AI became a mainstream, public-facing technology when OpenAI announced ChatGPT. That was the first time I used an AI tool, and it felt like magic. It provided surprisingly useful responses, and one of my favorite use cases was cooking. You can simply input a list of ingredients, and it will generate a great recipe—adjusting measurements, temperatures, and even unit conversions. I usually trust it because, well, I’m still alive! But if I had asked Google AI, I’d be more skeptical—especially after hearing that it once suggested putting glue in pizza sauce to keep the cheese from sliding.

Beyond personal use, I’ve seen developers in my team using ChatGPT and GitHub Copilot to assist with coding queries. However, interestingly, some developers avoid Copilot altogether and don’t talk much about it.

The Power of Deep Learning

So, how does ChatGPT actually work? At its core, ChatGPT is trained by analyzing vast amounts of text from the web, recognizing patterns, and predicting what the next word should be. Essentially, it’s learning to respond by mimicking the structure of human language.

AI’s Expanding Role

AI is now integrated into many aspects of daily life. Writers use it to generate content, developers rely on it for coding assistance, and—surprisingly—some people even use it as a friend. There have been real cases of people falling in love with chatbots like Replika, and some have even gone as far as marrying them. The CEO of Replika was even invited to chatbot weddings—this sounds straight out of a sci-fi movie, but it’s happening.

The most recent shift in AI development is its transition from simple chatbots to autonomous agents—AI that can take actions without human intervention. In 2022, we saw early autonomous agents, which evolved in 2023 into Large Action Models (LAMs)—AI that not only processes language but can also perform tasks. In 2024, these LAMs have been integrated into software applications and are now called Agents. A great example is Salesforce’s AgentForce, which acts as a digital workforce for businesses.

A recent Super Bowl ad even showcased a traveler struggling to find his gate at the airport, only to realize he hadn’t used AgentForce. This technology is already being deployed in flight booking systems, and according to the latest reports, direct consumer access is planned for Q2 2025.

Challenges and the Road Ahead

Despite AI’s rapid growth, companies like OpenAI are facing challenges. They are dealing with copyright infringement lawsuits over the use of copyrighted materials for training their models. Another major issue is profitability—OpenAI is reportedly $3 billion in debt, and not many consumers are willing to pay $200 per month for premium AI access. Meanwhile, new competitors like DeepSeek are emerging, challenging the traditional way AI models are trained using vast amounts of GPUs.

Apple has also entered the AI space with Apple Intelligence, but Siri is still widely criticized for failing at basic tasks, like finding flight details in emails.

Final Thoughts

While some believe AI is overhyped and could replace many remote jobs in the next five years, it is far from useless. AI will continue to evolve, offering both incredible opportunities and critical lessons along the way. The question is: How will we choose to use it?