Your vehicle is dirty. Should you walk or drive to the carwash?

AI Generated Image, Obviously

AI Generated Image, Obviously

It’s one of life’s great dilemmas. You’ve got a dirty car and a beautiful day. The nearest carwash is less than a block away. Should you walk or drive? 

This delightfully absurd scenario is making its rounds online. It’s the kind of brilliant question only a human could invent as a test of the reasoning capabilities of AI, and many of the AI tools failed. I love this test because it reveals three insights leaders need to know as they begin to adopt AI.

1. AI can make mistakes.

Everyone knows by now that AI can make errors. While I’m having a harder and harder time finding mistakes as I get better at using AI, I always use it assuming it will get some things wrong, and if it’s important, I always verify accuracy. Most of the mistakes I see happen when I overload it with too much information. I either check AI’s answers against a source I trust, or I just use my own judgement. In this case, I will not be walking to the carwash.

2. There’s a difference between knowledge and reason.

AI models could probably score better on any standardized test than you. I know it could score better than me. It can ace the SAT. It can easily pass the bar exam (the lawyer test), the MCAT (the doctor test), and the CPA, which is widely regarded as the hardest of the three. Much respect to my accountant friends who locked-in on that one in. 

AI is so good at these tests, researchers had to invent the hardest test ever created: Humanity’s Last Exam

These exams generally test knowledge, but they don’t test the reason necessary to answer questions that have never … ever been asked before. I think AI failed the carwash test because it’s a simple, but completely novel question.

All of the AI systems you use have a whole lot of knowledge under the hood, but the free versions most people use do not have strong reasoning capabilities. Free AI can tell you how to wash your car and where the nearest carwash is, but it might, unreasonably, suggest you walk to get there. 

AI’s reasoning capabilities have only started to grow quickly over the past few months, and if you’re using the free version of ChatGPT, Claude, or Gemini, you’re not getting the most reasonable AI.

3. If you’re using AI for work, you need a professional account.

Professionals need professional tools, and there are so many reasons to upgrade. Most importantly, data security. Most professionals today are using insecure free AI accounts because they know how useful it is, but they don’t understand the risks. The info uploaded to these chatbots are not only getting sent to the companies running them. The data is also being used to “train” the models, which could, theoretically, be found by anyone else using the same AI tool. Even if you’re paying for a personal account, your data is likely training the model. This exposes organizations to ethical and legal risks, and that can erode your most valuable asset: trust.

The best enterprise-grade AI tools are designed to keep your data secure. They do NOT train the model. Organizations should use AI to avoid getting left behind, but using personal accounts carries huge risks. The good news is that enterprise-grade AI is surprisingly inexpensive ($20-30/month) and relatively easy to deploy. That alone is a good reason.

The other reason to invest in pro tools is (sorry) reason. 

If you’re using free AI, you’re using unreasonable AI. Here’s the result I got in one of my secure AI accounts using advanced reasoning.

Almost all organizations are using AI to boost efficiency. The leaders who are using AI most effectively today are using it to truly innovate. Innovation is about way more than just using the latest technology. It’s about solving new problems in new ways. Asking questions that have never been asked. That requires human expertise and judgement, and it can be massively supported by professional-grade AI tools with their rapidly expanded knowledge and reasoning capabilities. 

If you’re interested in learning more about how you can get the most out of AI for your organization, I am hosting The AI Onramp in Carmel, California on April 23. It is a small hands-on workshop for leaders who know they need to incorporate AI into their operations, but are not sure where to start. You’ll walk away with the knowledge and tools to deploy AI securely and effectively. 

AI did not write this. I did, which is why it’s too long. To thank you for getting through it all. Here’s a coupon code for $40 off the course: IREADTHEWHOLETHING

Learn more and register for the class here: gapjump.ai/onramp

Kevin Brookhouser

GapJump AI was founded by Kevin Brookhouser, a Google Certified Generative AI Leader. With over two decades in innovative technology adoption, Kevin offers a pragmatic, hands-on approach, helping teams deploy powerful technologies to streamline work, automate workflows, and find joy and strategic advantage in using AI. Our approach ensures that every client benefits from a unique fusion of visionary thought leadership and practical, implementable strategies.

https://gapjump.ai
Next
Next

How to train your ai