Wild West of Intellectual Property in AI
In the past, intellectual property issues in AI were often overlooked. Technology is changing rapidly, most systems published in the study literature rarely evolve beyond proof of concept, and industrial force platforms have caused model changes only in niche markets and rarely patent applications.
As someone who founded AI in 2014 at the end of the Deep Learning Revolution, I still have some memories of how it started. In 2014 and 2015, deep neural networks began to process people in a variety of tasks, including recognizing basic images and computer games. And while most of this research has been published openly, some companies have begun to build huge portfolios of intellectual property. I remember that after Google’s acquisition of DeepMind we analyzed the IP aspect, many published ideas had pending patents. We decided to follow their lead. At that time, for AI to confirm the assessment, investors recommended patents for some core technologies, and we obtained several licensed patents on transcriptomic, proteomic, and microbiomic predictors of biological age. Chemistry, biology, and even methods connect the two fields. Today, even a high school student can do it again, but a decade ago, when in-depth scientists were scarce and expensive, participating in this type of research was very dangerous. However, when it comes to the application of these patents, it turns out that it is not a normal practice. In addition, violations are difficult to verify in server-side industrial systems. However, we have continued to patent even to show some partners that the software they are licensing is based on the original work. And so far we have not prosecuted the perpetrators. The market for AI in drug discovery is small, and most of the prices are not in IP around AI, but in patents around specific therapeutic applications.
It is likely that Google has followed a similar pattern when it comes to many AI startups. Their AI patent litigation just doesn’t have much commercial meaning, and building an AI ecosystem is a community effort. However, as ChatGPT, DALL-E and other prototyping tools are taking over the market, storming with Google as the dominant player in AI and undermining its core business – Google’s search for IP protection is subject to change.
Google keeps important IPs in self-contained system
And when it comes to IP protection in AI, I do not know any company stronger than Google. I remember our team spending weeks ensuring that the architecture was unique to avoid violating Google’s IP. Even DeepMind, a wholly owned subsidiary of Alphabet, occupies a strong IP position in space. A simple WIPO search returns over 800 results.
Google is also the originator of the core self-care approach used in transformational architecture. It was Google scientists who broke the lesson in the transformed neural network that paved the way for GPT-3. In 2017, at a conference on information processing systems (NIPS, later renamed NeurIPS), Google scientists presented a workshop document entitled “Attention is what you need”. As of January 2023, the paper has been cited more than 62,000 times, making it one of the most widely cited documents in AI.
And simple patent search WIPO and Google create comprehensive and comprehensive patents that cover this powerful approach.
It is clear that the patent application was filed on June 28, 2018, about a year after the seminal paper appeared on the previous print server on June 12, 2017, and the current situation confirms that the patent has been granted. Provide.
Famous OpenAI Patent Avoidance
OpenAI, the dominant AI power since its inception, was initially established as a non-profit organization. I tried to find a patent under OpenAI and found it a difficult task. The WIPO search returned only a few results for OpenAI srl, which seemed irrelevant to the company. A simple Google search did not yield any meaningful results, so I asked ChatGPT myself.
It also accurately describes the patent situation in in-depth learning. Of course, IBM, Google, Microsoft, Samsung and Baidu have huge IP portfolios.
And according to ChatGPT, while GPT uses self-care, it is not clear whether Google’s patents will cover the use of self-care in the GPT architecture.
Will Google try to test or strengthen its patents?
While Google has a strong IP portfolio that covers many areas of AI and it has patents on the use of attention-based sequencing in neural networks, it is not known for Flex its IP muscles in the artificial intelligence. However, as next-generation AI continues to dominate the market by storm, and Microsoft integrates OpenAI across its vast product ecosystem, this position could change rapidly. And ChatGPT has the perfect answer for it.
Do not be a bad person
In 2014, Elon Musk, one of the co-founders of the famous OpenAI, made all of Tesla’s patents public to protect the performance of unfair competition in the EV industry. And OpenAI’s decision not to waste time on patenting core AI technology is likely to come from this philosophy.
While Google has pioneered the key theoretical aspects of the next generation of AI, and is largely prioritized in core neural technologies using self-care, it is unlikely that legal action will be taken against OpenAI. . Patents are a great way to prioritize and protect against piracy, but suing other scientific groups in AI is not a common practice. In particular, consider the fact that OpenAI emerged from a nonprofit that was largely driven by people dedicated to making the world a better place.