Mind Readings: Generative AI Backup Plans

In today's episode, let's talk about your generative AI backup plans. Today's market leader, like OpenAI, could be tomorrow's Google Reader. For those who may not know the reference, uh, once upon a time, eons ago in internet time, there was an absolutely amazing blog reading software platform service from Google was free called Google Reader. It was fast, it was convenient, it was free, and as Google often does, one day they just turned it off. And said, nope, we're not doing this anymore.

Which, again, you know, it's free. And it left this massive gap in the marketplace that no company is ever really successfully filled. Although many worthy companies provide similar functionality, companies like Feedly. But Google Reader's absence is notable, partly because of its utility with really good software, and partly because of its transience. Here today, gone tomorrow.

Now, to be clear, we weren't paying for it, and you get what you pay for. But this is one in a long legacy of Google products like a Casa and Orcut and a variety of others that just vanish. And it's not just Google. Obviously, there are tens of thousands of software companies that are uh uh dried up husks on the side of the road, if you will. But this lesson, this transience of companies and services, is critical for anyone working in AI today with tools and software and services and models.

AI today is in its infancy, and the thousands of thousands of AI companies popping up like weeds everywhere, for the most part, are probably not going to have staying power. Some are gonna run out of runway and close up shop because AI is expensive to provide, and you can you can only be a lost leader for so long before you run out of investors' funds, right? You just gotta close up shop and say, you know what, we can't afford to do this anymore. Others will be acquired and then shuttered, as Google has done many times, as many other tech companies have done many times. They buy the IP, they buy the staff, maybe.

Um, and they're like, Yeah, you know what, we didn't really want that company, we just wanted the people who were developing it. And in AI, there's a very high probability of this happening because there are a lot of companies that don't have a lot to them. They are a wrapper, a UI on top of somebody else's model. And that gets expensive. If an AI app or service or platform, whatever leaves the testing phase at your company and becomes part of how you do business, becomes part of your standard operating procedures, becomes part of your value proposition, you owe it to yourself and your organization to have alternatives on deck.

If Chat GPT, uh OpenAI runs out of money, or Google Gemini and Google says, yeah, we're not doing this anymore, it's really expensive, or Anthropex Clause is, yeah, we're out of money. If they all closed up shop today, what would you do for generative AI for large language models tomorrow? Suppose it's you know, Friday, they announce they're going out of business. What do you do Monday morning? If you've integrated generative AI into your standard operating procedures.

What do you do? If tools like DOLI or Mid Journey or Meta Image Generator or whatever went offline today, what would you use to create imagery tomorrow? There's a lot of options, but do you know what those options are? Have you tested them? Are you ready to go?

This is part and parcel of why local AI models and tools are so essential. Having models that you download and run on your own hardware is your insurance policy against the worst case scenarios. When you see a cool new AI service, asking yourself, hey, can I build that myself? Could I create my own version of that? That's a vital question to ask.

Everyone is fawning over Google's free notebook LM software, right? The reference software where you upload documents and it can create, among other things, study guides, frequently asked questions, and an audio podcast. Sit with simulated AI generated voices. People love this tool and it is a good tool. To be clear, it is a good tool.

It is useful, it is rag locked, which means retrieval augmented generation. It will not give you answers from if you don't provide the data for those answers. It's terrific. It's free. It's from Google.

What is the probability that Google says, wow, this is really expensive to operate? Maybe we're just going to turn it off. There's a lot of precedent for Google to do that. If Google did that, what would you use to replace it? If you love this tool, if you're using this tool, if like some of the folks I know, you're building a business around this tool, what would you use to replace it?

Do you know about the dozen open source projects that exist to replicate part or all of its functionality? Do you know how to use a tool like ChatGPT or Claude to code it yourself to make something that is so totally yours that no one can ever take it away because it's yours? Could you set that up? I am a big advocate of having a backup plan for any mission critical service or technology. And this has never been more true than an artificial intelligence.

As AI matures, as the market space matures and changes, you're gonna want to have those backups ready so that when the worst inevitably happens, because as the as the expression from World of Warcraft goes, uh no king rules forever, when the worst happens, you're ready and you don't miss a step. You're like, oh, chat GPT is gone. What are we gonna do today? I'll just open up anything LLM, turn on uh Llama 70B, and we're we're good to go. That's where you want to be.

You want to be at a point where if the the utility of your choice just shuts down, it's not a problem. You just keep on trucking. You you you build you execute your backup plans perfectly. That's gonna do it for today's episode. Thanks for tuning in.

Talk to you soon. If you enjoyed this video, please hit the like button. Subscribe to my channel if you haven't already, and if you want to know when new videos are available, hit the bell button to be notified as soon as new content is live.