3D rendering of an AI robot, a law scale, and a judge’s gavel is the representation of cyber law or internet law.
Obtaining a photograph from Getty is a possibility. This could be done by obtaining the rights to use one of their images.
A Frontier in the Arena of Artificial Intelligence and Intellectual Property Rights
The Wild West analogy can be used to describe the present-day state of intellectual property in relation to artificial intelligence (AI). There is a great lack of clarity in regards to the regulations and laws that protect these rights. Consequently, it has created an atmosphere of uncertainty and confusion.
In the past, the subject of intellectual property in the realm of AI was not given much attention. The technology advanced rapidly and the majority of systems that were published in academic literature did not go beyond the stage of proving feasibility. Even the platforms developed for industrial use only managed to cause shifts in niche markets and, as a result, patent enforcement was rarely necessary.
When I established my AI business back in 2014 at the start of the deep learning revolution, I still recall the beginning. In 2014 and 2015, neural networks started to outshine humans in various activities like image recognition and basic gaming. Even though much of this research was open to the public, some organizations built substantial intellectual property portfolios. After Google’s purchase of DeepMind, we looked into the IP landscape and found that a number of the published concepts had pending patents. Therefore, we decided to pursue the same route. At that time, it was suggested by venture capitalists that AI companies should patent their core technologies in order to validate their worth, so we obtained multiple patents for transcriptomic, proteomic, and microbiomic predictors of biological age, as well as many generative concepts in chemistry, biology, and methods linking the two fields. Even though a high school student could do this work today, back then deep learning researchers were limited and expensive, so it was a risk to engage in this kind of research. However, when it came to enforcing these patents, it became clear that this was not a common practice, and demonstrating infringements in server-side industrial systems is difficult. Nonetheless, we kept patenting to make clear to our partners that the software they were licensing was based on the original work. As of yet, we have not prosecuted any possibly infringing entities. The market for AI in drug discovery is relatively small, and most of the value is in the patents associated with specific therapeutic programs rather than the IP around AI.
When it came to the many AI startups, Google possibly had a similar idea in mind: patent prosecution didn’t make much sense, and building the AI environment was a collective effort. Now, however, tools like ChatGPT, DALL-E and others are making a big impact on the market, potentially putting Google’s position as the leading AI player at risk and posing a threat to its core business – search. In light of this, it’s possible that Google’s views on IP protection will change.
Google possesses essential intellectual property in systems that employ self-attention.
When it comes to IP protection in AI, Google is one of the most secure companies out there. I recall my team members devoting a large amount of time to making sure their designs were distinct enough that any infringement of Google’s intellectual property could be avoided. Even DeepMind, a subsidiary of Alphabet, has a firm grasp on IP in this area. A quick WIPO search returns more than 800 results.
At the 2017 Conference on Neural Information Processing System (formerly NeurIPS), Google scientists introduced a paper titled “Attention is all you need“, which was the originator of the core self-attention methods employed in transformer architectures. This study was a major breakthrough in the field of transformer neural networks and was instrumental in the development of GPT-3. By January 2023, this paper had been cited over 62,000 times, making it one of the most-referenced works in Artificial Intelligence.
A screenshot of a significant publication on transformer neural networks is presented.
Dr. Alex Zhavovonkov
A comprehensive and extensive patent covering the self-attention methodology can be found through a WIPO and Google patent search on this link
A picture of the Google Patents output for a patent related to self-attention can be seen below. [+]
Doctor Alex Zhavoronkov is the holder of a PhD degree.
The filing of the patent took place on June 28th, 2018, which was almost a year after the original paper was uploaded to the preprint server on June 12th, 2017. The patent has now been granted.
OpenAI Chose Not to Pursue Patents
From its very beginning, OpenAI, a well-known AI powerhouse, was formed as a non-profit. I attempted to locate the patents registered under OpenAI and found them to be a difficult assignment. The WIPO search gave only a few outcomes for OpenAI s.r.l., which appears to be an unrelated organization. An easy Google search did not provide any significant results, so I inquired about ChatGPT itself.
A capture of the output of ChatGPT concerning OpenAI and DeepMind’s patent standings can be seen.
Dr. Alex Zhavoronkov is the name that comes to mind.
A precise account of the patent scope for deep learning was given, highlighting that IBM, Google, Microsoft, Samsung, and Baidu all possess extensive intellectual property holdings.
What company has the most patents concerning deep learning that are held by them? This is a question that can be answered through ChatGPT.
Though self-attention is utilized in GPT, it is uncertain if Google’s patent would encompass the application of self-attention in GPT’s structure.
A picture of the output from ChatGPT when a query is asked if Google has patented GPT.
Dr Alex Zhavoronkov is a person of academic distinction.
Does Google Have Plans to Assess and Implement Their Patents?
Google has an extensive IP portfolio in the area of AI, and holds a patent on the application of attention-based sequence transduction in neural networks. Despite this, they have not been particularly aggressive with their AI intellectual property. However, with the rise of generative AI and Microsoft’s integration of OpenAI tools into their products, this stance may soon become more assertive. ChatGPT is well-equipped to face this challenge.
Considering the possibility of Google utilizing their patents to confront OpenAI, ChatGPT has provided its opinion.
Be Morally Upright
In 2014, Elon Musk, a co-founder of OpenAI, made a well-known decision to make all the patents of Tesla available to the public, as a way to stop unfair competitive practices in the electric vehicle sector. It is possible that OpenAI’s choice to forgo patenting the primary AI technologies is due to this same philosophy.
Google was the first to develop the key aspects of generative AI and its neural network technologies involving self-attention. Nonetheless, it is improbable that they will bring a lawsuit against OpenAI. Patents are an excellent way to demonstrate priority and fend off patent trolling, but suing in the AI domain is not a normal practice. This is especially true, considering OpenAI was launched as a non-profit by people committed to making the world a better place https://www.newyorker.com/magazine/2016/10/10/sam-altmans-manifest-destiny.
Using AI, a picture was created at the midway point of a patent war between Google and another entity, which was depicted by …[+]
Dr. Alex Zhavoronkov, and Midjourney are connected.
It is possible to avoid plagiarism by transforming the structure of the text without altering its context or semantic meaning. Ensuring the original markdown formatting is maintained is also vital.
