You Ask, I Answer: What About New AI Experts?

In today's episode, Chris asks, does it bother you that so many people are now, quote, AI experts all of a sudden? No. And here's why. I've been working in machine learning and artificial intelligence for about a decade now, a little bit more than a decade. 2012 is when I really started getting interested in data science in machine learning, uh the R programming language, etc.

And in that time, it's been an uphill battle trying to explain to people what artificial intelligence is, why they might need it, uh, what it can do for them. And uh adoption has been slow. Adoption has been challenging to get people to recognize that the technology delivers better results and that they can use those results. When you have stuff like chat GPT and DALI and stable diffusion, etc., creating all these things, uh, you know, writing copy, making images, composing music, etc. That's a that's a good thing, right?

That's a good thing. We want people to use these tools. We want people to understand what these tools can do for them. And yeah, are there a lot of people who are claiming to be AI experts and chat GPT experts overnight? And, you know, all these companies, uh, startups doing large language models.

Of course, there's a lot, there's tons of them. There's a lot of opportunity there. It's what we've been saying for 10 years, there's a lot of opportunity here. When you have kind of this gold rush mentality, it's a good thing for everybody because it creates more visibility, right? Talking to somebody about attribution modeling and markov chains and such like that, that doesn't really get anyone excited, right?

It's very complex math. Telling somebody, hey, this machine can write a blog post for you, they understand that. They understand that, they get it, they see what it can do for them. And will you have a lot of uh snake oil salesmen that pop up when you have some a new field of technology? Of course, of course.

We had tons of podcasting experts come out of the woodwork in the early days of podcasting. Remember all those folks telling you what exact microphone to buy and uh all these different things, and you know, most of them were flash in the pan folks that just kind of vanished over time. Uh we had no shortage of SEO experts uh you know and and webmasters back in the day. Uh we had no shortage of crypto experts and NFT experts uh back in the day. And now we have no shortage of AI experts who may or may not have any expertise.

But the conversation, the chatter, the awareness is good for everybody, right? We want people to be trying this stuff, we want people to be adopting it and implementing it. And this is the part that I think is probably the most reassuring for people who have actual expertise. The definition of an expert, I'm sorry, my definition of an expert, is someone who knows what will go wrong, right? An expert is someone who knows what will go wrong so that they can avert it or mitigate it, or you know, deal with the consequences.

When you have a lot of snake oil salesmen rushing into a space, you know, it kind of reminds me of the ongoing joke. Uh, you know, what do you call an aerobics instructor? You know, someone who's taken one class more than everyone else in the class. You have a lot of folks like that in the AI space now who don't have actual expertise. And that's okay.

Um, because again, it creates visibility and it creates opportunity. And the opportunity is this. Those who don't have expertise, when they try to convince somebody or try to lead somebody through a complex project involving artificial intelligence, things will go wrong. And when those things go wrong, if you don't have expertise, you tend to make a pretty big mess of things. You tend to do things pretty badly.

And if you have actual expertise, you can mitigate some of these harms up front. And more importantly, when you run into somebody who's like, oh, you know, we we we tried doing this AI thing with this other person and um and it didn't really work out, you can ask them very specific questions, say, Well, did this happen? Did this happen? And they're like, Oh, yeah, yeah, this, you know, that these are all the things that went wrong. Uh, and they're like, How'd you know?

Like, well, it's what people who actually know what they're doing would would do about it. We see this all the time with things like Google Analytics, right? Where someone who's a Google analytics expert, but not really, um, goes in and makes a hash of things, and then you come in and they're like, well, okay, here's the five things that they did really wrong. This is set up backwards, fix this here, change this in Tag Manager, and and you're fine, right? And so the fakes, I wouldn't call them fakes, because you know, they do have some knowledge, just not very much.

The uh wannabe experts create a lot of opportunity for the real experts to clean up their messes. Um because of that, you can often you can also often bill more, right? You know, um, if somebody commissions a model for for uh usage and you know they charge them a quarter million dollars and the model just doesn't do what they wanted to, you say, like, yeah, I can either try to retune the model that you've got, or we can just start from ground up, but it's gonna cost you half a million dollars to do it this time. Now, in the long term, is that bad? Yes, in some ways, because it can it can scorch some of the earth, right?

There'll be people who will try an AI project with a wannabe expert and be turned off by it. They're like, nope, didn't work for us, it was a failure, etc. Not realizing it's not the technology at fault, it's the person who's leading them through the technology. But the field overall, because it is on solid ground, because it is backed up by real science and real expertise, isn't going anywhere. So eventually those folks might be persuaded to give it another try.

But for the rest of the folks who know what they're doing, it's a good thing. We want more eyes on the space because we want people to benefit from these tools and and do cool stuff, right? So does it bother me that there are so many uh quote AI experts? In relatively short order, people will figure out who knows what they're doing and who doesn't based on the things that will or won't go wrong. If you would like to avert this when you're talking to an AI expert, and you're not sure whether they have expertise or not, start asking them questions about the things that will go wrong, right?

Um and you'll for real experts, you will get some very, very specific questions back about your use case, about what it is that you're trying to do, and then you're gonna start getting really technical questions, right? Show me your data set. What kinds of tuning have you done in the data set? What kinds of um uh detection for uh anomalies are in the data, what kinds of thing uh biases are in the data, and you'll spend a lot of time hearing about your training data set. You'll spend a lot of time hearing about what model choice, what application, what API you're gonna use.

And that's when you know you've crossed over from hey, here's a cool chat GPT prompt to oh, okay, this is the this is real enterprise software. Um that's where the rubber meets the road is in the implementation of complex systems and all the things that go wrong. So it's a really good question. Um I can definitely see where there would be times and places where people with real expertise are like, well, I've been working at this for 10 years, and why is this guy over here who is hawking um you know cryptocurrencies last week, why is he getting all the limelight? I can totally get what that comes from.

But at the same time, if there was no interest in the guy over there hawking uh cryptocurrencies, who's now hawking AI solutions, it would mean that the market still wasn't ready for us, right? Still isn't ready to even have the conversation. And if it takes the the guy selling snake oil to open some doors, maybe that's his purpose, right? Maybe that's his role in the AI revolution. Who knows?

Anyway, really good question, complex question, kind of a uh a loaded question, but a good one nonetheless. Thanks for watching. If you like this video, go ahead and hit that subscribe button.