Iterating towards disaster

Yes, that happens, we all consider it, but did anyone thought it through? You see, innovation is essential in staying ahead, iteration tends to give you a 2 year advantage, innovation gives you a 5-7 years leap. That is not new, it has been a ‘fact’ of life for 3-4 decades. Yet that premise is about to change, it will change a lot and it will change towards the bad side of the pool. To see this we need a few items, the first is an article, an article that the Guardian gave us with ‘I’m sorry Dave I’m afraid I invented that: Australian court finds AI systems can be recognised under patent law’ (at https://www.theguardian.com/technology/2021/jul/30/im-sorry-dave-im-afraid-i-invented-that-australian-court-finds-ai-systems-can-be-recognised-under-patent-law), you see there is a danger here, even as the Guardian gives us “Allowing machine inventors could have numerous consequences, both foreseeable and unforeseeable. Allowing patents for inventions churned out by tireless machines with virtually unlimited capacity, without the further exercise of any human ingenuity, judgment, or intellectual effort, may simply incentivise large corporations to build ‘patent thicket generators’ that could only serve to stifle, rather than encourage, innovation overall.” This we get in the article from Australian patent attorney Dr Mark Summerfield, and he is right, you see, there is a larger danger here. It is not merely that only a few companies can AFFORD such an AI, the larger stage is that if we combine this and we add a little statistics to the pile, we get a new setting. 

SPSS (now IBM Statistics) has something called the conjoint analyses. To understand this, we need to take a look at the manual. There we see:

Conjoint analysis presents choice alternatives between products defined by sets of attributes. This is illustrated by the following choice: would you prefer a flight that is cramped, costs $225, and has one layover, or a flight that is spacious, costs $800, and is direct? If comfort, price, and duration are the relevant attributes, there are potentially eight products:

Product Comfort Price Duration
1 cramped $225 2 hours
2 cramped $225 5 hours
3 cramped $800 2 hours
4 cramped $800 5 hours
5 spacious $225 2 hours
6 spacious $225 5 hours
7 spacious $800 2 hours
8 spacious $800 5 hours

Given the above alternatives, product 4 is probably the least preferred, while product 5 is probably the most preferred. The preferences of respondents for the other product offerings are implicitly determined by what is important to the respondent. Using conjoint analysis, you can determine both the relative importance of each attribute as well as which levels of each attribute are most preferred.

This is all statistical science and it works, but the application can be changed. If data is the only premise here, we see the application in another way. What if the AI is taught the categories that enable a unique stage to own ANY patent field. Consider that this is not about a flight, what if this is about a processor.

Product Speed Processor Sampling
1 X Sycamore Bozon
2 X Sycamore Instantaneous Quantum Polynomial
3 X Tangle Bozon
4 X Tangle Instantaneous Quantum Polynomial
5 Y Sycamore Bozon
6 Y Sycamore Instantaneous Quantum Polynomial
7 Y Tangle Bozon
8 Y Tangle Instantaneous Quantum Polynomial

I am merely making a fictive sample with existing names, but what if the math of conjoint is tweaked to cover the quantum field to a larger degree, a computer can do this faster than any person and it can even start making the documents, so the AI can create a set of patents that cover the entire field, with a setting where less than 20 patents will stop commercial competitors to get traction in this field and this is not merely speculation, I feel that this is where we go to and now the big tech companies will own it all and the AI’s will have the entire patent field. Yes, there will be holes in the beginning, but as patent filing will overturn normal filings, the patent field will end up being owned by Google, IBM and Amazon. I have nothing against any of these three, but this is not what I (or anyone else) signed up for. I might just put all my 5G IP online making it all public domain, just to temporarily deflate the AI premise.

And personally, there is no way that either of the three had not considered this application, making the AI patent field a lot more debatable and I reckon that the larger law field is looking into that. In 2012 a total of 1,892 filings were made, now consider that an AI could cover a larger field with a mere 300 filings. That is not out of the realm of considerations, as such the Australian case we see in the Guardian could well end up with all kinds of nasty surprises if the stage of “The decision by the Australian deputy commissioner of patents in February this year found that although “inventor” was not defined in the Patents Act when it was written in 1991 it would have been understood to mean natural persons – with machines being tools that could be used by inventors” is not overturned. Will it? I cannot tell, but it opens a whole range of doors and some of them will end up being rather nasty.