The AI Revolution - Transformers and Large Language Models (LLMs) and Why You Believe The Things You Do

The AI Revolution - Transformers and Large Language Models (LLMs) and Why You Believe The Things You Do

In recent years, one of the most significant breakthroughs has been the emergence of Transformer models in the field of natural language processing (NLP). These models, initially invented at Google but later adopted by OpenAI, have paved the way for the development of large language models (LLMs) such as GPT-1 and GPT-3. Transformers and NLP, though still in their infancy, are expected to play a crucial role in the next five years.

Language is at the core of many enterprise activities, from legal contracts and code to invoices and sales follow-ups. The ability of machines to robustly interpret and act on information in documents will bring about a transformative shift, similar to the impact of mobile or cloud technologies. Already, we can see the applications of LLMs in tools like GitHub Copilot for code generation or sales and marketing tools like Jasper or Copy.AI.

For startups, the challenge lies in determining whether to build a de-novo product/market or to enhance existing products with AI capabilities. Sometimes, the best approach is simply to try it out. Startups thrive on iteration and taking action, and overthinking or overanalyzing can hinder progress. The possibilities for consumer applications, enhanced search, interactive chatbots, and even an intelligent agent to replace Google search are immense. Smart commerce is another area where LLMs can have a significant impact.

Looking beyond the realm of startups, large-scale language models have the potential to revolutionize professions such as healthcare and law. AI could potentially replace many tasks carried out by doctors and lawyers, transforming the way these industries operate.

However, the journey towards fully harnessing the power of LLMs in startups and various industries poses a question: are the challenges primarily scientific or engineering problems? While there is room for advancements in algorithms and architectures, incremental engineering iteration and efficiency gains also play a vital role. Semiconductor innovation can dramatically enhance the performance of AI systems, and historically, major technology waves have been accompanied by the emergence of underlying semiconductor companies.

In the realm of AI research, there is ongoing debate about the timeline for achieving Artificial General Intelligence (AGI). Some experts believe it could be anywhere from 5 to 20 years away, similar to the perpetually "5 years away" status of self-driving cars. Whether AGI becomes a reality sooner or later, its development will undoubtedly impact various fields and industries.

On a different note, let's delve into the realm of beliefs and why we believe the things we do. Our beliefs are often influenced by how much we want them to be true. The more something offers hope or helps us deal with uncertainty, the more likely we are to believe it. This is especially evident during challenging times, such as the Great Plague of London, when people were more susceptible to prophecies and old wives' tales.

In the legal world, there exists a phenomenon known as Gibson's Law, which humorously states that for every PhD, there is an equal and opposite PhD. This highlights the fact that people can find expert opinions to support almost any argument, often driven by motivations such as justifying past actions, protecting reputations, or maximizing income. The allure of a belief can overshadow its truth.

Our memories also play a significant role in shaping our beliefs. We tend to emphasize certain memories while discarding others, based on what makes good stories, confirms stereotypes, or connects dots between different experiences. It is easier to believe falsehoods than admit mistakes, leading us to cling to false beliefs and resist changing our minds.

A scientific lifestyle, however, revolves around changing one's mind when faced with conflicting information. Intellectual inertia should be avoided, and the ability to adapt and revise beliefs is crucial. Richard Feynman, a renowned physicist, emphasized the importance of distrusting experts and embracing skepticism. However, herd mentality and blind faith in authority figures still prevail in society.

Logic should form the basis of our decision-making, especially in scientific reasoning. Yet, wishful thinking, irrational fears, and cognitive biases often take precedence. Beliefs are not solely about knowledge; they also serve as social signals, showcasing our confidence, intelligence, and ability to convey reliable information.

Ultimately, the desire to eliminate uncertainty often leads us to believe things that have little relation to reality. We seek experts who are willing to change their minds, but we also gravitate towards individuals who exude unwavering confidence. The pursuit of truth is often overshadowed by the need to alleviate uncertainty.

Combining the realms of AI revolution and human belief systems, we can draw actionable advice for navigating both landscapes:

1. Embrace experimentation and iteration: In the world of AI startups, it is essential to embrace a mindset of trying things out and iterating based on the results. Overthinking and overanalyzing can hinder progress. Similarly, in the realm of beliefs, being open to new information and willing to change our minds is crucial for personal growth and intellectual honesty.

2. Foster a culture of skepticism and critical thinking: In the AI field, it is important to question assumptions and challenge existing paradigms. This applies to both scientific and engineering aspects. Similarly, in our personal beliefs, fostering a culture of skepticism and critical thinking can help us navigate through the sea of misinformation and false beliefs.

3. Strive for intellectual humility: Recognize that our beliefs are not infallible and that changing our minds is a sign of growth and intellectual honesty. In the AI domain, this humility can drive advancements and prevent stagnation. In our personal lives, it allows us to learn from our mistakes and make more informed decisions.

In conclusion, the AI revolution driven by transformers and large language models holds immense potential for transforming industries and startups. Simultaneously, understanding the intricacies of human beliefs and the factors shaping them is essential for personal growth and intellectual integrity. By incorporating actionable advice such as embracing experimentation, fostering skepticism, and striving for intellectual humility, we can navigate these realms more effectively and drive progress in both domains.