AI Progress is a Poor Excuse for Seizing the Intellectual Property of Others
It was interesting to read that the New York Times was suing Open AI and Microsoft for copyright infringement (NY Times sues OpenAI, Microsoft for infringing copyrighted works | Reuters).
While there have been other copyright-type challenges to generative AI, this is probably the most significant one to date; and perhaps most importantly, because it seems like an issue of fundamental fairness – on a base level, shouldn’t you have to compensate someone for utilizing their work, or at least provide adequate attribution?
We are in the early innings of the generative AI revolution. So, maybe it’s not a surprise that the rules of the road are not yet fully established and agreed to. We need to be patient and give transformative technology a chance to flourish, right? The NY Times counters that “…there is nothing ‘transformative’ about using [its] content without payment to create products that substitute for [it] and steal audiences away from it.” Well said.
Certainly, plagiarism has always been an issue, but there are well defined standards that have developed over time that most respect and follow. While not foolproof, they acknowledge the intellectual property of others and provide a framework for reuse. From a young age we were all taught to provide proper attribution for someone else’s ideas by referencing a source either in the body of a work or through a footnote – even the footnote itself has a particular order and structure! At a minimum, this made us sensitive to the issue of passing off the hard work of others as our own. Even in today’s world, we re-tweet, or re-post, attaching our comments to the original work.
Now, with generative AI, we are entitled to usurp anyone’s ideas and call them our own as long as the information is in the public domain. The first time I used ChatGPT, I was struck by the fact that the program returned results as if it had been directly responsible for the original insights. Where was the attribution? Where were the links? Most generative AI advocates have defended this behavior under the “fair use doctrine” – a legal theory that allows for the reuse of copyrighted material without permission under certain, specific circumstances such as for use in parody, criticism, or teaching. After all, it’s very hard and costly to provide proper attribution or compensation to everyone – enlightened progress demands that we ignore this unfortunate complication and ensure that generative AI wins.
Ironically, the calls for progress at any cost seem to contradict the recent momentum in the direction of enhanced data rights and the acknowledgement that personal data cannot be used by third parties without permission or compensation. This may at least partly explain why Apple has largely been absent in fielding its own generative AI solution. Remember, Apple has made defending the integrity and privacy of user data with respect to third parties their calling card, especially relative to their mega-cap technology peers. Now, can they really start scraping other’s data and still hold the moral “high ground”? Apple needs to move deliberately in carefully navigating this potential minefield, especially if public investors are willing to give the company the benefit of the doubt in the meantime. Further, Apple is one of the best examples of a “fast follower” – letting the first movers sort out the kinks, then gobbling up market share based upon its financial and structural advantages.
Unfortunately, like many things in our polarized world today, there is little opportunity for nuance or distinction. Increasingly, every choice must be binary – either wholly embrace the brave new world of AI without reservation or reject it completely and remain a dinosaur. We object to this notion. Adobe is a great example. While they have been an early leader in incorporating Firefly, their generative AI tool, into their products for generating images, they were also one of the first to introduce an Intellectual Property indemnification policy. Essentially, to protect users from potential claims of copyright infringement, Adobe has trained its Firefly models on licensed content like its own Adobe Stock, expired copyright public domain material, and other non-copyright or openly licensed material.
For our part, we have so far steered clear of making solely generative AI-based private investments, preferring companies focused on applying AI to solve functional problems using proprietary data sets – one of our co-investors has dubbed this “blue collar” AI. A better way to participate in the technological revolution without the “baggage” associated with “scraping” someone else’s data.
Up until now, content providers have largely been unwilling to challenge the generative AI players, preferring instead to cut licensing deals – in fact, some believe that the NY Times legal action is part of a strategy to negotiate better terms. Probably. In our estimation, a last-minute settlement that skirts the fundamental issues at hand would be detrimental to creators of content that don’t have the resources, clout or public presence to warrant a direct deal (i.e. most of the rest of us). While progress is important, we must strive to find ways to enable the adoption of transformative technology while respecting fundamental principles that provide the foundation for our collective economic success.
Progress is a poor excuse for seizing the intellectual property of others. It’s always better to address the challenges of generative AI early to provide a fair and responsible playing field so that all members of the ecosystem can benefit.
Securities highlighted or discussed in this blog have been selected to illustrate Validus’s investment approach and/or market outlook and are not intended to represent any strategy or portfolio performance or be an indicator for how strategy or portfolio have performed or may perform in the future. Each security discussed in this blog has been selected solely for this purpose and has not been selected on the basis of performance or any performance-related criteria. The securities discussed herein do not represent an entire portfolio and, in aggregate, may only represent a small percentage of a strategy or portfolio holdings. The strategies and portfolios are actively managed, and securities discussed in this blog may or may not be held in such strategies or portfolios at any given time. These individual securities do not represent all the securities purchased, sold, or recommended and the reader should not assume that investments in the securities identified and discussed were or will be profitable. Nothing in this blog shall constitute a recommendation or endorsement to buy or sell any security or other financial instrument referenced in this letter.
Validus Growth Investors, LLC seeks to invest in companies at every stage of their growth. From startups to publicly traded companies, our research identifies inflection points that have the potential to produce meaningful growth and income for the clients we serve.
Investment Advisory Services are offered through Validus Growth Investors, LLC (“Validus”), an SEC Registered Investment Adviser. No offer is made to buy or sell any security or investment product. This is not a solicitation to invest in any security or any investment product of Validus. Validus does not provide tax or legal advice. Consult with your tax advisor or attorney regarding specific situations. Intended for educational purposes only and not intended as individualized advice or a guarantee that you will achieve a desired result. Opinions expressed are subject to change without notice. Investing involves risk, including the potential loss of principal. No investment can guarantee a profit or protect against loss in periods of declining value. All information is believed to be from reliable sources; however, we make no representation as to its completeness or accuracy. Opinions and projections are as of the date of their first inclusion herein and are subject to change without notice to the reader. As with any analysis of economic and market data, it is important to remember that past performance is no guarantee of future results.