Do we address text that could lead to hate or violence? Yes.
Claim 1 of our two General Risk Patents No. 9,552,548 and No. 9,760,850, pertains to “threats or risks of interest.” While the Specification presents a Proof of Concept where the models score yesterday’s emails that could put an enterprise at risk for specific types of litigation, the claims in these two patents are broader. They cover the identification of text that inspires hate or which may incite violence, so that any enterprise with the right to use these patents may receive an early warning of text which human reviewers assess to be in violation of its policies.
Our 8th deep learning patent describes a breakthrough.
On October 9, 2018, Intraspexion achieved a fundamental breakthrough: Our 8th patent is the first U.S. patent to explain how to use blockchain to enable the creation of deep learning models when the amount of training data is small. The inspiration for this invention stems from a problem that many U.S. companies face: a potential violation of the Foreign Corrupt Practices Act.
Our Priority Date is July 1, 2016.
On July 1, 2016, with assistance from co-founders Dan Cotman and Obi Iloputaife of CotmanIP Law Group PLC (Pasadena, CA), Intraspexion’s founder Nelson E. (Nick) Brestoff filed the provisional application for our first patent.
On September 27, 2016, the formal application was filed for “Using Classified Text and Deep Learning Algorithms to Identify Risk and Provide Early Warning.”
On December 12, 2016, as a first Office Action, we received a Notice of Allowance.
On January 24, 2017, our first patent issued as U.S. Pat. No. 9,552,548.
Intraspexion’s other patents are continuations of or relate back to our first patent. They all have the following characteristics:
Priority Date: July 1, 2016.
Expiration Date: September 27, 2036.
Our patents-first strategy.
Our first instinct was to build the software system to implement our first patent. We did.
However, because Intraspexion’s leadership believed that “deep learning” — a form of Artificial Intelligence (AI) — was hot and getting hotter, we also adopted a unique AI “patents-first” strategy.
Who’s Got Deep Learning Patents?
With a request via the Contact Us page, will provide an Excel spreadsheet to you free of charge. The spreadsheet lists all of the “deep learning” patents issued by the USPTO from January 1, 2013 to the current date of August 13, 2019.
Here are the Year over Year results (boldface indicates the # of patents): 2013 (3), 2014 (4), 2015 (4), 2016 (36), 2017 (81), 2018 (162), and 2019 (through the current date) (211).
The deep learning patent “land rush” began in 2016, when the level jumped from 4 to 36. You can see a doubling from 2016 to 2017 and from 2017 to 2018. With 211 new patents in 30 weeks, the total for 2019 is on track to reach approximately 365 new patents. If the current pace holds, we’ll see yet another doubling in 2019. As of the current date, INTRASPEXION’s patent portfolio is still the only startup in the Top 10. The Leaderboard Top 10 patent assignees are:
1. IBM (45)
2. Google (35)
3. Siemens Healthcare (24)
4. Microsoft (23)
5. NEC (12)
5. Samsung (12)
7. Amazon.com, Amazon Technologies and A9.com (subsidiary) (11)
8. Adobe Systems (9)
8. Baidu (9)
10. INTRASPEXION (8)
The Intraspexion patent family
On January 24, 2017, Patent No. 9,552,548 was issued and assigned to Intraspexion. The title is “Using Classified Text and Deep Learning Algorithms to Identify Risk and Provide Early Warning.”
On September 12, 2017, No. 9,760,850 was issued. This patent goes by the same title but has different claims. Both patents pertain to “threats or risks of interest.”
On September 5, 2017, the USPTO issued patents related to identifying risks in connection with a Product Defect (No. 9,754,205), Contract Invalidity (No. 9,754,206), Entertainment and Publishing projects (No. 9,754,219), and Medical Diagnoses (No. 9,754,220).
On September 5, 2017, the USPTO also issued a patent for “Using Classified Text and Deep Learning Algorithms to Identify Support for Financial Advantage and Provide Notice” (No. 9,754,218).
Note: This patent is for identifying “financial advantages of interest” to obtain rather than a risk to avoid.
On October 9, 2018, the USPTO issued Patent No. 10,095,992 for "Using Classified Text, Blockchain and Deep Learning to Identify Low-Frequency Adverse Situations, and Provide Early Warning."
Examples of low-frequency, adverse situations include the following: potential violations of the Foreign Corrupt Practices Act, Product Liability litigation, and losses due to the theft of trade secrets.