Top 5 Common Mistakes Business Leaders have about AI
8/18/2024 | nmb@konfitech.com
Welcome to today’s insights in IT. We will go through some common AI mistakes we see many leaders in business make with artificial intelligence. We as a strategic IT sourcing partner for finding the right solutions and implementation partners have unique insights into what business leaders are thinking about AI mistakes. We want to share these today.
- It is all the same: AI, Machine Learning, Generative AI
These things are very similar, but with very different functional applications. Artificial narrow intelligence which is the AI we are mostly familiar with is only focused on doing specific tasks. Not considering anything else. One of the most famous examples of narrow artificial intelligence is Deep Blue. Which was a computer purposely built to only play chess, it could beat the best player in the world at the time Gary Kasparov. However, if tasked with something else Deep Blue would utterly fail. To summarize, narrow artificial intelligence has its sole purpose defined and it is extremely good at that one thing.
General Artificial intelligence on the other hand is not given a single task to exceed at. The goal with General Artificial intelligence is to think and act of its own and learn from it. So that the next decision is better than the last. (While on could argue, such as Yann LeCun on the Lex Fridman Podcast, that solving problems in a human reality could be defined as a narrow task. However, treating wide and narrow problems similarly can lead to significant AI mistakes.)
Machine Learning is a subset of artificial intelligence, which focuses on how machines or these algorithms that represent intelligence learns new problems. Such as supervised, unsupervised, reinforcement, and deep learning. Meaning that machine learning is the improvement mechanism inside of an artificial intelligence.
Generative Artificial Intelligence, such as ChatGPT, Gemini, and other chatbots are not something that is a uniquely new technology. It is just that the examples mentioned have made it work well. Generative AI is two machine learning algorithms that are designed to learn from each other. One is a generator that generates responses to inputs – like prompts from users. The other is a discriminator or a fact checker, which checks if what the generator is creating reliable output. Like factually true statements about the world.
- AI will solve all our problems.
AI is extremely exciting and seems like it is being applied to everything. However, it will not be able to deliver efficient solutions without data and proper data. One does not need to give giant amounts of data, but quality data. Data that has a high significance of indication on how to solve what you will give it as a problem. Poor data quality is among the common AI mistakes. There is no easy open-source plug in you can add to your stack. The AI will need proper training before deployment, and continuous training after to make sure that it learns.
- You must be an expert to get ROI.
While the machinery behind the technology and advanced algorithms demands a high level of expertise, there are so many low code options to build your own models that can generate returns. These prebuilt models can empower businesspeople to build their own models. Make sure to experiment a lot with AI, and do not fear trying because you’re not an expert. Understanding and avoiding AI mistakes can significantly improve your ROI.
- Implementing AI will make me lose my job.
When the car came and replaced the horse and carriage, do you think there was a reduction in need for work? In a short-term perspective it may have caused a shut down of horse farmers and carriage makers, the market economy shifted that more workforce and labour was needed in the auto industry. The same will happen with AI, we are in an open market world, where the AI will replace some traditional works, we still need to have humans in the loop for autonomous processes. Checking quality or enhancing output of single workers. AI will not saturate the market of demand for labour but rather shift the demand for labour to somewhere else. Being aware of potential AI mistakes, like improper implementation, helps ensure smoother transitions.
- You will need enormous amounts of data and resources to successfully deploy AI.
You don’t need terabytes of data, and it would not be smart either. An analogy of this could be if you have two identical cars, and one requires more fuel data for the engine than the other it would not be efficient. Do the same for your business, focus on fuel efficiency, and don’t waste time and resources building your own engine. Use prebuilt pieces like GPT-4 from OpenAI. No need to reinvent the wheel, use Azure OpenAI or AWS and feed them high significant data for the process you want to automate.
Want to learn more? Check out our page on Data & AI: https://www.konfitech.com/data-and-ai