Good afternoon, Mr. Chair, vice-chairs and distinguished members of the committee.
My name is Christian Troncoso and I'm the director of policy for BSA The Software Alliance.
BSA is the leading advocate for the global software industry before governments and policy-makers around the world. Our members are at the forefront of software-enabled innovation that powers the global economy and helps businesses in every industry compete more effectively.
Because copyright policy is a critical driver of software innovation, we're deeply appreciative for this opportunity to appear before the committee today.
This committee's review of the Copyright Act comes at a very timely moment.
With the recent announcement of the pan-Canadian artificial intelligence strategy, Canada has staked out an ambitious goal of becoming a global leader in the development of AI. This committee has a critical role to play in helping realize that vision.
To become a global leader in AI, Canada will need to set in place a policy environment that will enable its R and D investments to flourish. One critical competitive factor will be access to data. AI research often requires access to large volumes of data so that software can be trained to recognize objects, interpret texts, listen and respond to the spoken word, and make predictions. Ensuring that Canadian researchers can compete with their counterparts in other AI leading nations will therefore require careful examination of government policies that affect their ability to access data. Copyright is one such policy.
As currently enacted, the Copyright Act may place Canadian researchers at a disadvantage relative to their international competitors. For instance, unlike the U.S. and Japan, the Canadian copyright system creates uncertainty about the legal implications of key analytical techniques that are foundational to the development of AI. This committee can shore up the legal foundations for Canada's investments in AI by recommending the adoption of an express exception to copyright for information analysis.
Many of the most exciting developments in AI are attributable to a technique called machine learning. Machine learning is a form of information analysis that allows researchers to train AI systems by feeding them large quantities of data. This so-called training data is analyzed for the purposes of identifying underlying patterns, relationships and trends that can then be used to make predictions about future data inputs.
For instance, developers have created an app called Seeing AI that can help people who are blind or visually impaired navigate the world by providing audio descriptions of objects appearing in photographs. Users of the app can take pictures with their smart phones and the Seeing AI app is able to translate for them, through an audio description, what is occurring in front of them. To develop the computer vision model capable of identifying those objects, the system was trained using data from millions of photographs, depicting thousands of the most common objects we encounter on a daily basis, such as automobiles, street crossings, landscapes and animals.
It would be understandable if you were wondering right now what any of this has to do with copyright. The uncertainty arises because the machine learning process may involve the creation of machine readable reproductions of the training data. In some instances, the training data may include works that are protected by copyright. In the case of Seeing AI, that would be the millions of photographs that were used to train the computer vision model to identify common objects.
To be clear, the reproductions that are necessary for machine learning are used only for the purposes of identifying non-copyrightable information from lawfully accessed works. However, the Copyright Act currently lacks an express exception to enable that type of informational analysis. Therefore, there is considerable uncertainty about the scope of activity that is permitted under current law. This uncertainty poses a risk to Canada's AI investment, and provides a competitive advantage to those countries that provide firmer legal grounding for AI development.
Japan is one such example. In 2009, Japan passed a first of its kind exception for reproductions that are created as part of an “information analysis” process. Earlier this year, the Japanese diet amended the exception to make it more broadly applicable for AI research. Analysts now credit these legal reforms for transforming Japan into what they call a machine learning paradise.
In the U.S., courts have also confirmed that under the fair use doctrine, incidental copying to facilitate informational analysis is non-infringing. In September, the European Parliament voted in favour of a new copyright provision that would provide member states with the flexibility necessary to create broad exceptions for information analysis. Singapore and Australia are currently considering the adoption of similar exceptions.
These developments reflect an emerging consensus that the creation of machine readable copies for purposes of information analysis should not be considered copyright relevant acts. Copyright protection was never intended to prevent users from analyzing a work to derive factual, non-copyrightable information, so it makes little sense for copyright law to prevent such an analysis merely because it's being performed by a computer.
To ensure that Canada's significant investments in AI will pay dividends long into the future, the Copyright Act should be modernized to provide legal certainty for this common sense proposition.
The reproductions that are made in the course of training AI systems are unrelated to the creative expression that copyright is intended to protect, are not made visible to humans, and do not compete with or substitute for any of the underlying works. In other words, exception for information analysis poses no risks to the legitimate interests that copyright is intended to protect.
Copyright is ultimately intended to provide incentives for the creation of new works. An exception for information analysis advances this objective by stimulating the creation of new research and enabling the discovery of new forms of knowledge. By recommending the adoption of this exception, this committee will ensure that the Copyright Act remains fit for purpose in an age of digital intelligence.
Thanks, and I look forward to the questions.