years. If your phone unlocks by recognizing your facial features, then you are using AI. However, the widespread adoption of AI generative language models, such as ChatGPT, has recently led to increased interest. ChatGPT is an AI language model, and there are several other AI models that generate different outputs.

If you look at its impact, for example, the education sector is scrambling to see how to move forward. They are currently divided between those who want to outright ban the use of this technology or adopt, adapt, and embrace this technology. Historically, banning disruptive technologies has never been effective in the past.

Similarly, two opposing camps argue for and against the importance of embracing generative AI technology for IP. Those in favor believe that it will drive innovation and improve the IP cycle for the common good. Meanwhile, opponents argue that the value of work will be diminished, and there will be no incentive for people to create works of intellectual property.

Since IP is a vast field, attempting to predict when and where issues will arise is a futile exercise. As I said, the way these tools are used will determine how problems arise and how we interpret existing laws and regulations to solve said problems.

Let's take a look at one aspect of IP law, for example, copyrights. In the US, the output of a generative AI cannot be granted a copyright owned by the user or the human element. We all know the case with the monkey who took a selfie that was taken before the courts. Only a person can have a copyright. While many current IP issues can be settled using existing laws and regulations, new questions will arise. Other jurisdictions have different interpretations, such as the UK?(patent law allows humans that use?AI?to devise inventions to be named as inventor in most cases).?

So, if the output cannot have a copyright, some might argue that the user's efforts to input a specific prompt might result in being open to obtaining a copyright on the output. Prompts are the way in which you ask the model questions, and arguably prompts can be considered copyrightable as small literary works. There are companies that monetize prompt engineering; the more specific and robust the prompt is, the more specific and accurate the results of the output you obtain.

This means that someone could claim that because there is a human element involved in the creation of the prompt, the user deserves a copyright on the output due to their creative contribution. This will probably not hold up because, first, the prompt is considered an idea rather than an expression. Ideas cannot be protected by copyright, only their expressions can. Second, it's important to note that the answer or output to an identical prompt will most likely change somewhat each time it's asked.

It's possible that there will be claims of work-for-hire in relation to AI-generated outputs. In the same lines as before, the argument may be that the owner and/or creator of the AI model had some involvement in the creative process and should be able to claim copyright protection for the output based on the work-for-hire doctrine, arguing that the output was ultimately the result of the instructions given to the AI model by the human element. However, the question of whether such claims would be successful is still up for debate.

Infringement is also a key issue to think about, particularly the potential for copyright infringement when training AI systems. One issue is the source of data used to train the AI. If data is obtained without proper I apologize for the abrupt ending. I will continue from where I left off.

If data is obtained without proper authorization or licenses, there could be legal consequences for copyright infringement. For example, using copyrighted images or other materials without permission could result in lawsuits. There are currently a couple of cases that have been presented in the US regarding the use of image databases used to train and create AI-generated works. Getty Images has a license to use their image database to train AI.

Another example would be Microsoft-funded ChatGPT, which is very good at writing code partly because code found on the platform GitHub, owned by Microsoft, was used to feed ChatGPT's database to train it. GitHub would need to ensure that the code used to feed the database was obtained legally and with the proper licenses. This would involve obtaining the proper authorization from the authors or owners of the code, or using open-source code that is licensed for such use.

Claims of fair use will probably arise, such as what constitutes fair use of information regarding AI and the information used to train these models. Other new issues will have to be addressed, such as whether particular styles can have copyright protection. Generative AI in its current state can create work in the style of a famous author, painter, or songwriter, and style is not protected under copyright law.

As you can see, the ramifications are endless, and the disruption of IP law will have to deal with the issues as they arise, and in some instances, fundamentally change. Many of the issues we could address are regulated by the current IP legal norms and regulations. Like predicting the issues for IP that arose from the internet in 1999, any attempts to try and predict how, when, or where the issues for IP will arise from AI will not age well.

So, what do we do with them now and in the future? How can we use these tools at our firms?

Right now, AI models are tools, and these tools are not ready to handle legal work. Do not use them for any legal work, do not use them for anything important, and do not input any confidential information, as you are feeding that information to the database.

What matters now are your skills related to this technology. The skills that matter now are knowing how to use the technology to help you find information you need and applying critical reasoning to get accurate and accountable results. Your team needs to understand how this technology works yesterday. Know the limitations of the technology, the possible perils of its use, and its potential.

If you have someone on your team that needs to improve their interpersonal skills, this can help. If you aren't a native-speaking IP office, this can help better communicate and improve tone and structure. It can help improve client service, and you can use it for brainstorming and quick, content-rich outlines. Use it to help you improve your services and communication. AI language models are particularly good at this, but don't use it for any real work.

In the past, using search engines was new and awkward; now, search bars are commonplace in almost all software user interfaces, and they are all very familiar to us now. In the same context, now is the time to learn how to use these tools, be an early adopter, and make sure you're familiar with how to use them, but don't use them for any real legal work.

In the future, we will have instant research. The advent of generative AI is only in the limelight for about six months, and already it is going through changes. The monetization of ChatGPT through recently introduced ChatGPT+. This week they introduced ChatGPT 4, which is said to be an improvement in every possible way.

ChatGPT's API kit, which will be released shortly, allows developers to easily integrate the power of GPT into their applications and create advanced language-based features, such as natural language understanding, text generation, and chatbot capabilities. This makes it ideal for use in chatbot applications within your organization, as it can understand and respond to user queries in a way that feels natural and human-like. The API also has the ability to generate text, which can be used to create automated responses, summaries, or even entire articles. This can be particularly useful in content creation and summarization applications applied to your particular business and dataset.

The fine-tuning of AI models is where you take all of the language-based AI training from one model, such as the API, and apply it to an existing database, for example, Lexis Nexis. All of a sudden, you have an AI model that will be able to answer legal inquiries in real-time with complete accuracy and context, for a modest price, of course.

We will have instant legal subject matter expertise; it will be like having an expert sitting at your desk with all of the global knowledge and can communicate it, explain it, and generate new works at a click's speed, but we are not there yet. This is around the corner; the faster you and your team know how to use this new technology, the better.

As AI continues to evolve and improve, it is essential to stay informed and adapt to the changes in the legal landscape. Firms should invest in training and education to ensure that their teams are well-equipped to leverage AI tools effectively and responsibly. By embracing the potential of AI while understanding its limitations, legal professionals can enhance their services, boost efficiency, and stay ahead in an ever-changing world.