I totally agree that the AI bubble is just that. It has all the usual ingredients. Excellent summary of the market forces and underpinnings thank you. For me AI has been a bit of a productivity tool, and I selectively use it to craft some patient education, and every once in a blue moon i use it to brainstorm diagnoses and possible testing strategies with opaque problems. But I’ve spent 0 dollars on this as I’m sure 99% of the rest of the world has not either. I have a post in the Doximity queue (floundering there for 2 months!) about AI ambient progress notes… anyway, looking forward to following this with you!
Thanks for the thoughtful comments, Ryan! I do spend $20/month on OpenAI's premium ChatGPT tool; how useful I find it seems to vary by the week. Initially, the image generation and computer vision analysis tools were like magic. I was able to feed in Excel spreadsheets and get it to instantly generate tables and figures and provide insightful feedback. It was able to assist with lit reviews by summarizing PDFs.
Then over time, many of those functions degraded. One of the frustrations I, and surely many others, have with OpenAI is they *constantly* change their models in the background without much transparency about what they are doing. After they got sued by the New York Times on plagiarism grounds, they HEAVILY "nerfed" the PDF summarization features. Some of the analysis tools got crappier. Academic studies found the basic math function fluctuated between GPT-3.5, 4, and beyond. To be fair, some things got *better* and GPT-4o clearly runs much faster than GPT-4, but it is hard to feel good about spending money on a tool whose performance fluctuates more than a moody human!
Of course, most of the conversation here is around a small and specific subset of artificial intelligence: Large language model (LLM) based generative AI. In the background, other modalities are quietly making their way into society with less fanfare. Computer vision tools already allow us to unlock our iPhones with our faces, instantly decode scanned documents into text, provide driver assist features in our cars (although I would discourage turning your brain off in a Tesla! 😅), and many more. AI is almost certainly going to find its way into medical applications like radiology and pathology, although I increasingly doubt it will fully replace humans. Simpler machine learning applications are ubiquitous in banking, databases, retail, etc.
The first "dot com" bubble is a great analogy here: While many of those companies were overvalued and/or snake oil voodoo, the invention and adoption of the internet really WAS the real deal and changed society forever (for both good and bad)
I think partially yes, partially no. When you see what some of the big venture capital firms have invested, it is hard not to think about recent tech bubbles. However, unlike say crypto/blockchain/metaverse, I do maintain there is real value here. Whether it matches the euphoric expectations and spending like a drunken sailor is another question...
I totally agree that the AI bubble is just that. It has all the usual ingredients. Excellent summary of the market forces and underpinnings thank you. For me AI has been a bit of a productivity tool, and I selectively use it to craft some patient education, and every once in a blue moon i use it to brainstorm diagnoses and possible testing strategies with opaque problems. But I’ve spent 0 dollars on this as I’m sure 99% of the rest of the world has not either. I have a post in the Doximity queue (floundering there for 2 months!) about AI ambient progress notes… anyway, looking forward to following this with you!
Thanks for the thoughtful comments, Ryan! I do spend $20/month on OpenAI's premium ChatGPT tool; how useful I find it seems to vary by the week. Initially, the image generation and computer vision analysis tools were like magic. I was able to feed in Excel spreadsheets and get it to instantly generate tables and figures and provide insightful feedback. It was able to assist with lit reviews by summarizing PDFs.
Then over time, many of those functions degraded. One of the frustrations I, and surely many others, have with OpenAI is they *constantly* change their models in the background without much transparency about what they are doing. After they got sued by the New York Times on plagiarism grounds, they HEAVILY "nerfed" the PDF summarization features. Some of the analysis tools got crappier. Academic studies found the basic math function fluctuated between GPT-3.5, 4, and beyond. To be fair, some things got *better* and GPT-4o clearly runs much faster than GPT-4, but it is hard to feel good about spending money on a tool whose performance fluctuates more than a moody human!
Of course, most of the conversation here is around a small and specific subset of artificial intelligence: Large language model (LLM) based generative AI. In the background, other modalities are quietly making their way into society with less fanfare. Computer vision tools already allow us to unlock our iPhones with our faces, instantly decode scanned documents into text, provide driver assist features in our cars (although I would discourage turning your brain off in a Tesla! 😅), and many more. AI is almost certainly going to find its way into medical applications like radiology and pathology, although I increasingly doubt it will fully replace humans. Simpler machine learning applications are ubiquitous in banking, databases, retail, etc.
The first "dot com" bubble is a great analogy here: While many of those companies were overvalued and/or snake oil voodoo, the invention and adoption of the internet really WAS the real deal and changed society forever (for both good and bad)
Is AI another tech bubble?
I think partially yes, partially no. When you see what some of the big venture capital firms have invested, it is hard not to think about recent tech bubbles. However, unlike say crypto/blockchain/metaverse, I do maintain there is real value here. Whether it matches the euphoric expectations and spending like a drunken sailor is another question...