Artificial intelligence is now everywhere, thanks in large part to the massive viral success of generative AI apps. In particular, ChatGPT has become the fastest growing user base app of all time, sparking a lot of discussion about what this technology can do and what it might do in the future.

Of course, as with all this excitement, there is a fair amount of misunderstanding and misinformation floating around. And with so much money thrown in, it’s not uncommon to see some pretty grandiose claims about what this can do or how smart it actually is. So, I’ve put together five of the most common misconceptions I regularly encounter when it comes to AI and machine learning today.

AI is intelligent

Intelligence is an inherent property of living things that enables us to learn, communicate, understand, empathize and make decisions. AI is an attempt to use machines to simulate it or produce similar results, but it is still just a mechanical simulation that seems to produce some of the same results as natural intelligence.

When we talk about AI today, especially when we refer to its use in businesses and online applications, we usually refer to machine learning. It’s a type of AI that uses algorithms trained on data to get better and better at performing specific tasks. This could include playing games, recognizing images, translating languages, driving a car, answering questions, and many other tasks that machines can perform if given the right information. there is. These are all examples of so-called specialized AI, as they only perform the specific tasks they are trained to do.

AI that is truly as intelligent as humans (but much faster) and capable of performing all kinds of tasks that humans can do is known as General Purpose AI. But we still have a long way to go to achieve this.

AI is expensive and difficult to implement

Traditionally, AI has been expensive and available only to large or well-funded research organizations. This is because collecting, cleansing, and storing all the data needed to train machines to make decisions and provide the computing power they need to process the data is costly. For example, the cost of training ChatGPT is estimated to be around $5 million, but the cost of training larger models expected to emerge in the future will be much higher.

However, most companies and organizations that benefit from the use of AI will not need to train their own model, let alone one the size of ChatGPT. The availability of AI services through a cloud platform means that they can be accessed and used at low cost without specialized knowledge or technical skills. This means that, unlike five years ago, AI learning and decision-making will now be available not only to AI and data experts, but also to domain experts, creating a “democratizing” effect and making it more accessible to more companies. and organizations can now reap its benefits.

AI is about to take over human jobs

It will probably be inevitable that some human jobs will simply be replaced by machines that can perform them faster, more accurately and more cost-effectively. This is true even after other major industrial revolutions such as mechanization, electrification, digitization and now automation.

But it will also create more jobs, and will likely be better paid and more rewarding jobs than the jobs lost. In 2020, the World Economic Forum released a report stating that by 2025, automation will replace 85 million jobs in manufacturing, insurance underwriting, customer service, data entry and long-haul truck driving, while creating 97 million new opportunities. said to be born Created.

AI is neutral and impartial

Because AI was born from machines, it makes sense for people who aren’t familiar with how AI works to assume that AI is always fair, balanced, and unbiased. Unfortunately this is not true. AI algorithms are trained on data, so they can’t “know” anything, and this data is often created or managed by humans. This means that it is almost inevitable that human bias will creep in and affect the output of the algorithm, especially for large datasets. AI performance is determined by the data used to train it. A common warning about any computer system is that “garbage in = garbage out.” Academics in the AI ​​research field have warned for some time that in a world where computers can make decisions for humans, bias is one of the main dangers we must be aware of. and a lot of research has been done on the ethics of AI. We are interested in ensuring that the risk of bias in our data is eliminated or minimized.

AI will conquer the world and enslave humans

This has been the premise of many popular and highly entertaining science fiction novels, such as the Matrix and Terminator series. But the idea may not be all that far-fetched. At least some very famous and intelligent people like Elon Musk, Dr. Stephen Hawking and Bill Gates say this might be the case.

The absolute truth is that no one knows where AI will end up, and much depends on how we humans develop, implement, and regulate it. For this reason, ethics and oversight are very important factors in the work of understanding and creating AI today. Today’s state-of-the-art AI, such as ChatGPT, does not pose an existential threat to us as a species because it has no ability to harm us or act in ways other than the way it was programmed. . It means helping us. Basic tasks with information. Nor do they have the instinct of self-preservation that motivates machines against humans in science fiction stories. It’s simply because it wasn’t programmed.