The advent of a new generation of generative AI applications could represent a sea change in how we see creativity and ownership. Jude Umeh CITP FBCS reflects on what the new wave of technology could mean for human input and intellectual property.
A new breed of generative AI applications, including ChatGPT, has taken the world by storm with its ability to generate highly realistic content including text, art, code, music, poetry, videos and games. Although still in its early days, this new technology promises a double-edged sword of improved productivity and efficiency alongside the spectre of disintermediation and role rationalisation across the board.
Naturally, this raises a host of questions: what makes generative AI applications so compelling? Is it the seemingly miraculous leap in capability, application and user experience? Could it be something to do with the timing (read: at the tail-end of a great pandemic) and/or the tantalising promise of even greater things to come? And what has all this got to do with Intellectual Property (IP)? Is this finally the right time to update and upgrade that most artificial of property rights into something more fitting for the 21st century?
What is generative AI?
The current plethora of good-quality generative text and image applications signals a sea change in the scale, speed and production of creative content. The World Economic Forum considers generative AI to be “a game changer for which society needs to be prepared”.
A Generative Pre-Trained Transformer (GPT) is a type of neural network and deep learning algorithm or model trained on vast amounts of (usually internet) data. It is also known as a Large Language Model (LLM), which works by learning from large datasets and can recognise, summarise, predict, translate and generate content remarkably well.
OpenAI’s ChatGPT broke records by becoming the fastest-growing consumer application ever, with over 100 million users within 2 months of release. OpenAI also has other GPT-powered applications such as Codex (for code generation), Dall-E (text-to-image generation), and Print-e (a 3-D model generator), so there is something for everyone. Other popular text-to-image models such as Stable Diffusion and MidJourney have gained equal fame (or notoriety) due to their amazing outputs of new or enhanced images, art, animation, videos, games and architecture.
Though Microsoft and Google were quick to launch GPT-enhanced search applications in a battle for market share, we’re really only just starting to see what these things can be made to do. There are already directories available to find the right AI model for your projects or interests, as well as for adult-oriented and NSFW applications; for better or worse, the image-laden porn industry is right at the forefront of the charge to adopt AI-generated imagery.
Furthermore, and in a triumph of user experience, GPT-based applications are starting to bridge the chasm of usability which historically made other AI applications inaccessible to most users. When an easy-to-use technology captures people’s imagination it can gain unstoppable momentum towards mass adoption; IBM’s Wolfram Alpha and other equally sophisticated AI systems may provide more verifiably accurate outputs, but many lack that hook of simplified accessibility that ChatGPT brings to the table.
What does generative AI mean for the future of human input and creativity?
There’s a great fire of expectation washing over the AI landscape, with frenzied experimentation on generative models approaching fever-pitch. Could this be a much-anticipated AI Summer, as opposed to the notorious AI Winters of the past?
Though there is the fear that existing roles and jobs may fall victim to the quality, speed and scale of generative AI outputs, it is also becoming clear that AI-generated content may not yet be consistently reliable or accurate enough to take over the reins. There is still need for humans in the loop to ensure the best possible outcomes.
As generative AI systems mature, the nature of human input will likely shift towards more human-optimal activities within emergent systems – e.g.: plain common sense, fact-checking and overall governance based on human values and ethics. For example, new disciplines such as prompt engineering are emerging to address certain gaps and limitations of generative AI in order to create optimal output that can incorporate fact-checking, auto-correction, divergent opinions and stateful, complex conversations.
Finally, it is worth considering the potentially huge implications of generative AI on ethics and IP, including:
- Intentional or inadvertent generation of abusive, offensive and discriminatory content
- Infringement of the moral rights of creators whose works are used for training generative AI models that can and do generate or mimic their works or style
- The ability to generate reputation-damaging “deepfakes” of celebrities or private individuals, brings its own set of ethical challenges around privacy, defamation and libel
- The question of whether AI itself should be granted IP rights was a hotly debated topic at last year’s World IP Forum event
- It is possible that owners or users of content-generating AI may themselves run the risk of having their IP applications refused or revoked (if cases like the AI-generated comic book are anything to go by)
- Claims of software piracy, such as in the class action lawsuit filed against the creators of GitHub’s AI powered Copilot
- Issues of plagiarism through training data: for example, Google’s MusicLM, a text-to-music generative AI application, remained unreleased apparently because some outputs included minuscule direct copies of IP-protected music from its training dataset.
The latter points are consistent with training data sourced from the internet without the knowledge or consent of the relevant rights owner. Perhaps there ought to be a new class of IP rights to grant compensation for rights-holders whose contents are included in AI training datasets!
How can we respond to the inevitable transformations brought about by AI?
How should the average person respond to what could amount to a critical, transformative step in collective human experience and cultural evolution? Here are some suggestions:
Firstly, it’s important to keep some perspective. Emerging technologies are, by their nature, emerging, so their cumulative impact on human society and culture is unclear. They simply can’t do it all yet, and this can make the landscape in which they work appear to be a puzzle within a puzzle. For example, statistical AI-based models (e.g. LLMs such as ChatGPT) are not optimised for explainability, unlike their logic-based symbolic-AI counterparts which are focused on high-level human reasoning and classical logic. Reaching the holy grail of Artificial General Intelligence will likely require a joint statistical and symbolic AI system.
Recycling internet data for AI training can lead to inaccurate AI-generated content and this, according to an MIT Technology Review article, can lead to the “slow poisoning of the internet”. Additionally, the history of AI is replete with the painful repercussions of over-inflated, unmet expectations of exciting developments, leading to so-called AI Winters.
However, none of this means we shouldn’t get excited at all - it is undeniable that something special is happening, and at an extraordinary point in time. For example, the simplified AI user experience is here to stay – ChatGPT’s accessibility and ease of use is an experiential gulf which, once crossed, cannot be uncrossed. This is significant for mass adoption and bodes very well for generative AI and other models.
Generative AI also creates opportunities to explore ideas and topics in ways not previously imagined or viable: for example, visual artist Malik Afegbua’s ‘The Elders Series’ uses AI generated images to bring to life an alternative view of ageing in a way that is important and uniquely engaging – but impracticable in reality. Businesses are not immune to the excitement either, and many organisations are now asking how to use GPTs to transform their business.
Be part of something bigger, join BCS, The Chartered Institute for IT.
Finally, it’s important to be prepared for whatever happens. Change is happening at an unprecedented scale and speed, not only with AI applications but with other game-changing technologies that continue to supercharge innovation and drive emergent outcomes. The keys to riding this wave are educating ourselves, keeping pace with developments, and getting involved with the technology.
Humanity’s superpowers are adaptability and resilience, usually underpinned by irresistible curiosity and thirst for knowledge; keeping pace with current innovations and separating the noise from the substance can be daunting, but businesses and their leadership must also lean into these human characteristics if they are to survive. Many game-changing applications did not – possibly could not – exist even 15 years ago, and such rapid developments will require the right balance of diverse input to deliver value to the many. The field is open to anyone with an appetite for experimentation and discovery.
In the true spirit of enquiry, ChatGPT’s views on the future of Generative AI and IP are captured in the conversation below, with sub-minute responses to two questions by this author:
What are the key implications of generative AI for intellectual property such as copyright?
- Difficulty in determining original authorship: AI models can generate content that is similar to existing works, making it difficult to determine who hols the rights.
- Lack of legal protection: Currently, most copyright laws do not extend to AI generated content.
- Potential for infringement: AI generated content could potentially infringe on existing copyrights, which could lead to legal disputes.
- New legislation: The increasing use of AI in content creation may prompt the need for new legislation to address these issues.
What type of new legislation might be useful to address those issues?
- Extension of copyright laws: To cover AI generated content and determine who holds the rights.
- Moral rights protection: To ensure that authors have control over the modification or destruction of their works.
- AI creators liability: To hold AI creators responsible for any infringing content generated by their models.
- Fair use clarification: To clearly define the limits of fair use in the context of AI generated content.
- Registration and licensing: To provide a clear and transparent process for registering and licensing AI generated works.
From this response, it is obvious that generative AI presents some key challenges and opportunities for IP. It’ll take a combination of foresight and the ability to harness AI capabilities - along with a good dollop of common sense and compassion - to evolve the right system for the new world we’re entering.
- Generative AI – a game-changer society needs to be ready for (World Economic Forum)
- Futurepedia - The Largest AI Tools Directory
- Meet Unstable Diffusion, the group trying to monetize AI porn generators
- Will AI replace software engineers?
- Prompt Engineering 101: Introduction and resources
- AI image generation tech can now create life-wrecking deepfakes with ease
- WIPF2022: A different mindset for emerging technologies and intellectual property
- AI-Generated Comic Book Could Lose Copyright Protection
- The lawsuit that could rewrite the rules of AI copyright
- Google created an AI that can generate music from text descriptions, but won’t release it
- Can you steal the ‘feel’ of a song?
- AI can't do it all, yet
- Symbolic vs. Subsymbolic AI Paradigms for AI Explainability
- MIT Technology Review - How AI-generated text is poisoning the internet
- History of the first AI Winter
- Are NFTs the future of digital IP and the creative world, or just a remix of DRM and all its woes? (Part 5)
- The Nigerian AI artist reimagining a stylish old age
- Generative AI is here: How tools like ChatGPT could change your business
- “Please slow down”—The 7 biggest AI stories of 2022 – Ars Technica
- ChatGPT and Intellectual Property