Well, everyone is chatting about ChatGPT. I will too.
My data science career to this point has been more of an analyst or maybe statistician than a data scientist. Working in small healthcare tech for the past almost 8 years, my work has never gone anywhere beyond a simple xgboost model and 99% of the time involves basic analytics and regression. But outwardly, we see things like AI, autonomous diving, and now ChatGPT being hot topics in the press and on professional and social media outlets. For me, there’s this huge disconnect in what I see being the more popular definitions of data science and what goes on at work.
So, I should be super impressed with ChatGPT, right? I mean, you can’t deny the impressive technical capabilities. I read that GPT-4 has a 100 trillion feature space. Companies I worked for have a hard time getting 10 features in a model (or getting a model at all for that matter), so 100 trillion is a hard number to get my head around. It’s also very nice and round, so immediately suspect to me. Nevertheless, it’s clearly an amazing technical feat. The Verge has a nice, quick summary on it. OpenAI must have a compute and electricity bill that rivals most anyone. I’m sure DHH at 37Signals would salivate while critiquing their tech stack spending.
I was doom scrolling through LinkedIn earlier and came across a post from Generative AI that gave an example of ChatGPT and its usage. The use case was a chat session with Steve Jobs about the covid-19 pandemic. It should be obvious that Steve Jobs passed away 8 years prior to the beginning of the covid pandemic. Is this really what we are working on? Seems like the uncanny ability of the data science world to focus on all the wrong things would make congress blush.
I think AI is developing a bit of a PR problem. While there’s no doubt interesting and valuable uses for ChatGPT and the business cases appear almost endless, we are seeing more things like this: chatting with a fake Steve Jobs, writing research papers in lieu of actually doing thinking for yourself, writing code when you don’t know how to code, taking the bar exam, and developing fabricated photos. I mean, call me old fashioned, but I’d rather go ride my bike up in the mountains and eat a PBJ with a bestie while watching the sunset as opposed to sitting in my underwear chatting with a fake Steve Jobs on a slow Friday night. And just like 10-15 years ago with machine learning, horribly misguided executives are rushing around with FOMO to find a problem with a ChatGPT solution and hire a data science team only to get rid of them 3-5 years later because of lack of ROI. I can’t even. It feels like another cycle all over again.
Even the Verge article above talks about applications of ChatGPT in the context of fabricating art and creativity, synthesized music, and test taking. Again, impressive from a tech standpoint, but my question is, do we even want or need this? It makes it feel more like a plaything than a serious technology. Can we focus on something useful, that’s making life better for us? One interesting application I’ve seen is summaries of meeting notes being generated automatically. Yes please, more of things like that. We don’t need to tell the difference between pictures of sloths and blueberry muffins. But is the juice worth the squeeze, so to speak? It seems like a lot of time and money investment for things like better chatbots. I posit that we really don’t know what to do with this technology yet. That’s likely the reason we are seeing a lot of fluff. My fave podcaster Prof G discussed business cases on the latest episode of Pivot, one being a tool to profile and risk assess banks for failure. Interesting, ok, but doesn’t machine learning methods already do things like this? Maybe ChatGPT is just packaged up nicely and is more sleek than a group of nerdy PhDs digging around in your data for 6 months.
I’m not being any kind of naysayer here. I think it’s an amazing step forward in our AI journey, but it’s early still. I suspect like ML, it’s going to be the hotness for a while, followed by a lull, then a more thoughtful time of application and investment. Enjoy the ride.
Excuse me while I put on my stretchy pants I go ask Steve what he thinks about the war in Ukraine.