Artificial intelligence (AI) has become an increasingly prevalent and transformative technology in today’s world. The extraordinary pace of progress is best seen in generative models, such as DALL-E-II, Stable Diffusion, MidJourney, and ChatGPT, producing images, text and speech that are near-indistinguishable from the work of humans. For what can only be described as a monumental leap forward in AI research, awe sits equally beside trepidation, as the prospect of more advanced and widespread AI systems raises sobering concerns. This has prompted an important shift in the public discourse around the consequences of ubiquitous AI in its current form, as well as the interplay between ideas around AI risk and value alignment. In this article, we’ll explore the history of these ideas, why people think they are important, and how they play into our thinking and policies here at Harrison.ai.

In March 2023, the Future of Life Institute published an open letter, calling for a moratorium on the training of large language models (LLMs), the intelligence engine underpinning applications such as ChatGPT. The letter garnered the signatures of more than 25,000 supporters, among whom are famous AI researchers, scientists and public intellectuals such as Max Tegmark, Yoshua Bengio, Yuval Noah Harari and Stuart Russel. This show of intellectual solidarity followed several months of groundbreaking discoveries of what this new class of models can do. Never before have we seen AI so compellingly generate original jokes, explain complex topics in science and mathematics, write and debug code, generate Q&A exchanges and translate between multiple languages. Most profoundly, ChatGPT displays emergent properties which weren’t explicitly part of its training repertoire. In their ability to seemingly create novelty, we face the undeniable risk that such systems are inherently unpredictable. When we stretch our imaginations of what would be capable with even larger models trained on even more data, the uncertainties of intended and unintended consequences grow rapidly.

Visions of the Future

In 1843, Ada Lovelace was one of the first people to understand the transformation that computers would have on society, guessing that one day they “might compose elaborate and scientific pieces of music of any degree of complexity or extent” (see OpenAI JukeBox). However, she was sceptical that it could “originate anything* [on its own, doing ] whatever we know how to order it to perform”. In other words, humans would remain at the centre of knowledge creation. In 1950, Alan Turing referred to this as her “Objection”, countering it with the idea that if put in control of its own design, and operation, “…at some stage, therefore, we should have to expect the machines to take control*”. Recursive self-improvement would naturally lead to an “intelligence explosion”, as described by Irving John Good, a state which would see machines far exceed our (human) ability to keep pace. Such a vision has fuelled the curiosity and nightmares of people ever since.

Problems and Knowledge Creation

It is true that the promise of AI is equally matched by its peril, the possibilities of both should be taken seriously. Yet, whatever path we choose, ever-present will be new problems across all scales. Small problems, severe problems, dangers, all the way up to existential dangers. Some at the extinction level, and others causing suffering and tragedy on such a scale as to merit the same amount of concern as the former. We always have faced such dangers, and always will. The continuous stream of problems is only resolved via our constant stream of knowledge creation. Ultimately, it is our unwavering commitment to the pursuit of knowledge that will determine whether we thrive or falter in the face of all ever-present dangers, including those posed by AI.

For our society, investing in education and research is essential to promote a culture of innovation and critical thinking that encourages the exploration of new ideas and perspectives that promote optimism about AI. Such an approach is crucial because AI has the potential to revolutionize our understanding of the world, offering new ways to solve complex problems and advance scientific discovery. For instance, in just one year, Deep Mind’s AlphaFold has dramatically expanded our knowledge of 3D protein structures from 1 million to over 200 million, approaching the expected total number of structures encoded by the human genome. This underscores the tremendous potential of AI to accelerate scientific progress and drive positive change across a range of domains.

A Positive Stance on Problem-Solving

At Areté Informatics we understand that there are risks ahead. That is why we’ve declared our commitment to safe, ethical and beneficial AI in our AI Principles statement. It covers a range of topics, including

1. Safety: we are committed to developing AI that is safe and secure, and that minimizes the risk of harm to people and the environment.
2. Transparency: we are committed to promoting transparency in the development and deployment of AI systems, including making information about how the systems work and their potential impact on society more widely available.
3. Fairness and non-discrimination: we recognize the importance of ensuring that AI is developed and deployed in a way that is fair and does not discriminate against any particular group or individual.
4. Privacy and data protection: we emphasize the importance of protecting the privacy of individuals and ensuring that personal data is handled in a responsible and transparent way.
5. Human control of AI: we are committed to ensuring that humans retain control over AI systems and that these systems are developed to augment human capabilities rather than replace them.
6. Societal and environmental impact: The document highlights the importance of considering the broader societal and environmental impact of AI systems and ensuring that these systems are developed and deployed in a way that promotes sustainable development and benefits society as a whole.

Overall, the AI Principles document serves as a guide to ensure that the development and deployment of AI align with the organization’s values and commitment to responsible innovation.

As AI continues to evolve and shape society, we must work towards developing AI systems that align with our values, address potential risks, and promote transparency and accountability. We understand that while the risks associated with AI cannot be entirely eliminated, we must remain committed to our pursuit of knowledge and innovation, working to minimize the potential for harm and maximize the benefits of this powerful technology.