.svg)
Large Language Models (LLMs) like ChatGPT are helping to make AI part of our everyday experience whether we realize it or not. LLMs are quickly evolving as is our understanding of how to use them to be more creative and productive.
We all probably know someone so well that we feel we can complete their sentences and vice versa. We have spent so much time interacting and learning their nuances that it feels like we can anticipate what they will say next. This is kind of like Large Language Models (LLMs), which are deep learning models that have been pre-trained on a vast amount of text (massively huge!) and over time, are able to emulate human speech that they can fairly accurately autocomplete your sentences as you start typing.
Creating an LLM is rigorous undertaking and relies on highly trained professionals to build and supervise the machine learning algorithms. The basic process, however, begins with assembling a large corpus of text data sourced from diverse materials like books, articles, and web content. This corpus is then utilized to train the LLM, instructing it to comprehend and emulate human-like text. The model scrutinizes the given input and applies its training to generate a coherent response that aligns with the context. Key concepts like tokenization, which involves breaking down the text into smaller units or 'tokens', play an essential role in this process and how we interact with the models. LLMs require ongoing training and modifications to enhance its contextual understanding and response quality.
The continuous refinement is vital for maintaining the LLM's relevance and evolving to be less biased and more inclusive. Being less biased means that the Large Language Model (LLM) is not favoring or discriminating against certain groups or ideas based on the data it was trained on. More inclusive means it is able to understand, interact with, and generate responses that are respectful and appropriate for all different types of people, regardless of their background, culture, or personal characteristics.
The lack of inclusiveness, inherent biases, data privacy breaches, and recent legal issues over intellectual property rights are causing some significant challenges for LLM owners. AI is outpacing regulations and ethical standards, which is not completely unexpected from an emerging technology growing at an exponential rate. Still, these challenges could affect public trust, slow down the rate of adoption, and hinder companies looking to pivot to AI as a new revenue source.
While Large Language Models (LLMs) exhibit impressive capabilities in text generation, it's important to clarify that they are not a form of pure artificial intelligence. These models are created by humans, who program them with specific instructions on how to analyze and respond to text. Their capacity to generate human-like text is a result of extensive programming and training, rather than an innate ability to understand or think.
Additionally, the responses generated by LLMs are based on statistical probabilities rather than any inherent understanding. When an LLM generates a response, it's essentially estimating the likelihood of a particular response based on the patterns it has observed in the data it was trained on. It does not truly "understand" the text in the way a human would. Instead, it uses complex algorithms to predict the most likely response. This distinction is important to keep in mind when considering the role and capabilities of LLMs in the broader context of artificial intelligence, which can be thought of in 3 distinct types: Artificial Intelligence (AI), Artificial General Intelligence (AGI), and Artificial Super Intelligence (ASI).
AI, or Artificial Intelligence, refers to machines or software that mimic human intelligence, performing tasks such as recognizing patterns, learning from experience, drawing conclusions, making predictions, or taking actions that help it achieve specific goals. Examples are Siri on your iPhone or recommendations on Netflix.
AGI, or Artificial General Intelligence, is a type of artificial intelligence that has the ability to understand, learn, and apply its intelligence to any intellectual task that a human being can. It is a more evolved and advanced form of AI and can understand, learn, plan, and exhibit wise judgement. It's like a human mind but in machine form. However, AGI doesn't exist yet; we are still in the phase of developing AI capabilities.
ASI, or Artificial Super Intelligence, refers to a time when the capability of computers will surpass humans. ASI is currently the stuff of science fiction and far from our current technological capabilities. It's a hypothetical AI that doesn't just mimic or understand human intelligence and behavior; ASI can improve on it, inventing new languages and even making discoveries humans can't even imagine.
LLMs are gaining popularity in our digital age. In November 2022, OpenAI publicly release ChatGPT 3.5. The media attention surrounding its release quickly elevated the status and awareness of Artificial Intelligence in the public eye. It was a process years in the making. It marked the first time an LLM was really accessible, allowing people to interact with through a chat interface. Within its first few days, a million people were curious enough to give it a try.
The ability to "ask anything" allowed people’s imagination to run wild. After a few months, the enthusiasm slowly waned as people struggled to correctly prompt the models to get any kind of useful response. Over the last year, however, people have fine-tuned their prompting strategies...often referred to as prompt-engineering. With better prompts and better outputs, people are starting to figure out more interesting use cases in a variety of niches.
Today, with the right prompt, you can create amazing images, draft emails and social media posts, or write code, and more. Their ability to understand context and generate relevant, coherent responses is helping to push artificial intelligence into the main stream by making tasks that were once considered complex and labor-intensive more accessible and efficient.
There is an increasing trend in utilizing Language Learning Models (LLMs) as co-pilot tools for various tasks. These include tasks like summarizing meeting notes, generating to-do lists from emails, and creating posts for multiple social media platforms, among others. The primary advantage of these AI assistants is the significant boost in productivity they offer. They allow for less time to be spent on mundane and repetitive tasks and the possibility of having several "AI Personas" working simultaneously. This could potentially open up new income streams that might otherwise be unattainable for a single person.
What began as mild curiosity with ChatGPT to now having a wide breadth of browser plug-ins and paid co-pilot subscriptions, it’s clear that LLMs will be core components of our daily productivity. Like any emerging technology, it takes time to recognize and understand the purpose and usefulness in our daily lives. We are very early in this adoption cycle but momentum is building quickly.
AI Copilot is designed to be used in conjunction with the tools and workflows you are already using. Today, the most established of these co-pilot services target software developers in popular code editors such as Visual Studio Code, IntelliJ IDEA, and PyCharm. Recently, Microsoft has rolled out CoPilot in Microsoft 365 subscriptions and Google with its Google NotebookLM...all running on top of large language models. Productivity tools such as Notion also now have an AI paid plan integrated to help you summarize articles, generate new content, or recommend actionable tasks. Photo and Video editing software also has a lot of built-in AI functionality to create, enhance, and augment your digital creations without having to rely or switch to other third-party solutions such as Adobe's Generative Fill feature. We working on another article that will dive deeper into these AI "Co-Pilot" services and software plug-ins but the takeaway for now is that AI is ready and available today...in many of the tools we use and love everyday. And that list will continue to grow.
LLMs like ChatGPT are helping to advance the adoption of AI and allowing individuals and business to capitalize on its ability to optimize productivity and efficiency. With the release of GPT-3 and GPT-4 and several others like Google's Gemini (formerly Bard) or Anthropic's Claude, people now have choices to use a model that best serves their needs. With existing automation tooling and readily available APIs, the sky's the limit for tailored AI solutions.
AI hasn't yet reached it's "iPhone Moment" of mass adoption but it's quickly gaining momentum. With services like Microsoft 365 CoPilot being integrated into the software we use daily, the adoption rate will likely accelerate exponentially over the next couple years. Large tech companies like Microsoft, Google, and Apple are betting big on the future of AI so it seems only a matter of time before it’s baked into every device we own and part of everything we do…both for work and personal use. Regardless of how we feel about using AI and its implications, we are going to have to adapt and change our mindset in order to lead the change that's coming instead of risking falling behind.
For 25 years, Kyle has been a leader in digital transformation initiatives at large financial firms and owns and operates an AI automation agency focused on delivering practical, cutting edge AI applications for small and mid-sized businesses.
Read My Full Bio