This talk was given in March 2024 at Lehigh University Pennsylvania to a 200-person audience.
Ladies and gentlemen, esteemed guests, and fellow members of Rising Together @LV,
It is an honour to address you today at the March 2024 Rising Together event here at Lehigh University, where we celebrate Women’s History Month annually. As the founder of Rising Together, a community-driven organisation rooted in female excellence but inclusive of all genders and ages, I am proud to witness the growth and impact we’ve achieved since our establishment in 2017.
Our 2024 summit, a culmination of talks, activities, debates, awards, and entertainment, centres around a theme close to my heart: Artificial Intelligence (AI) – The Gift, The Challenge, and The Mystery. This theme reflects my deep-seated feelings about AI’s profound implications. It is befitting that I give the opening speech on the summit topic as I have done in years past.
At this pivotal juncture, we stand on the cusp of transformation across various aspects of AI, spanning both software and hardware. Delving into the hardware realm, considerations around silicon technology and data throughput become paramount. Semiconductor suppliers face the challenge of designing for immense data throughput with minimal power consumption and latency, ushering in a new era of optoelectronics and nanometer geometries. Yet, amidst these challenges lie incredible opportunities for innovation and exploration.
Turning our attention to the heart of AI’s impact, Generative Artificial Intelligence emerges as a central concept. Combining the prowess of AI to perform tasks traditionally reserved for humans with the creative ability to generate new content from partial information, Generative AI holds immense promise and potential. While not a new concept, recent advancements, notably exemplified by OpenAI’s GPT series, have propelled Generative AI into the limelight. The journey of the Generative Pretrained Transformer (GPT) backbone, the Transformer is worth tracing from countless researchers to Google Labs to FB DeepMind and eventually to the “amazing eight” from Google Brain who eventually birth the Transformer – from which just about every modern GPT/LLM algorithm now emerge!
Training Process on Neural Nets: Building a large language model (LLM) like ChatGPT involves extensive training on neural networks. This process requires a vast corpus of data collected from various sources, including the web, social media, and more. The data is then passed through a neural network, where the model learns to predict and complete sentences. Through iterative training, the model becomes more accurate in generating responses based on the input it receives.
Description of the Transformer: The transformer architecture, introduced in 2017, has become the king of AI architectures. As neural networks are trained, they learn and eventually culminate in self-supervised learning. This process involves truncating sentences, making predictions, and continually learning probabilities. With advancements like GPT-4, transformers have grown in complexity and size, with billions of parameters enabling them to understand and generate human-like text.
The journey from GPT-1 to GPT-4 exemplifies an exponential leap in AI’s capabilities, underscoring its transformative potential. The advent of models like GPT-4, boasting trillions of parameters, signifies a paradigm shift in AI’s scale and complexity. Yet, with great power comes great responsibility. The ethical considerations surrounding AI, from its energy consumption to its societal impact, demand our attention and vigilance.
As we navigate the intricate landscape of AI, the notion of fine-tuning emerges as a critical mechanism to align AI systems with human values and intentions. By imbuing AI with principles of helpfulness, honesty, and harmlessness, we strive to harness its potential for the betterment of society while mitigating potential risks.
However, as we tread this path of innovation, we must confront the inherent uncertainties and challenges that lie ahead. The question of AI’s ultimate trajectory, whether it heralds a utopian future or poses existential threats, remains shrouded in ambiguity. Yet, through informed dialogue, responsible governance, and collective action, we can steer AI towards a future that benefits all.
Going back to the theme – The Gift, The Challenge and The Mystery –
The Gift: We can acknowledge the gift that AI brings to our lives. AI, particularly generative models like ChatGPT, has revolutionised how we interact with technology. These models serve as multifunctional tools that can assist us in various tasks, from writing code to composing music. The rapid adoption of AI highlights its immense utility and potential to enhance human productivity and creativity.
The Challenges: However, along with its gifts, AI also presents significant challenges that we must address. One of the foremost challenges is the energy consumption associated with training and deploying AI models. As these models grow larger and more complex, they require substantial computational resources, leading to environmental concerns and increased costs. Moreover, there are ethical considerations regarding the potential misuse of AI, including the spread of misinformation and job displacement. Addressing these challenges requires collaboration between stakeholders to develop responsible AI practices and regulatory frameworks.
The Mystery: Despite the progress we’ve made in AI, many mysteries remain. One of the most intriguing aspects of AI is its ability to learn and adapt autonomously. Can AI systems replicate themselves and evolve without human intervention? Furthermore, what are the long-term implications of AI on society, particularly in terms of job displacement and societal inequalities? These mysteries remind us of the complexity and uncertainty surrounding AI’s future trajectory and compel us to approach its development with caution and foresight.
In closing, let us remember that the trajectory of AI, much like the arc of history, is shaped by human agency. It is up to us, as stewards of progress and guardians of our collective future, to chart a course that embodies the values of inclusivity, empathy, and sustainability.
Thank you for your attention, the logical step for us today and always, is to rise together towards a future where AI serves as a catalyst for positive change.
About Rising Together:
Rising Together is a flagship organisation anchored by female excellence and embraced by community, including men and youth. We believe that every robust success should involve all stakeholders. Our Rising Together platform celebrates Women’s History Month @ LV
Rising Together was founded in 2017 by Ngozi Bell
Some cool statistics:
Fun facts:
- GPT-4 GPT-4’s is said to be based on eight models with 220 billion parameters each, for a total of about 1.76 trillion. GPT-4 has more than a trillion parameters – Report (the-decoder.com)
- The human brain has roughly 100 billion neurons. Each neuron has roughly 7000 synapses. If (and this is a big ‘if’) each synaptic connection in the brain is roughly equivalent to one parameter in a Deep Neural Network, then the human brain has roughly 700T parameters. Prediction: Human-equivalent neural networks by 2030 (daveshap.github.io)
- Other large language models include Gemini, Ernie Bot, Llama, Claude, and Grok.
- Microsoft launched Copilot, based on OpenAI’s GPT-4.
- AI processing requires high-performance computing.
- GPT-4 can beat 90% of people on the SAT.
- GPT-4 excels in LSAT and MCAT exams.
- The human brain has approximately 100 trillion parameters with 2.5 petabytes of memory capacity.
