Hacker News new | past | comments | ask | show | jobs | submit login
AI drives down cost and drudgery of routine patent work (ft.com)
36 points by hhs 5 days ago | hide | past | web | favorite | 20 comments





A fun sensationalized headline for this would be "AI now capable of doing the work of Albert Einstein."

Lol. They should have ran with that.

I mean we don't really understand creativity. Maybe that AI will evolve to invent new physics - after all it spends all its time analyzing inventions for new ideas. /s

Somehow I have the feeling that like in medical the new technology will not drive down the cost for the customers of the patent lawyers.

Any cost savings likely will not be very noticeable, in part because the technology doesn't do very much. It is directed at finding prior art, classifying inventions, etc. That work already is often outsourced overseas. The stuff that makes patent legal services expensive (e.g. having an engineer explain how the invention works so a patent can be drafted) remains quite outside the reach of existing computer technology.

There are some tools out there that can help with the drafting of patent applications too. I've seen tools that can predict the art unit the patent might end up, catch errors in patent applications being drafted, and even tools that can draft a simple preliminary written description for a set of patent claims.

At first, probably not. Until that one firm comes in using the technology to undercut everyone else's prices. In fact, I believe Atrium[0] is doing exactly that. I'm not sure if their prices are lower right now, but since they built up around technology, they certainly have that lever available to them.

[0] https://www.atrium.co


Actually fees have actually been fairly flat in patent prosecution over the recent years. Unlike other legal work, it's very common for the work in patent prosecution to be performed for fixed fees (e.g., $10k/patent application). Thus regardless of the number of hours billed by the attorney, they will not be paid for the hours worked that exceeds the fixed fee they're charging for the task. As such, patent attorneys do have extra incentives to use these emerging AI tools to help them become more efficient because it allows them to stay below the fixed fees they're charging for clients.

Terminology question... is patent prosecution same as a patent lawsuit?

Patent prosecution is the process of drafting the patent application and working with the patent office to get a patent issued. So it's the work done to secure a patent before the patent is asserted in a lawsuit.

Is that the same for criminal prosecution?

Well it will make patent attorneys more productive which may mean they can take on more customers at different prices or even allow smaller firms etc...

Patents get really interesting when litigated because 40% of patents are found to be invalid.


Not disputing your number, but wherever you got it, it’s likely “40% of challenged patents” that are invalid. Only a relatively small subset of patents are challenged.

As an aside; patent attorney productivity is an elusive concept. Or any attorney, at that. And that is because every single attorney only cares about one metric: billable hours. Not volume of work produced, not client wins, but billable hours. Efficiency to an attorney means lower effort to bill more hours. Quality is not a concern, only hours. I cannot stress enough how traditional concepts of efficiency or success or effectiveness don’t apply to lawyers. It’s a deeply flawed industry. And in patent work, it’s difficult for clients to determine if the lawyer is doing a good job or not. Concrete results are many years down the road, and often the in-house attorneys supervising aren’t qualified or given enough time to evaluate ongoing prosecution efforts across a large portfolio.


You could amend payment function to be sublinearly increasing to cope for hours inflation.

There’s a lot of good ideas to reduce costs for patentees, but this article doesn’t really touch on any of them

If you're interested in using patent data for AI, check out https://www.kaggle.com/bigquery/patents and https://www.kaggle.com/ostegm/plotting-similar-patents

Thx for posting this. Very interesting dataset: https://bigquery.cloud.google.com/table/patents-public-data:...

Do you know by any chance how the `embedding_v1` vectors were generated? The data field description says "Machine-learned vector embedding based on document contents and metadata, where two documents that have similar technical content have a high dot product score of their embedding vectors."

Could this be word2vec, GloVe, or something else like that? Maybe produced from the tf-idf-transformed sum of the word tokens in the title+abstract of each patent?


We (I run Google Patents), generated them using Wsabie (https://research.google.com/pubs/archive/37180.pdf) trained on the set of words of the full text -> Cooperative Patent Classification codes. So summed word embeddings trained for a classification task, which works well on similarity too.

What will happen to the patent system when people start claiming AI generated patents? Like those generated with GPT-2 on this site: https://aipatent.wordpress.com/




Registration is open for Startup School 2019. Classes start July 22nd.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: