You Ask, I Answer: Staying Current on AI Tools?

In today's episode, Julie asks there are so many competitive tools in AI that are continually evolving. How do you stay current and align with the best tools based on what is needed? The answer to that question is at the very end of your question. You gotta know what's needed, right? You have to know what is needed in order to evaluate tools.

It's uh the Trust Insights 5P model purpose, people, process, platform performance. What is the purpose you're trying to accomplish? Once you know that, the rest can start to fall in place. But if you just go in tool shopping with no clear idea of what you want or what you need, it's not going to work out very well for you. Imagine going to you know, Home Depot or Walmart or Target or the store of your choice, and just start buying kitchen appliances.

Like you don't have anything you want to cook in specific, you just start buying appliances, a blender, frying pan, air fryer. And you're by the time you're done, you're like, oh, well, this looks like the best uh blender. And then later on, retroactively you go, oh, you wanted us to make steak. Guess this blender won't help all that much, right? That's the problem with the tool centric focus.

A tool centric focus is one of those things where, yeah, there's you're gonna pick the wrong things because you're you're not focused on what you need, not the tools. Now, how do you keep up with the sheer number of tools? In the AI space, here's the thing there are hundreds of new AI tools every week, and these are mostly what I call wrappers. A wrapper is exactly as it sounds, you know, like a candy wrapper, something that is that is around a piece of something else. These tools are wrappers around other people's APIs, like OpenAI or Anthropic, or you know, name the service of your choice.

And as a result, they are there may not be a whole lot of value added on top of what they have built, right? It depends. It's a case by case thing. But many of these tools, when you when you look at them, when you read about them, and you go, huh, this tool is really just an interface onto GPT 4 or Claude 2. And it's an interface that's essentially just translating what you what would be a prompt you would typed into something slightly easier to use, but it doesn't have a huge amount of value add on top of it.

And then you have a company essentially charging money for something you probably could have done on your own. So once you know what you need, the second thing you should understand are the foundational technologies, right? What are the pieces that you would need if you wanted to build it yourself? I'm not saying you have to. In many cases, you probably should not.

But knowing what's running behind the scenes, what's under the hood can be a great help in narrowing down what it is that you want to do. So if you're doing language modeling, you should start by saying, okay, well, what are the foundational ways to use language models? There are models that are APIs, like ChatGPT is essentially an interface onto the GPT 4 model. So it's an it's an interface. You can you connect directly with the GPT-4 model, and many, many software companies are doing exactly that.

So that's one sort of foundational piece. You could connect to anthropic. You could use be using the llama family of models from Facebook or the hundreds of derivatives of that. You could if you're doing voice, you could be doing uh interface with something like 11 Labs uh or Tortoise Text to Speech. Knowing the foundational pieces and how you would build something, then makes it a lot easier to vet vendors, right?

You know, you know that enough to know what's under the hood, even if you can't take it apart, right? It's like opening up your car hood. You can say, okay, well, there's the engine block, there's the battery, there's the the you know, windshield wiper fluid tank, and you put the hood back down. I don't need to take apart my car to open the hood and go, okay, I know what the the general areas of my engine are, but knowing that gives me some understanding about how the vehicle works, about what it can and can't do. So if you have a foundational understanding of the different technologies, you know, voice generation, image generation, text generation, then when you start talking to vendors and evaluating different tools, you can say, okay, this vendor is probably an ensemble of three or four different tools, or maybe this vendor has their own thing that they're doing that is custom and unique to them.

You know, maybe that's that's something that there's lasting value there. I personally would be looking at vendors that are using open source models, because a system, a service like, for example, the GPT 4 API from OpenAI. If OpenAI went out of business tomorrow, not saying they're going to, but if they did, you'd kind of be stuck. And if you're using a tool that is going to be part of your bread and butter, you know, you the the strategic advantage of your marketing, you want something that's going to stick around, right? If Meta went away tomorrow, the Llama model would still exist.

Because it's been mostly open sourced, it it's going to exist. It's independent of Meta now. So they've made it public and now people can run with it and do things with it that they can't control anymore. So that's an aspect as well. It's looking at tools and how much control a vendor has over their underlying tech stack.

These are all questions you would ask to the salesperson, to the sales engineer, especially when you're eventuating software, and you're evaluating, you know, is this the right tool for the job? What is the job? What's the right tool for it? So those are some of the ways that I like to look at the myriad of technologies and the questions I'd like to ask to see who is who's who's the real deal and who's not. One last piece of advice.

If a company will let you talk to engineering unchaperoned by sales, there's a good chance that that company knows what they're doing, right? That they're not selling smoke and mirrors. Because engineers generally tend to be fairly honest, right? Uh I remember talking with one vendor, the sales person or the sales guy was like, oh yeah, our product does this, this, and this. And I went and talked to the engineers.

The engineers are like, we don't do any of that. Uh what the heck is he talking about? They said a little more rudely than that. But it's like he's making things up. Our product does not do those things.

Our product does X, Y, and Z. When I when I vet vendors for myself or our clients, I try to talk to engineering. I try to ask engineers, like, what is the product actually do? What are its limitations? What challenges are you running into right now?

And the more access I'm given to engineers at a company, the more faith I have in that company that they don't have anything to hide because they're willing to let they'll let the nerds talk to each other, essentially, and the sales folks in particular at those more progressive companies, understand that, yeah, the engineers may say things that could be deal breakers, but probably not if they feel comfortable enough letting people talk to engineering directly. Anyway, that's how I stay current on AI tools. Um your process will probably vary, but those are some things to think about. Thanks for the question. We'll talk to you next time.

If you like this video, go ahead and hit that subscribe button.