Running With AI
The discussion about AI is so pervasive that it cannot be escaped. It has fully lodged itself in our psyches and in the current zeitgeist. Despite the numerous discussions about it, most people have a difficult time understanding its practical implications in their daily lives. I often talk to my friends about it and they ask me the same thing:
“Beyond looking things up, I’m struggling to understand how I can use AI?”
This is because the vast majority of people are using these tools as a more intelligent search engine. I understand the confusion as there is always fog around early-stage transformative technology.
The magnitude of the upcoming change is also difficult to wrap our heads around. The best analogy is that of electricity and its subsequent ubiquity during the last industrial revolution. If you were around in the late 1800’s when electricity was just kicking off, and you had a basic understanding of how it worked, all you would see around you would be ways to utilize it to change people's lives. Similarly, having a basic understanding of AI will help you see ways it will be transformative. Moreover, unlike electricity in the early days, everybody has access to AI and the ability to “play” with it.
You can go online today, talk to an AI agent about any piece of software you want to build, and have that agent create it for you and iterate with you. Don’t believe me? Stop reading this article and go spend just 5 minutes at replit.com. If your mind is not blown when you get back to this article, you probably don’t have a pulse.
The incredible pace of model strength and growth has shocked me. It feels like models are improving and dropping on a weekly cadence now. We used to wait in excitement and anticipation for a model to release every few months -- just 6 months ago -- and even that was incredibly fast, but now these exponential technical leaps are occurring almost on a weekly basis. Just over this past week and a half, Anthropic’s most powerful model – Claude 4 – was released. A few days earlier, Google announced Med Gemma, and yesterday OpenAI announced reinforcement fine-tuning using o4 mini. Practically speaking, you may be wondering what all of this means.
The early days of GenAI conversations focused a lot on limitations such as hallucinations, the inability to recall information, or to reason effectively. These conversations started entering research circles and went from murmurs to conference presentations, but what was not clear to everyone watching was that those very conversations became stale the moment they reached the presenters’ lips. That’s because reasoning or “thinking” models were already arriving. Every impulse by us to say “but what about…” is being met with an answer much faster than we could have anticipated. This trend will continue and the pace of better and stronger models arriving will only accelerate. This is the hallmark of exponential growth.
Behind the scenes, these models are being incorporated into practical and highly integrated technical stacks, including healthcare. I was at SIIM recently and spoke with a colleague and friend Jae Zayed, VP and CIO at Southwest General Health Center who showed me a fully integrated AI Agent embedded at his hospital. Jae demonstrated the agent calling him on his phone, speaking to him in such natural language that I could not tell that it wasn’t a real human, rescheduling an appointment for him, and answering any questions he had. After he hung up, the AI agent orchestrated the scheduling of the appointment in the hospital’s scheduling system and fired off alerts to clinical and clerical staff.
There are agents that will create production grade applications for you if you don’t know how to code. There are agents that will create endless marketing content for you. There are agents that will answer support calls for you. There are agents that will track meeting mins for every meeting you attend. The list of agents available to you is mind-blowing and endless. Also – let me clear – this is not like the beginning of GenAI where agents were mostly hit or miss. Like every technological revolution, reputable companies with products used by millions of daily active users have formed. Compounding all of this are chatbots like ChatGPT that have become ever more effective in providing guidance and advice on using these tools. Businesses will need to adapt and adapt quickly. We’re in a fortunate position where RADPAIR was founded embracing these technologies from the very beginning. Our entire business is optimized by AI and we use these tools every day, in all facets of the entire business lifecycle – be it in coding, code management, operations, marketing, and even down to minute taking in meetings.
If you’re motivated and want to take advantage of this exceedingly rare time we live in, then go and try these tools. Don’t know where to begin? Just start from first principles – ask the most basic questions with your LLM of choice and go down the rabbit hole. You’ll be stunned by how one question leads to trying something new, and that new thing you tried may lead to solving a previously intractable problem.
Human ingenuity progresses in a step-wise fashion and by using these incredible tools effectively, we can start running instead of walking.
_
Written by Dr. Avez Rizvi | CEO - RADPAIR
