My summary of the AI talks at the CogXfestival2023

As a recruiter in the AI space, I believe it's crucial to stay updated on the latest AI innovations, concerns, and successes. My company, Logikk, has supported this and encouraged me to attend the Cog X Festival 2023 in London, which took place yesterday.?

During the event, I had the opportunity to attend some inspiring talks and gained valuable insights. Some talks left me feeling excited about the possibilities of AI, while others shed light on the potentially ominous aspects of AI's future impact. I'd like to take a moment to share some of the highlights from these talks.?

In the opening talk, an interesting idea was raised: one of the major fears associated with AI is the possibility of it taking over our jobs. The speaker posed the question: is this really as frightening as we think? Much of our daily routines involve mundane tasks, and with AI automating these repetitive aspects of our work, we could have more time to nurture creativity and engage in activities we truly enjoy. An analogy was drawn to the way we use machines to wash our clothes, saving us from the tedium of handwashing. This shift would naturally necessitate changes in how we are compensated. Instead of being paid based on hours worked, as has been the historical norm, we should transition to a system that values the quality, skill, and outcomes of our work. At Logikk, we have already embraced this forward-thinking approach with a four-day workweek and flexible working hours, provided we meet our objectives. This progressive mindset will become increasingly important as AI assumes a more significant role in our professional lives.?

I also had the privilege of attending a talk by Lila Ibrahim, the Chief Operating Officer at Google DeepMind. She shared how AI is being harnessed to address some of the world's most pressing challenges, such as climate change, disease control, plastic pollution, and industrial pollution. DeepMind's work is described as "Science at a Digital Speed" due to the remarkable progress made with AI. They have developed superior models, enhanced predictive capabilities, and gained a deeper understanding of the world through AI. For example, their weather forecasting model, which employs Deep Generative Models for rainfall predictions, is hailed as the most accurate available.?

Lila emphasized the positive impact of tech advancements on infrastructure, reducing energy consumption and costs. DeepMind is also pioneering sensor-fusion solutions to generate power more sustainably. However, she highlighted the complexities of pioneering responsible and ethical AI while striving for growth. To address these challenges, she outlined three critical angles: responsible government involvement in shaping global legislation, inclusive research that considers diverse datasets, and ensuring a responsible impact by involving all stakeholders, including marginalized communities.??

Lila also menitoned SynthID, a tool for watermarking and identifying AI-generated images. As AI assumes a larger role in our lives, trust and transparency become paramount. Identifying whether content was created by humans or AI is essential.?

Overall, Lila's talk left me feeling optimistic about the positive strides we can make with AI, provided we remain committed to responsible development.?

Next, I had the privilege of listening to Jürgen Schmidhuber, a luminary in the field of AI. He provided insights into the history of AI development, from its inception to the present and beyond. Since the age of 15, he has aspired to create a self-improving AI capable of surpassing human intelligence. His labs in the 80s and 90s pioneered Deep Learning AI neural networks, generative adversarial networks, transformer variants, and meta-learning machines. He noted that he is closer than ever to realizing this dream, though not quite there yet.?

One intriguing aspect he discussed was the declining cost of computing power, which becomes 10 times cheaper every four years. This trend allows computers to perform more tasks at a lower cost, making highly skilled AI accessible to the masses. Schmidhuber provided historical context by mentioning the Z3, the first general-purpose computer, which could perform only one instruction per second in 1941. Ten years later, it could execute ten instructions per second for the same cost. By 2024, we will be able to perform one billion times one billion instructions per second for the same cost. This exponential growth suggests that AI will increasingly outperform human capabilities in the near future.?

However, Schmidhuber pointed out that we still have much work to do. He used ChatGPT as an example, explaining that while it excels at indexing human knowledge and answering natural language questions, it lacks true creativity. It relies on existing knowledge rather than generating novel experiments and data. Schmidhuber's vision is to create AI that can drive innovation by generating new patterns and data, potentially transforming society in profound ways.?

He correctly noted that, while ChatGPT is currently one of the most exciting developments in AI, it will likely be considered primitive in 30 years, as AI continues to advance exponentially.??

Another talk I attended was by Yuval Harari, the Israeli historian, author, and philosopher. His perspective offered a broader view of AI, though it left me with some apprehensions. Harari began by emphasizing that throughout history, all significant inventions and creations were made by humans, empowering them in the process with humans having control and making important decisions. AI represents a unique departure from this trend, as it has the potential to create a society where ideas and actions originate from AI rather than human minds. He likened this scenario to an alien invasion, where aliens promise to solve humanity's problems but demand control and power in return, a prospect I found quite alarming.?

Harari discussed the concept of time scales, pointing out that societal adaptations typically occur slowly over generations. However, AI's rapid digital growth presents challenges. Historically, humans learned through trial and error over time, but with AI's ability to impact society quickly and on a massive scale, we have limited room for mistakes. Harari drew an analogy to early aircraft development, focusing on faster engines without considering wing design—a potentially catastrophic oversight.?

He also drew parallels with totalitarian regimes such as the Nazis and the USSR, highlighting their inability to monitor all citizens and analyze all the information they collected on individuals. With AI and the widespread use of smartphones, we can be tracked, listened to, and assessed continuously. Our devices collect data about our movements, preferences, and interactions, which can be analyzed to extract valuable insights easily. This underscored the potential for misuse of AI in the wrong hands.?

Harari proposed that we have a limited window of 5-10 years to retain control over AI's trajectory. He suggested strategies such as distinguishing between development and deployment, allowing AI companies to only deploy once they can ensure safety. He also advocated for allocating a portion of AI companies' profits to focus on making AI safe. Additionally, he emphasized the importance of regulations, proposing a 20% tax to fund institutions that work closely with industry experts to create understandable and effective regulations.?

Harari highlighted China's progress in AI regulations, particularly the decision to ban bots from social media and remove products from the market if rules are violated. He urged nations to come together and form a united front, as AGI's potential impact rivals that of nuclear weapons or global pandemics.?

Finally, I had the privilege of listening to Stuart Russell, a renowned figure in AI and the co-author of the book "Artificial Intelligence: A Modern Approach." He began by asserting that we must assume that AGI, capable of exceeding human capabilities in all relevant dimensions, will emerge within the next decade. He emphasized that while large language models (LLMs) perform admirably in imitating general-purpose AI, they are essentially advanced text predictors.?

Russell revealed that a staggering 10 billion pounds per month are being invested in AGI, a sum equivalent to the entire budget allocated by the USA for all non-medical science and technology research. Significantly, a negligible portion of this budget is dedicated to safety, responsibility, and ethics. He likened this approach to sprinting towards a precipice at full tilt, with the hope that someone, somewhere will develop a parachute in time. This metaphorical parachute represents regulations, safety research, and a shift in mindset toward responsibility.?

In line with Yuval Harari's suggestion, Russell proposed that governments establish rules, with severe consequences for companies that breach them, including immediate product removal from the market and a fine amounting to 10% of revenue. This approach, he argued, would incentivize companies to prioritize ethical considerations. Russell also mentioned China's existing law, which mandates the removal of LLMs that disseminate false information from the market.?

Furthermore, Russell highlighted that the United States and the United Kingdom currently lack clear red lines (although the UK plans to enforce an act in 2025). He emphasized the upcoming summit and the necessity for global cooperation to establish a united front, as AGI possesses the potential for devastation on par with nuclear weapons or global pandemics.?

A novel proposal introduced was the training and licensing of practitioners in the field of AI, equipping them with a set of guiding principles.?

Stuart concluded by quoting Aristotle, who, in 350 BC, envisioned a future with automated machines that would create music and weave, rendering human labor unnecessary.?

In conclusion, While I find it challenging to envision a future without a need for workers, I can certainly perceive how humanity's relationship with work has evolved alongside technological advancements. In the late 19th century, the average working week in the US exceeded 60 hours; today, it hovers around 33. The COVID-19 pandemic has reduced the necessity for commuting to offices, with many now working remotely. Additionally, progressive companies have experimented with and adopted a four-day working week. This has allowed us to dedicate our work time to more meaningful and vital aspects of our roles, rather than mundane and time-consuming tasks. I have no doubt that we will continue on this path of enhanced efficiency, affording us more time to focus on important pursuits, ultimately benefiting us and future generations.?

At times during the day, I couldn't help but imagine a dystopian future where AGI falls into the wrong hands, leading to a society devoid of human input and control. Nevertheless, I left the event with a sense of optimism. I believe that by increasing awareness of the importance of responsible AI practices and making strides in terms of regulations and rules, we can guide ourselves toward remarkable achievements that enhance the lives of people worldwide.?

?

?

?